Culturing pancreatic islets in microfluidic flow enhances morphology of the associated endothelial cells.

Abstract

Pancreatic islets are heavily vascularized in vivo with each insulin secreting beta-cell associated with at least one endothelial cell (EC). This structure is maintained immediately post-isolation; however, in culture the ECs slowly deteriorate, losing density and branched morphology. We postulate that this deterioration occurs in the absence of blood flow due to limited diffusion of media inside the tissue. To improve exchange of media inside the tissue, we created a microfluidic device to culture islets in a range of flow-rates. Culturing the islets from C57BL6 mice in this device with media flowing between 1 and 7 ml/24 hr resulted in twice the EC-density and -connected length compared to classically cultured islets. Media containing fluorescent dextran reached the center of islets in the device in a flow-rate-dependant manner consistent with improved penetration. We also observed deterioration of EC morphology using serum free media that was rescued by addition of bovine serum albumin, a known anti-apoptotic signal with limited diffusion in tissue. We further examined the effect of flow on beta-cells showing dampened glucose-stimulated Ca(2+)-response from cells at the periphery of the islet where fluid shear-stress is greatest. However, we observed normal two-photon NAD(P)H response and insulin secretion from the remainder of the islet. These data reveal the deterioration of islet EC-morphology is in part due to restricted diffusion of serum albumin within the tissue. These data further reveal microfluidic devices as unique platforms to optimize islet culture by introducing intercellular flow to overcome the restricted diffusion of media components.

Full text

Page 1

Culturing Pancreatic Islets in Microfluidic Flow Enhances
Morphology of the Associated Endothelial Cells
Krishana S. Sankar2., Brenda J. Green1., Alana R. Crocker1, Jocelyne E. Verity3, Svetlana M.
Altamentova3, Jonathan V. Rocheleau1,2,3*
1Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada, 2Department of Physiology, University of Toronto, Toronto,
Ontario, Canada, 3Toronto General Research Institute, University Health Network, Toronto, Ontario, Canada
Abstract
Pancreatic islets are heavily vascularized in vivo with each insulin secreting beta-cell associated with at least one endothelial
cell (EC). This structure is maintained immediately post-isolation; however, in culture the ECs slowly deteriorate, losing
density and branched morphology. We postulate that this deterioration occurs in the absence of blood flow due to limited
diffusion of media inside the tissue. To improve exchange of media inside the tissue, we created a microfluidic device to
culture islets in a range of flow-rates. Culturing the islets from C57BL6 mice in this device with media flowing between 1 and
7 ml/24 hr resulted in twice the EC-density and -connected length compared to classically cultured islets. Media containing
fluorescent dextran reached the center of islets in the device in a flow-rate-dependant manner consistent with improved
penetration. We also observed deterioration of EC morphology using serum free media that was rescued by addition of
bovine serum albumin, a known anti-apoptotic signal with limited diffusion in tissue. We further examined the effect of flow
on beta-cells showing dampened glucose-stimulated Ca2+-response from cells at the periphery of the islet where fluid
shear-stress is greatest. However, we observed normal two-photon NAD(P)H response and insulin secretion from the
remainder of the islet. These data reveal the deterioration of islet EC-morphology is in part due to restricted diffusion of
serum albumin within the tissue. These data further reveal microfluidic devices as unique platforms to optimize islet culture
by introducing intercellular flow to overcome the restricted diffusion of media components.
Citation: Sankar KS, Green BJ, Crocker AR, Verity JE, Altamentova SM, et al. (2011) Culturing Pancreatic Islets in Microfluidic Flow Enhances Morphology of the
Associated Endothelial Cells. PLoS ONE 6(9): e24904. doi:10.1371/journal.pone.0024904
Editor: Kathrin Maedler, University of Bremen, Germany
Received June 6, 2011; Accepted August 19, 2011; Published September 22, 2011
Copyright: ? 2011 Sankar et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants to JVR from the Heart and Stroke Foundation of Ontario (NA-6591), Canadian Institutes of Health Research (NGM-
94489), Canadian Foundation for Innovation and Ontario Research Fund (18301). The funders had no role in study design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: Jon.Rocheleau@utoronto.ca
. These authors contributed equally to this work.
Introduction
Pancreatic islets are highly vascularized micro-organs designed
to effectively sense glucose and secrete insulin directly into the
bloodstream. The characteristically dense vasculature of islets is
comprised of endothelial cells (ECs) that are thin and highly
fenestrated, and therefore directly connect the insulin secreting
beta-cells to the blood [1]. Islets maintain this vasculature once
harvested for ex vivo studies. However, it is well established that
during ex vivo culture, the density of intra-islet ECs decreases
approximately 50% in the first day and is almost completely gone
by day four [2]. This loss of ECs during culture severely limits
long-term study of the interaction between beta-cells and ECs in
the tissue, and is directly relevant to the treatment of type 1
diabetes due to the critical role of the ECs from donor islets in the
revascularization of this tissue during transplantation [2,3,4]. The
final reconnecting vessels are formed by EC from the donor and
recipient, which makes it vital to maintain these cells prior to
transplantation [2,4]. Our goal was therefore to delay the loss of
islet-ECs during culture.
Islets isolated from the pancreas are severed from the
vasculature and once in culture rely on diffusion for nutrient
supply and by-product removal. We postulate that the ECs inside
islets in culture die in part due to an absence of blood flow to
provide media exchange. It is well established that ECs in culture
undergo apoptosis when serum is reduced in the culture medium
[5]. However, in isolated islets, serum proteins such as albumin
will not diffuse significantly into the tissue due to large
hydrodynamic radius and transient binding properties [6].
Alternatively, any nutrients that are consumed (eg. O2, glucose)
or products that are secreted (eg. insulin, lactate) will form
gradients that may adversely affect the ECs. Overall, limited
media exchange during islet culture can have detrimental effects
on the ECs within the islet. It should also be noted that flow-
induced shear can have a positive effect on ECs, acting as a
biomechanical stimulus and anti-apoptotic signal [7,8,9]. Con-
versely, flow can negatively affect epithelial cells [10], and may
therefore negatively impact glucose-stimulated insulin secretion.
To test the effect of fluid flow on maintenance of islet ECs, we
developed a hemodynamic environment using a custom-designed
microfluidic device. Our devices place islets in laminar flow,
similar to devices previously used to monitor bulk islet O2-
consumption rate using perfusion column chambers [11,12,13]. In
contrast to perfusion column designs, microfluidic devices require
PLoS ONE | www.plosone.org1September 2011 | Volume 6 | Issue 9 | e24904

Page 2

higher pressure-drops to induce flow [14] and are completely
compatible with fluorescence microscopy [14,15]. We aimed to
use these devices combined with fluorescence microscopy to treat
islets with media flow and image the effect on ECs and beta-cells.
Materials and Methods
Microfluidic Device Fabrication
Devices were fabricated using elastomer polydimethylsiloxane
(PDMS) (Dow-Corning) as described previously [16,17]. The
cured mould was irreversibly bonded to a 24650-mm coverslip
(VWR Scientific) by oxygen plasma treatment (Harrick Scientific,
Ossining, NY) and Tygon tubing was inserted directly into the
cored-out port holes.
Microfluidic Device Design
The microfluidic device design and process are described in
detail in the Supplemental Material (Text S1 & Fig. S1, S2, and
S3). Briefly, islets were brought into 125 mm tall 6300 mm wide
microfluidic channels through inlet port tubes. Islets moved along
these channels by media flow until they reached a dam wall. This
dam structure was 25 mm tall, 1,800 mm wide, and 0.13 cm long,
and allowed solution to flow past but blocked the movement of
islets. Microfluidic devices provide only non-turbulent or laminar
flow at the flow rates used in these studies [14]. Two main designs
were used for these studies: a three-channel microfluidic device
and an independent-channel microfluidic device (Fig. S1). The
three-channel device placed islets in three different flow-rates
based on varied channel lengths and by using a single syringe
pump. The independent-channel device provided a single flow-
rate in parallel channels to facilitate subsequent live cell imaging or
effluent collection.
Ethics Statement
Animal procedures were approved by the Animal Care
Committee of the University Health Network, Toronto, Ontario,
Canada in accordance with the policies and guidelines of the
Canadian Council on Animal Care (Animal Use Protocol #1531).
Pancreatic Islet Isolation and Treatment
Pancreatic islets were isolated from 8- to 12- week-old C57BL6
male mice by using collagenase digestion (Roche) [18,19]. Islets
were cultured in RPMI medium 1640 containing 11 mM glucose,
10% FBS, 5 U/ml penicillin-streptomycin, and 20 mM HEPES.
Control islets were incubated in a non-treated culture dishes in a
humidified incubator at 37uC and 5% CO2. Device-treated islets
were loaded into the microfluidic devices shortly after isolation and
incubated in a custom-built desk-top incubator described in detail
in the Supplemental Material (Fig. S2). Briefly, the microfluidic
device was submerged in a stirred water bath at 37uC with flow
driven by a syringe pump. The reservoir media was also
submerged in a separate water-bath maintained just above 37uC
to reduce formation of air bubbles in the microfluidic channel. A
cap of mineral oil was placed on top of the media in the reservoir
to reduce evaporation and drift in pH.
Immunofluorescent Detection
Control and device-treated islets were labelled within the
microfluidic device. Islets were fixed (2% PFA, PBS, 1 hr, 100 mL/
hr), blocked (PBS, 0.1% TritonX-100, 4uC, 10% Normal Goat
Serum, 4 hr, 15 mL/hr), incubated with primary antibody (PBS,
0.1% TritonX-100, 4uC, 1% Normal Goat Serum, 1:100 dilution
rat anti-mouse PECAM-1, 3 hr, 3 mL/hr), rinsed (PBS, 0.1%
TritonX-100, 4uC, 10% Normal Goat Serum, 1 hr, 15 mL/hr)
and incubated with 1:500 goat anti-rat Alexa 633 (PBS, 0.1%
TritonX-100, 4uC, 10% Normal Goat Serum, 3 hr, 15 mL/hr).
Islets were imaged using a 2060.50 NA objective lens in 3 mm
steps for 16–20 slices (dependant on the size of the islet) on an
Olympus FluoView 300 microscope.
Images were analyzed using ImageJ 1.41o (Wayne Rasband,
NIH, USA). EC fractional area was measured by calculating the
area of PECAM-1 labelling divided by the total area of the islet (6
slices/islet). Total connected length was determined using Simple
Neurite Tracer throughout the entire stack of images. PECAM-1
labelled regions were tracked through the stack and the resulting
summation was normalized to the circumference of the islet. The
average of individual islets was used for statistical analysis. A two
tail, unpaired, two sample student’s t-test was used and significance
was determined with p,0.05 (*) and p,0.01 (**).
Islet intracellular calcium
Islets were preincubated with either 4 mM of Fura-2 or Fluo-4 in
imaging buffer supplemented with 2 mM glucose for 1–2 hr prior
to imaging. Fura-2 images were collected using the 4061.3 NA oil
immersion lens of an Eclipse Ti-S inverted microscope with
340 nm and 380 nm excitation and 510640 nm emission filter
(PTI EasyRatioPro System, USA). Fluo-4 images were collected
on an LSM710 Microscope, 2060.80 NA objective lens, 488-nm
laser line, and 493–630 nm emission. The device and its reagents
were placed on the microscope stage inside a 37uC stage-top
incubator (Okolabs).
Two-photon NAD(P)H Imaging
NAD(P)H imaging was done as previously described using a
4061.3 NA oil immersion objective lens and internal detector
(385–550 nm) of a LSM710 microscope (Zeiss) [20,21]. The
Ti:Saph laser was tuned to 705 nm and attenuated to ,3 mW at
the sample (Coherent).
Insulin Secretion
Islets in a single channel were stimulated with 200 mL/hr flow
and 37uC. Islets were subsequently treated for one hour with
2 mM and 11 mM glucose, each providing 200 mL of effluent.
The flow rate was increased to 200 mL/min and 1% TritonX-100
introduced to release the total insulin content. The effluents were
diluted 506(2 and 11 mM glucose) and 2006(1% TritonX-100)
prior to insulin ELISA detection.
Results
A Microfluidic Device to Mimic in vivo Hemodynamics
Islets placed in culture no longer experience blood flow. To
introduce intracellular flow, we developed a microfluidic device for
long-term culture (Fig. 1 & S1). Similar to our previous
microfluidic device designs, islets were trapped at a channel dam
(Fig. 1A & B) [22]. The dam was fashioned by dropping the
channel height from 125 to 25 mm, effectively preventing islets
from moving further down the channel while still allowing fluid to
flow. The device shown had three channels of varying length with
individual inputs and one common output (Fig. 1C). This design
allowed us to culture islets in three different flow-rates using a
single syringe pump.
To measure the effect of time (0, 24, and 48 hr) and flow rate
(no flow and flow between 1 and 7 ml/24 hr) on EC morphology,
islets cultured in dishes (no flow) or in the three channel
microfluidic device were assessed by immunofluorescence imaging
of PECAM-1 (Fig. 2). These flow rates were chosen to
approximate a physiologically relevant range of shear stress (1–
Pancreatic Islet Culture in a Microfluidic Device
PLoS ONE | www.plosone.org2September 2011 | Volume 6 | Issue 9 | e24904

Page 3

14 dyn/cm2) based on the device geometry [23]. In comparison,
individual rat islets in vivo receive ,0.03 ml/24 hr of blood flow
[24,25]. Islets fixed immediately post-isolation were heavily
vascularised with EC that branched throughout the islet
(Fig. 2A). The morphology of the ECs in control dish-cultured
islets changed significantly in 24 hr, appearing to coalesce
(Fig. 2B) [2]. In contrast, the ECs of islets treated in the device
with flow had greater density and vascular structure throughout
the islet compared to control (Fig. 2C). To quantify these changes
in morphology, we measured the average relative length and
fractional area of the ECs in a number of islets (Fig. 2D–F). After
24 and 48 hr of culture, the length and area was significantly
decreased in both control and device-treated islets compared to
freshly isolated islets (Fig. 2D & E, Day0 vs. Cntl(0)). However,
we observed significantly denser and longer ECs in islets cultured
in the device at each time point compared to control islets. We also
measured a dramatic decline in EC density at extreme flow rates
(0.9860.08 and 1.1660.10 fold-EC fractional area at 0.3 and
30 ml/24 hr compared to control, respectively) consistent with an
envelope of optimal flow rate. Even though microfluidic devices
are inherently permeable to oxygen [26] and we are flowing at
rates normally experienced in vivo (0.03 ml/24 hr/islet) [24,25],
we may be limiting nutrient supply (glucose, O2) at 0.3 ml/24 hr.
We may also be inducing shear damage to the tissue at 30 ml/
24 hr. To explore the effect of islet size on EC morphology, we re-
examined the data shown in Figure 2E based on islet diameter
(Fig. 2F). Grouping the data from large and small islets (defined as
above and below the median islet diameter (178 mm), respectively),
we observed no difference in EC fractional area of freshly isolated
islets (Day0). After 48 hrs of culture in the absence of flow, large
islets had significantly lower EC fractional area than small islets.
These data are consistent with a more significant loss of EC
morphology in large islets. In contrast, culturing islets in flow (1–
7 ml/24 hr) resulted in no difference in EC fractional area
between large and small islets, consistent with flow-rate rather
than islet size determining EC morphology. Overall, these data
show that islets treated to media flow in the device (1–7 ml/24 hr)
maintained more than twice their EC density and connected
length when compared to islets traditionally cultured in the
absence of flow.
The Mechanism of Enhanced Morphology
To determine where media gained access in the tissue, we
cultured islets in flow (3 ml/24 hr) for 24 hr and subsequently
imaged media supplemented with fluorescent dextran (Fig. 3A &
B). These and subsequent studies were done using a single channel
microfluidic device with on-chip reservoir, which allowed us to
quickly exchange the media (Fig. S3). Fluorescent dextran does
not enter living cells, and therefore reports on the volume of media
in the intercellular space. We observed media surrounding each
individual cell as well as filling connected pathways that had
similar shape to ECs (Fig. 3A & B). By fixing and staining these
same islets for PECAM-1, we further confirmed that the
connected pathways corresponded to the ECs of the islet
(Fig. 3A & C, arrows). These data show that media can access
the cells in the center of the islet; however, our PECAM-1 data
(Fig. 2) suggested that this access is limited at slower flow rates
(0.3 ml/24 hr). To examine the flow-rate dependence of media
penetration, we measured the time to fill the intercellular space at
Figure 1. A microfluidic device to treat pancreatic islets with
laminar flow. (A) A schematic of a microfluidic channel is shown to
demonstrate how pancreatic islets are held in laminar flow. Pancreatic
islets (oval) are brought into microfluidic channels (outlined) and
trapped when the channel height drops from 125 to 25 mm. This stops
the islet from moving down the channel while allowing fluid to flow
past the islet (arrows). (B) A representative image of islets held at the
‘pea pod’ shaped dam in a channel. In this experiment, the islets were
cultured for 24 hr prior to loading. The islets are sitting in the tall
channel (125 mm) against the dam wall to the right. Flow in the device
is past the islets from left to right. (C) An image of the three channel
microfluidic device used for varying fluid flow-rates past islets. The
device shown is filled with bromophenol blue solution to highlight the
channels and tubing. The device has three individual inputs (tubing)
with a common output (not shown). The three microfluidic channels
(125 mm) each have a dam for islet retention (spherical region, 25 mm)
followed by varying lengths of output channels (125 mm). These varied
channel lengths allowed media flow to be controlled with one syringe
pump, and media flow at three different flow-rates relative to length.
Flow rates were ultimately determined by measuring the height of the
input reservoirs before and after incubation.
doi:10.1371/journal.pone.0024904.g001
Pancreatic Islet Culture in a Microfluidic Device
PLoS ONE | www.plosone.org3September 2011 | Volume 6 | Issue 9 | e24904

Page 4

Figure 2. The effect of fluid-flow on EC area and connected length. Pancreatic islets were cultured in either the microfluidic device with flow
(device-treated) or in non-flowing media in an incubator (control) followed by anti-PECAM-1 immunolabelling. (A) A representative Day 0, freshly
isolated islet shows heavy vascularization of the tissue prior to culture. (Scale bar 125 mm) (B) A representative control islet cultured in no-flow for
24 hr displaying characteristic coalescence of EC. (Scale bar 100 mm) (C) A representative islet cultured for 24 hr in a microfluidic device displaying
increased EC density and connected length compared to control islets (Cntl(0)). (Scale bar 150 mm) (D) The total EC connected length relative to the
circumference (EC Length L/c) of freshly isolated islets (hatched bar), control islets (Cntl(0)) at 24 (black bars) and 48 hr (open bars), and device-treated
islets exposed to fluid flow (1–3, 3–5, and 5–7 ml/24 hr). (E) The EC area relative to islet area (EC percent area) of freshly isolated islets (hatched bar),
control islets (Cntl(0)) at (24 (black bars) and 48 hr (open bars), and device-treated islets exposed to fluid flow (1–3, 3–5, and 5–7 ml/24 hr). The EC
length and fractional areas were measured from the islets isolated from at least 3 different mice on independent days. The data shown is the mean 6
sem for the following number of pooled islets: 43 (Day0), 14 (Control, 24 hr), 12 (24 hr, 1–3 ml/24 hr), 19 (24 hr, 3–5 ml/24 hr), 10 (24 hr, 5–7 ml/
Pancreatic Islet Culture in a Microfluidic Device
PLoS ONE | www.plosone.org4 September 2011 | Volume 6 | Issue 9 | e24904

Page 5

different flow rates (Fig. 3D & E). A representative image series is
shown of a single islet as fluorescent dextran is introduced at a flow
rate of 3 ml/24 hr from right to left. In this setup, we measured
the time between dye front arrival (as dye surrounds the islet) and
dye filling the intercellular space. The top panel of Fig. 3D is the
image frame immediately prior to the arrival of dye at the islet
(0 s). At this time the media and islet are both dark reflecting the
absence of dye. At 2 s the dye completely filled the channel, but
the islet remained dark indicating the dye had not significantly
penetrated the tissue. The visibly dark flow tail streaming behind
the islet (at left) occurs due to non-fluorescent media being pushed
from the intercellular space. By 30 s, the dye completely filled the
intercellular space of the islet and the clear flow stream
disappeared. Data collected from multiple islets at multiple flow
rates shows that the time for dye to penetrate the islet is
exponentially dependent on flow rate (Fig. 3E). These data
suggest that at the flow rates previously shown to enhance EC
morphology (3–6 ml/24 hr), media exchange occurs in less than
30 s. In contrast, at flow rates that did not enhance EC
morphology (0.3 ml/24 hr) media exchange was a least 56slower
(.150 s). Importantly, this graph is asymptotic as it approaches
0 ml/24 hr, indicating that in normal static culture the exchange
rate due to diffusion is likely much slower (min - hrs). We also re-
grouped the data based on islet size, but observed no differences in
dye penetration at any of the flow rates between large and small
islets. This result was consistent with our previous data showing a
24 hr), 16 (Control, 48 hr), 10 (48 hr, 1–3 ml/24 hr), 7 (48 hr, 3–5 ml/24 hr), and 7 (48 hr, 5–7 ml/24 hr). (F) The EC fractional area for the data shown
in Fig. 2E re-grouped into large and small islets based on the median diameter (178 mm). Data shown are the mean 6 sem for the following number
of pooled islets: 21 (Large Islets, Day0), 22 (Small Islets, Day(0)), 9 (Large Islets, Cntl(0)), 8 (Small Islets, Cntl(0)), 12 (Large Islet, 1–7 ml/24 hr), and 12
(Small Islets, 1–7 ml/24 hr). The * and ** indicate p,0.05 and p,0.01 compared to control (Cntl(0)) or where otherwise indicated.
doi:10.1371/journal.pone.0024904.g002
Figure 3. Media access in a microfluidic device measured using fluorescent dextran. To explore the nature of the fluid flow inside treated
islets, real-time fluorescent imaging was applied to 24 hr device-treated islets (3 ml/24 hr). (A) We flowed fluorescent dextran (3,000 MW) into the
microfluidic device and imaged penetration of this dye into the tissue. The dye fills the channel and intercellular spaces of the islet. Connected (path-
like) regions that resemble islet-EC morphology (arrow) were immediately evident (black scale bar=40 mm). (B) A magnified image shows that
fluorescent dextran surrounds each cell in the islet (Scale bar=15 mm). (C) These same islets were immunofluorescently labelled for ECs using PECAM-
1 immunofluorescence. The connected regions identified by fluorescent dextrans co-labelled as ECs (compare arrows in A and C) (Scale bar=40 mm).
(D) To measure the kinetics of fluorescent dextran exchange with the intercellular space, we examined the arrival of the dye as it was introduced into
the channel. A typical time series is shown with media flowing 3 ml/24 hr from the right to left. The images shown were acquired at 0, 12, and 101 s
after the arrival of fluorescent dextran to the islet until it eventually fills both the channel and intercellular space. In the absence of dye, the channel
and islet are dark (0 s). At 12 s after the dye arrival, the channel had filled with dye except for a tail of non-labelled solution pushed from the
unlabelled islet. At 101 s after the dye arrival, the intercellular spaces were clearly visible and the dye completely filled the channel. (E) To determine
the dependence of media exchange on flow rate, we measured the time required to penetrate the islet at various flow rates. These data were
collected from the islets of 4 separate mice on different experimental days. The data is plotted as mean clearance time 6 sem for the pooled islet
data. (F) Islets were cultured in non-flowing media for less than 24 hr prior to imaging. Fluorescent dextran images were collected in a sequence
similar to (D), but with spatial resolution similar to (A). This imaging sequence allowed us to measure the intensity of fluorescent dextran at EC
structures (EC, black line) or neighbouring intercellular spaces (Intercellular, grey line). The arrow indicates the arrival of fluorescent dextran at the
islets. The fractional intensity was determined from the intensity at time zero and the maximal intensity reached at each region at .100 s. Data are
shown from the average 6 sem for 90 neighbouring pairs of regions from 7 different islets. No difference in arrival rate was measured between EC
and intercellular spaces indicating no preference for flow through the lumen of the ECs.
doi:10.1371/journal.pone.0024904.g003
Pancreatic Islet Culture in a Microfluidic Device
PLoS ONE | www.plosone.org5September 2011 | Volume 6 | Issue 9 | e24904