In situ formation, manipulation, and imaging of droplet-encapsulated fibrin networks.
ABSTRACT The protein fibrin plays a principal role in blood clotting and forms robust three dimensional networks. Here, microfluidic devices have been tailored to strategically generate and study these bionetworks by confinement in nanoliter volumes. The required protein components are initially encapsulated in separate droplets, which are subsequently merged by electrocoalescence. Next, distinct droplet microenvironments are created as the merged droplets experience one of two conditions: either they traverse a microfluidic pathway continuously, or they "park" to fully evolve an isotropic network before experiencing controlled deformations. High resolution fluorescence microscopy is used to image the fibrin networks in the microchannels. Aggregation (i.e."clotting") is significantly affected by the complicated flow fields in moving droplets. In stopped-flow conditions, an isotropic droplet-spanning network forms after a suitable ripening time. Subsequent network deformation, induced by the geometric structure of the microfluidic channel, is found to be elastic at low rates of deformation. A shape transition is identified for droplets experiencing rates of deformation higher than an identified threshold value. In this condition, significant densification of protein within the droplet due to hydrodynamic forces is observed. These results demonstrate that flow fields considerably affect fibrin in different circumstances exquisitely controlled using microfluidic tools.
- [show abstract] [hide abstract]
ABSTRACT: Severe infection and inflammation almost invariably lead to hemostatic abnormalities, ranging from insignificant laboratory changes to severe disseminated intravascular coagulation (DIC). Systemic inflammation results in activation of coagulation, due to tissue factor-mediated thrombin generation, downregulation of physiological anticoagulant mechanisms, and inhibition of fibrinolysis. Pro-inflammatory cytokines play a central role in the differential effects on the coagulation and fibrinolysis pathways. Vice-versa, activation of the coagulation system may importantly affect inflammatory responses by direct and indirect mechanisms. Apart from the general coagulation response to inflammation associated with severe infection, specific infections may cause distinct features, such as hemorrhagic fever or thrombotic microangiopathy. The relevance of the cross-talk between inflammation and coagulation is underlined by the promising results in the treatment of severe systemic infection with modulators of coagulation and inflammation.Cardiovascular Research 11/2003; 60(1):26-39. · 5.94 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: In agreement with earlier observations that the angular dependence of light scattering by fibrin gels obeys the theory for light scattering by very long and thin rigid rodlike particles (intensity proportional to the square of half the scattering angle), we find that the turbidity, tau, of the less opaque gels varies as the inverse third power of the wavelength, lambda. Mass-length ratios of the fibers calculated from these two measurements closely agree. For fibrin gels containing fibers with a very high mass-length ratio (of which we had not been able to obtain interpretable scattering data), the turbidity is found not quite to vary as 1/lambda3. For these opaque gels, the fiber diameter is no longer negligible with respect to the wavelength. It is shown how the radius of gyration of the fiber cross section (and therefore the radius of cylindrical fibers) can be obtained from the ratio of slope and intercept of a plot of 1/tau lambda3 vs. 1/lambra2. The square of the radius of the fibers is found to be proportional to the mass-length ratio. The density of the fibers is calculated to be 0.28. This corresponds to a ratio of fiber volume to volume of protein contained in the fiber of 5.0.Macromolecules 01/1978; 11(1):46-50. · 5.52 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: The concentration dependence of the structure of fibrin gels, formed following fibrinogen activation by thrombin at a constant molar ratio, was investigated by means of elastic light scattering techniques. The scattered intensity distributions were measured in absolute units over a wave-vector range q of about three decades ( approximately 3x10(2)-3x10(5) cm(-1)). A set of gel-characterizing parameters were recovered by accurately fitting the data with a single function recently developed by us [F. Ferri et al., Phys. Rev. E 63, 031401 (2001)], based on a simple structural model. Accordingly, the gels can be described as random networks of fibers of average diameter d and density rho, entangled together to form densely packed and spatially correlated blobs of mass fractal dimension D(m) and average size (or crossover length) xi. As previously done for d, we show here that the recovered xi is also a good approximation of a weight average, namely, d approximately sqrt[ (w)] and xi approximately (w). By varying the fibrinogen concentration c(F) between 0.034-0.81 mg/ml, gels with 100> or =xi> or =10 microm, 100< or =d< or =200 nm, 1.2< or =D(m)< or =1.4, and constant rho approximately 0.4 mg/ml were obtained. The power-law c(F) dependencies that we found for both xi and d are consistent with the model, provided that the blobs are allowed to partially overlap by a factor eta likewise scaling with c(F) (2> or =eta> or =1). Recasting the whole dataset on a single master curve provided further evidence of the similarity between the structures of all the gels, and confirmed the self-consistency of the model.Physical Review E 08/2002; 66(1 Pt 1):011913. · 2.31 Impact Factor
In situ formation, manipulation, and imaging of droplet-encapsulated fibrin
Heather M. Evans,aEnkhtuul Surenjav,aCraig Priest,abStephan Herminghaus,aRalf Seemannac
and Thomas Pfohl*ad
Received 17th November 2008, Accepted 13th March 2009
First published as an Advance Article on the web 30th March 2009
The protein fibrin plays a principal role in blood clotting and forms robust three dimensional networks.
Here, microfluidic devices have been tailored to strategically generate and study these bionetworks by
confinement in nanoliter volumes. The required protein components are initially encapsulated in
separate droplets, which are subsequently merged by electrocoalescence. Next, distinct droplet
microenvironments are created as the merged droplets experience one of two conditions: either they
traverse a microfluidic pathway continuously, or they ‘‘park’’ to fully evolve an isotropic network
before experiencing controlled deformations. High resolution fluorescence microscopy is used to image
the fibrin networks in the microchannels. Aggregation (i.e. ‘‘clotting’’) is significantly affected by the
complicated flow fields in moving droplets. In stopped-flow conditions, an isotropic droplet-spanning
network forms after a suitable ripening time. Subsequent network deformation, induced by the
geometric structure of the microfluidic channel, is found to be elastic at low rates of deformation. A
shape transition is identified for droplets experiencing rates of deformation higher than an identified
threshold value. In this condition, significant densification of protein within the droplet due to
hydrodynamic forces is observed. These results demonstrate that flow fields considerably affect fibrin in
different circumstances exquisitely controlled using microfluidic tools.
Microfluidic tools hold great promise in nanotechnology.
Reduced sample volumes, advanced sample handling, and high
throughput dramatically increase the productivity of bioassays
and combinatorial chemical reactions.1,2The ability to generate
precise sample aliquots as droplets in an immiscible continuous
phase enables new schemas involving fluid compartments of
picoliter volumes.3–5Experiments on this scale allow finer control
in studies of complex biomolecular processes and provide a more
efficient and cost-effective experimental design. Furthermore, the
use of microfluidic tools has facilitated investigations into the
measurements of the early stages of protein folding,6single actin
filament confinement,7and DNA self-assembly processes.8
This study of fibrin is motivated by the protein’s biological
relevance and attractive network-forming properties. Fibrin
forms via the assembly of subunits of fibrinogen under conditions
that dictate the physical properties of the network formed.
Hemostasis, or healing of injured blood vessels via blood clot-
ting, is initiated by the aggregation of platelets, which are small
as evidenced by
aggregation at the injury site is then aided by the presence of
fibrin. In some cases aggregation can also be detrimental, such as
when foreign therapeutic scaffolds induce the formation of
excessive blood clots.9The biodegradable nature and good tissue
tolerance of fibrin networks have already been demonstrated in
terms of commercially available wound covering agents. In
addition, there is much interest in the deactivation of hemostasis
with respect to artherosclerosis,10inflammation-induced coagu-
lation,11nervous system diseases such as multiple sclerosis,12and
certain cancers.13In addition, from a materials point of view
single fibers of fibrin have been found to show remarkable elas-
ticity.14Looked at as a three-dimensional gel, fibrin networks
display a high elastic modulus that increases with increasing
strain (i.e. strain-stiffen).15This is similar to other biopolymer
networks, but fibrin formation is relatively straightforward and
more reproducible in comparison to other biomaterials: using
only two components, an effectively crosslinked network can
easily be formed.
The building blocks of fibrin are the monomeric protein,
fibrinogen, which comprise about four percent of the total
protein in blood plasma. Each fibrinogen molecule has a length
of 45 nm, and consists of a central domain flanked on either end
by coiled-coil regions. Although the in vivo process of blood
clotting involves a multi-step biological pathway, a minimal
system consists of the protein ‘‘monomers’’ of fibrinogen and the
serine protease thrombin (see figure 1). A three-dimensional
fibrin network is initially catalyzed by the presence of thrombin,
which selectively cleaves fibrinopeptides located in the central
domain of fibrinogen. This cleavage creates activated fibrinogen
molecules that reassemble into protofibrils via knob-hole
aMax Planck Institute for Dynamics & Self-Organization, Bunsenstraße
10, 37073 G€ ottingen, Germany
bIan Wark Research Institute, ARC Special Research Centre for Particle
and Material Interfaces, University of South Australia, Mawson Lakes,
South Australia, 5095, Australia
Saarbr€ ucken, Germany
dDepartment of Chemistry, University of Basel, 4056 Basel, Switzerland
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1933–1941 | 1933
PAPER www.rsc.org/loc | Lab on a Chip
interactions. In this sense, the protofibrils can be considered as
‘‘sticky’’ objects. Subsequent aggregation of nearby fibrils results
in fibers of thicknesses thoughtto be limited bythe twisting of the
fibers.16Due to this growth and the branching of fibers, a three-
dimensional network is formed that spans micrometers, with
properties that depend on salt and thrombin concentration.17–20
Due to its biological and material relevance, fibrin has been
intensely probed using a variety of techniques. Previous studies
of fibrin formation have used bulk measurements, such as optical
density or light scattering, to characterize fibrin gels.17–19,21
Historically, most microscopy of fibrin networks in the literature
resulted from electron microscopy measurements, which require
staining and fixing of samples in processes that may alter the
native structures.22–24While all of these previous studies have
made important contributions to the understanding of fibrin
formation, there is a dearth of direct measurements of fibrin in its
natural state. Recent measurements suggest that the use of high
resolution light microscopy provides information about the
network that correlates well with some macroscopic measure-
ments, such as turbidity assays.25However, the intrinsic sticki-
ness of fibrin clots is a significant obstacle to their controlled
study in dynamic environments.
The use of microfluidic devices allows for a real-time, direct
measurement of the assembly of fibrin gels under well-controlled
conditions. Furthermore, the implementation of microfluidic
droplets enables the in situ formation, manipulation, and visu-
alization of fibrin networks in a manner that was previously
inaccessible. The systems reported here take advantage of the
microfluidic generation of small volumes of proteins and
enzymes to create an enclosed three-dimensional protein micro-
environment. Rather than focusing on high throughput appli-
cations, high resolution microscopy is utilized to characterize the
finer aspects of fibrin network formation and its behavior within
the microenvironment. Ismagilov et al. showed the feasibility of
microfluidics to determine drug dosages based on blood clotting
times in droplets in a high-throughput approach.26In contrast,
the work reported here focuses on details of the network
formation within a droplet and elucidates the influence of the
microflow and confinement on the protein network.
In order to investigate the evolution and manipulation of this
robust network, we developed microfluidic chips that are capable
of initiating fibrin network formation within droplets. This is
achieved via targeted electrocoalescence between adjacent
droplets and allows the early stages of the fibrin network to be
observed directly. The use of droplets prohibits strong adhesion
of fibrin to the microchannel walls during network formation by
encapsulating the protein and enzyme within the dispersedphase.
Furthermore, the incorporation of geometric structures into the
microchannels enables the controlled deformation of individual
droplets containing fibrin networks. Here, we demonstrate two
types of devices: one in which an evolution to large-scale clotting
is observed, and another in which the elastic network behavior is
determined based on the rate of deformation.
Materials and methods
Microchannels were fabricated directly in SU8 photoresist using
UV-photolithography, resulting in a final device as shown in
figure 2. All photolithography solutions were purchased from
Micro Resist Technology GmbH, Germany. Glass disks (50 mm
diameter, VWR) were precoated with Omnicoat and a 5 mm thick
layer of SU8-2005, then exposed to UV radiation and treated in
plasma for 30 seconds to render the surface hydrophilic. After-
wards, SU8-50 was spin-coated to a thickness of 30–40 mm, soft-
baked, and exposed to UV for 30 seconds in the presence of
a photomask containing the desired microchannel structures.
After a post-exposure bake, the glass was rinsed in developer
until the channels were cleared of material. Inlets were drilled
through the glass plate and the channels were closed by thermally
bonding a 50 mm diameter round cover glass (VWR) at 160?C in
a mechanical press. Prior to bonding, gold electrodes and an
underlying chromium adhesion layer were thermally vapor
deposited at 10?6mbar ontothe coverglass and covered by a very
thin bonding layer of PMMA, which was spin coated and baked
onto the cover glass. Wires were glued to the electrodes using
conducting epoxy in order to apply an electric potential during
experiments. Teflon tubing was connected to the microchip via
Nanoports? (UpChurch), which were bonded to the microchip
scales at each stage in formation.
Hierarchical self-assembly of fibrin, with approximate length
Fig. 2SU8 microchip with ports, tubing, and electrodes.
1934 | Lab Chip, 2009, 9, 1933–1941This journal is ª The Royal Society of Chemistry 2009
according to the company protocol. The tubing was connected to
Hamilton microsyringes that were driven using custom-made
pumps controlled by programs written in LabView (National
Instruments Corporation). Additional to the devices described in
this manuscript, similar processes were used to manufacture
a microchip with a much wider chamber in order to stop the large
scale motion of droplets and study the network formation in
what we refer to as ‘‘stopped flow.’’
Human fibrinogen (plasminogen, von Willebrand factor, and
fibronectin depleted) was reconstituted in water to a concentra-
tion of 45 mM according to the supplier’s guidelines (Enzyme
Research Laboratories, UK). Fluorescent fibrinogen Alexa
Fluor? 546 conjugate was also prepared according to guidelines,
to a concentration of 4.4 mM (Mobitec GmbH). Fibrinogen and
fluorescent fibrinogen were diluted 10:1 in a fibrinogen buffer (20
mM Sodium Citrate-HCl, pH 7.4). Human a-Thrombin, activity
approximately 3000 NIH u/mg (Enzyme Research Laboratories,
UK), was reconstituted in water to a concentration of 94.6 mM
and further diluted in thrombin buffer (50 mM Sodium Citrate,
0.2 M NaCl, 0.1% PEG-10,000, pH 6.5). Relative molar ratios of
fibrinogen to thrombin are given in the text as [F]/[T]. For
comparative studies of beads in droplets, Fluoresbrite? carboxy
NYO 0.50 mm microspheres (Polysciences, Inc., Warrington PA)
were diluted 1:1000 in thrombin solution and flowed through the
microchip as usual.
Imaging and analysis
The droplets were imaged using a Zeiss Axiovert 135 microscope
with a 60? 1.25 NA oil objective and an ebx75 Xenon illumi-
nation source equipped with the appropriate filters to perform
fluorescence experiments using an excitation wavelength of
around 546 nm. Images were recorded using a PCO sensicamQE
camera and CamWare software. Shorter exposure times (typi-
cally, about 700 ms) were required to image fibrin in the serpen-
tine device, and longer times (up to 10 ms) were used in the device
with geometric confinements located after a reaction chamber.
Generally speaking, image contrast and fiber network resolution
were considerably improved with longer exposure times. There-
fore, the longest exposure time possible (considering the experi-
mental restriction of droplet velocity) was always chosen.
Further analysis was conducted for three to five images per data
point using ImageJ image processing software.27
Results and discussion
Droplet formation and fibrin gelation in droplets
Droplets of fibrinogen and thrombin solutions were formed in
a continuous oil phase consisting of a low viscosity organic oil
(Isopar?M, Exxon Mobil). In order to stabilize the emulsion,
a non-ionic surfactant (2 wt% Span? 80, Sigma Aldrich) was
added to the continuous phase. As shown in figure 3, the device
contains two T-junction configurations where the droplets of the
dispersed phase (in this case, protein solution) are formed.28,29As
the stream of the dispersed phase penetrates into the main
channel, a droplet develops. The pressure gradient and shear
force generated by the continuous phase distort the droplet in the
downstream direction until the neck of the dispersed phase thins
and breaks into a droplet (inset, upper left, figure 3).
Under ideal synchronized conditions, droplets from each
T-junction pair together at the following Y-intersection in the
device. The two droplets pass through the electrodes very near to
one another (inset, upper right, figure 3). An applied potential
(10 V, 50 Hz) causes two neighboring droplets to coalesce due to
an electric-field induced instability.30
Clotting in droplets
In order to visualize the formation of fibrin, a microchip incor-
porating serpentine channels was designed. This allows the newly
coalesced droplet, containing fibrinogen and thrombin, to fully
evolve into a fibrin network over the extended length of the
microchannel. The microscope remained in a fixed position on
the device for these experiments. A high magnification objective
(60?) was required to visualize the network in sufficient detail
and, as a result, the camera field of view was only large enough to
image one single droplet at a time. Imaging was carried out at
a number of fixed positions along the serpentine microchannel to
analyze the fibrin networks at various timepoints. Representative
images are shown in figure 4 for droplets prepared at a molar
ratio of fibrinogen to thrombin [F]/[T] ¼ 0.5 and a droplet
velocity v ¼ 5 mm/s. The onset of mixing of the two solutions,
seen in figure 4a, is especially clear since the fibrinogen is fluo-
rescent and the thrombin is not. Shortly after droplet coales-
cence, a homogeneous distribution of gray levels is measured
within the droplet (figure 4b). This confirms that complete mix-
ing of the two components has occurred, and is set as the time
point t¼ 0 because we assume thatthe mixing is much faster than
the fibrin-forming interactions between protein components. The
evolution of fibrin formation and clotting is shown in figures 4c–
4f, which are taken at increasing distances (and, correspondingly,
different reaction times t) along the device.
The presence of small, high intensity aggregates is found as the
fibrin network develops. The number and size of these aggregates
(inset, upper left) and fibrinogen (F) solutions form at the T-junctions.
Droplets pair at the Y-intersection and coalesce as they pass through the
electrodes (inset, upper right), and travel through the microchannel
(dimensions given in lower right box).
Serpentine microchip with 4 inlets. Droplets of thrombin (T)
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1933–1941 | 1935
appear to be inversely related, and eventually a large, single clot
of fibrin is found in droplets at the highest time point, cf. figure
4f. Moreover, the background intensity of the droplet transitions
from light gray, at smallest t, to virtually black, at higher t. For
a first analysis, the dimensionless parameter s*, defined here as
the normalized standard deviation of grayscale intensities in
a droplet, was used to quantify these observations. This
parameter is calculated in the following manner: s* ¼ (s/I)t/(s/
I)t ¼ 0, where I is the mean intensity and s is the standard devi-
ation of the intensities.
The values of s*, averaged over several droplets, are plotted as
a function of time in figure 4. A substantial increase in s* with
increasing reaction time is found. This parameter seems to
accurately describe the aggregation of fibrin as the droplets
proceed through the device. Additionally, values of s* for the
same timepoint (t ¼ 2.7 s) and droplet velocity (v ¼ 5 mm/s) but
with different [F]/[T] ratios 0.5, 1, and 5 are compared in the inset
in figure 4. These ratios are changed by varying the upstream
concentration of thrombin, [T]; in each experiment the concen-
tration of fibrinogen, [F], is kept constant so that the amount of
fluorophore is also conserved. It can be clearly observed that the
aggregation is faster for the smaller ratio [F]/[T] ¼ 0.5, for which
there is an excess of the enzyme thrombin compared to the
substrate fibrinogen. Furthermore, for a higher ratio, [F]/[T] ¼ 5,
the droplets remained homogeneously gray at all points in the
microchip, indicating that the accessible time scale (tmax? 30 s) is
not long enough in this case to visualize fibrin formation. The
evolution of large scale aggregation, and the corresponding
increase in s*, is particularly striking when compared to the
values of s* observed for fibrin networks formed under stopped
flow conditions. Using a microdevice of similar thickness but
wider opening, it was possible to coalesce droplets and then hold
them in a single location, thereby stopping the movement of the
droplet (device not shown). In these stopped flow conditions, s*
does not increase significantly from the starting value of 1, when
the two components have initially homogenously mixed. A good
example of the type of fibrin network formed in stopped flow can
be found in figure 6a. In these conditions, the formation of
a three-dimensional fibrin network results in a slight increase in
s*. However, this value remains very low over all measured
timescales (e.g. 1 hour, results not shown). Most importantly, s*
in stopped flow never approaches the higher values of s*
measured for moving droplets containing large aggregates.
The direct comparison between these two droplet conditions
clearly demonstrates that the flow fields in moving droplets play
an important role in the aggregation of fibrin networks. These
flow fields apparently induce and accelerate fibrin interactions,
culminating in protein aggregation. However, the circulating
flow fields in microdroplets are known to be complicated. Mixing
between two liquids coalesced into a single compartment has
been described as chaotic advection.31Recent measurements
using confocal-PIV to follow the fluid motion in plug-shaped
droplets indicate that indeed the circulating flow patterns are far
from straightforward, with several vortices found at droplet
edges.32,33The flow direction and magnitude depends on the
position within the droplet and is driven by shearing interactions
between the droplet and the microchannel wall. For the type of
adjacent (i.e. axial) mixing scenario used here, two circulation
vortices are seen when viewing the droplet in 2D, at the center of
which are stagnant zones.34For clarification, some streamlines
within a moving droplet are sketched in the inset of figure 5. In
similar plug-shaped droplets, the vortices at fluid-fluid interfaces
can have velocities nearly 25% higher than the bulk droplet
velocity (based on other measurements in our lab).
The complicated flow fields in the droplets directly impact the
fibrin growth in a manner that is not entirely straightforward.
The formation of fibrin protofibrils is an intermediate step in the
fibrin pathway, as illustrated in figure 1. The formation and
lateral growth of these fibrils into larger fibers can, for the most
part, be observed at optical resolution. As we cannot observe the
smaller protofibrils using optical methods, the results presented
here are indicative of larger fiber aggregation. The images taken
bar represents 60 mm. Right: Normalized standard deviation of droplet intensities s* at different time points in flow compared to a droplet in stopped
flow. Inset graph: For the same evolution time, t ¼ 2.7 s, s* varies with [F]/[T] and no fibrin formation is seen for high [F]/[T] ¼ 5 over the entire device
(denoted with *).
Fibrin formation in droplets ([F]/[T] ¼ 0.5, v ¼ 5 mm/s). Left: Representative images from (a) initial mixing stage to (f) final clotting. The scale
1936 | Lab Chip, 2009, 9, 1933–1941 This journal is ª The Royal Society of Chemistry 2009
at multiple timepoints clearly show the aggregation of protein
into small clumps and then further aggregation into a single,
large fibrin ‘‘particle’’. This is in stark contrast to the droplet-
spanning network of fibers recorded for stopped flow. Even in
this condition, however, it has been seen that movement of
neighboring protofibrils occurs during fibrin formation –
purportedly due to forces exerted by the newly formed fibrin
network – and this fibril contact induces network formation.25
The hydrodynamic conditions within a moving droplet will
contribute to additional, significant impacts on fibrillization.
Irreversible fibrin aggregation in moving droplets is enhanced by
local flow fields and is driven by the proximity of the small,
mobile, and active fibrils to one another and to larger aggregates.
The flow in the droplets increases particle–particle collisions,
where particles can be fibrinogen, thrombin, fibrin fibrils or even
fibrin aggregates. Thus, the flow field influences nearly all levels
of protein formation and behavior. The localized segregation
between fibrin aggregates and solution seen in figure 4 appears to
be a type of complex coacervation similar to what has been seen
with other proteins.35
In summary, the impact of droplet flow fields on protein
aggregation and clotting is an important consideration. The
hydrodynamic effects within a droplet appear to significantly
influence protein interactions and, therefore, comparisons
between in vitro protein behavior and in vivo behavior must be
carefully evaluated. However, it is plausible that flow fields in
blood vessels may also promote molecular crowding and coac-
ervation of fibrin in blood clots, both in blood flow as well as at
the vessel injury site.
Manipulation of fibrin networks
In order to reduce hydrodynamic influences on fibrin formation
and to enable the online manipulation of fibrin networks,
a second microfluidic chip was employed. The ability to ‘‘park’’
a droplet creates an environment for internal droplet processes to
proceed over long time scales.36In this microchip, a reaction
chamber allows the fibrin network to evolve in a stopped flow
condition before it is mechanically manipulated through a series
of microchannel geometries. The droplet velocity is reduced in
this reaction chamber by a substantial increase in the channel
width, as shown in figure 5. The velocity of the droplets in the
chamber is controlled not only by the inflow of liquid by the
syringe pumps at the inlets but also by the withdrawal of
continuous phase at an outlet from the side of the reaction
chamber. Consequently, the droplet velocity is tunable and
coalesced droplets can beheld in the reaction chamber for a given
residence time or ‘‘network ripening time’’, tn.
Using this channel design, a ripening time tnon the order of
tens of seconds in the reaction chamber is required to observe the
type of network seen in figure 6a, for [F]/[T] ¼ 2. At this time tn,
we observe an isotropic fibrous network that displays multiple
crossing points of fibrils but does not show any large scale
aggregation behaviors (i.e. those found under the flow conditions
in the microchip described earlier). When the coalesced droplet
sits in the reaction chamber, protofibril and larger fiber forma-
tion are able to proceed without any significant influence of
hydrodynamic effects. Also, the fibers’ thickness appears to be
conserved for long observation times (e.g. 1 hour), which
suggests that a steady state has been reached. The final result is
a genuine droplet-spanning fibrin network. In the following
analyses we treat this protein network as quasi-two-dimensional,
due to the high aspect ratio of the droplet width to depth (5:1)
and for simplicity of analysis. Some fluorophores are found at
the droplet-oil interface of coalesced droplets (possibly due to the
surfactant system used here), but the fluorescence intensity at the
is 40 mm. Negligible structural difference is observed in similar confinements (i.e. b and d, c and e), highlighting the elasticity of the fibrin networks at
Images of the fibrinnetwork in the same droplet ([F]/[T] ¼ 2, vnarrow? 200 mm/s) at positions a–e, corresponding to the sketch of device. Scale bar
elements for manipulating the fibrin networks. The velocity V2of the side
channel controls the velocity of single droplets traveling through the
microchannel outlet, V1. Some of the flow fields within a moving droplet
in the microchannel are sketched in the upper inset, although detailed
treatments can be found in the literature cited in the text.
Microchip incorporating a reaction chamber and geometric
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1933–1941 | 1937
interface does not change over long time periods and therefore
this effect may be disregarded in a first approximation. Addi-
tionally, in every case where the fibrin networks form, some
depletion of the network near the droplet-oil interface is
observed, indicating that there are no specific interactions
between the network and the droplet interface.
The fibrous networks are manipulated as they pass through the
microchannel, which contains several geometric obstacles. The
microchip shown in figure 5 contains narrow regions, which
result in a compression of the droplets, followed by wide regions,
which allow the droplet to regain its original circular shape. By
appropriately adjusting the flow at the side channel of the reac-
tion chamber (V2), the droplets are driven through the final
stages of the microchip at velocities ranging up to 1 mm/s.
The images in figure 6 show one droplet prepared at [F]/[T] ¼ 2
as it passes through the microchip. Due to the very slow flow
speeds achievable using these devices, individual droplets could
be followed by translating the microscope sample stage. There-
fore, direct comparisons of the fibrin network during and after
compression were possible. Comparing the squeezed networks in
figures 6b and 6d, one sees qualitatively similar features. This
finding is also true for further constrictions along the micro-
channel. Centrally located fibers extend along the flow axis and
a depletion region is found between the fibrin network and the
surface of the droplet. The fibrin networks in the wide regions of
figures 6c and 6e, where the droplet shape reforms to a nearly
circular cross-section, show a different fiber arrangement that
approaches the isotropic network of figure 6a. Similar features
can be identified when comparing the circular droplets in 6c and
6e, which indicates a restoration of general network features
throughout the droplet’s manipulation in the device. Despite the
existence of some weak plastic deformation, figure 6 clearly
shows that the fibrin networks formed under these conditions are
generally able to compress and extend in an almost reproducible
manner. This indicates a considerable degree of structural
integrity and network elasticity. Remarkably, this conservation
occurs despite the fluid flow that percolates the fibrin network in
the moving droplet.
To appropriately describe the experimental condition of the
droplets, we introduce a rate of deformation 3 ¼ Dv/l. Here, Dv ¼
vnarrow? vwideis the velocity difference of the droplet in the two
microchannel widths and l ¼ 160 mm is the typical length scale of
deformation. To a good approximation, the velocities of the
droplet depend on the width ratio of the deformation induced by
the channels, wnarrow/wwide¼ 80 mm /160 mm ¼ 0.5. This allows 3
to be calculated using only vnarrow, which was found to be
consistently accessible experimentally, especially in cases where
the fast moving droplet could not be tracked over long distances
using the microscope stage. Using these assumptions, the rate of
deformation can be calculated using 3 ¼ vnarrow/320 mm?1. Using
these parameters, the data in figure 7 correspond to droplets
undergoing a rate of deformation 3 ¼ 0.6 s?1.
Points of high intensity in the droplets are correlated to fibril
cross-over in the interconnected fibrin network. An analysis of
these points was conducted in order to quantify the structural
similarities (or differences) between the networks in compressed
and widened regions. Using image analysis software, the intense
network features were identified by a binary algorithm with
a conserved tolerance level. Examples of this process are shown
in figure 7, where the original image frame is shown above the
resulting point distributions for a droplet in both a wide (figure
7a) and narrow (figure 7b) channel. The positions of these
features were used to calculate the distance R of each point from
the droplet center (i.e. center of mass). Histograms of these
distances are plotted in figure 7, for a single droplet passing
through wide and narrow channel regions. For comparison, the
distribution of network positions within the droplet in the reac-
tion chamber, i.e. the native fibrin network, is also shown in
All of the distributions shown in figure 7a for the circular
droplets in wide regions approximately follow a linear scaling in
R up to a given threshold of around R ¼ 45 mm. Depletion of the
network from the droplet-oil interface is evident from visual
inspection of these images and results in a decrease in the
network point frequencies close to the physical droplet radius
(vertical gray line). These results agree with the expected distri-
bution of points contained within a circular area, except for the
depletion effect. The depletion is exacerbated by the presence of
the interacting protein network. A comparison of the distribu-
tion of fluorescent beads in droplets studied in a similar device
(black curve in figure 7a) highlights the significant depletion that
occurs with fibrin networks. Furthermore, a larger depletion of
the network is measured in the wide regions as compared to the
native structure found in the reaction chamber (arrows, figure
7a). This increased depletion is a consequence of droplet defor-
mation and may also be due to neighboring fiber interactions
driven by crowding effects. It is important to note that after the
droplet initially experiences a deformation, the extent of deple-
tion remains conserved. In fact, all of the distributions from wide
channel regions shown in figure 7a (with the exception of the data
from the droplet in the reaction chamber) overlay remarkably
well. This suggests that the fibrin network recovers completely
after repeated cycles of squeezing, which is qualitatively seen in
the microscopy images.
A similar reproducibility is found when considering the
distributions for the elongated droplets in the narrow channels,
figure 7b. In these regions the droplet assumes a plug of stadium
shape, which is neither a rectangle nor an ellipse, and is described
by a short axis asand long axis al. These two length scales emerge
in the distributions and are labeled with vertical gray lines in
figure 7b. For distances less than as, the distributions of inten-
sities scale linearly until reaching a threshold distance of about
25 mm, and then decrease reflecting depletion from the walls at as
in a manner similar to that found in the circular droplets.
However, the linear scaling in the plugs has a slope that is higher
than that measured in the droplets of figure 7a. This suggests an
increased densification of the network when the droplet is
compressed. The behavior of fibrin at distances between asand al
significantly deviates from measurements of fluorescent beads in
similar plugs (black line in figure 7b). Although the distribution
of beads remains high to distances of nearly 100 mm, the
frequency of point distances in the fibrin network substantially
decreases on the same length scale. However, the extent of
depletion is conserved upon repeated compression of the droplet.
There is no observation of delayed fibrin network recovery, as
evidenced by the fact that the distributions shown in figure 7 do
not vary significantly at similarly shaped device positions. The
impressive elasticity of these fibrin networks echoes the results of
1938 | Lab Chip, 2009, 9, 1933–1941 This journal is ª The Royal Society of Chemistry 2009
Wen et al., who reported elastic moduli ranging from 50 Pa to
a maximum of 900 Pa for fibrin clots undergoing stress in a cone-
plate rheometer.15These authors also found the dominant fibrin
deformation to be affine, meaning that the local strain at any
point followed the macroscopic strain of the system. The
recovery of the fibrin networks after repeated compressions
reported here suggests that the network exhibits properties
similar to those reported in rheology experiments – including
undergoing affine deformations.
Coacervation of deformed fibrin networks
At slow velocities and rates of deformation, 3, fibrin networks
centered within the droplet can survive several cycles of
compression. Furthermore, the network under these conditions
is elastically deformable. However, at higher 3, inelastic network
properties may be accessed if the deformation is faster than the
recovery of the network. This is achieved by increasing the
droplet velocity through the microchannel (of the microchip
shown in figure 5). At sufficiently high 3, two distinct phases are
present within the droplet. This is clearly visualized in figure 8c,
which shows a droplet with 3 ¼ 1.125 s?1(vnarrow¼ 360 mm/s).
The fibrin network has densified, concentrating toward the rear
of the droplet with respect to the direction of droplet movement
in the microchannel. For comparison, figure 8a shows an image
in the condition where 3 ¼ 0.5 s?1, in which case the network is
distributed across the entire droplet. Based on observation of
droplets moving at different velocities, the extreme separation
into dense fibrin is found for droplets experiencing a rate of
deformation 3 > 1 s?1.
A comparison of the fibrin network conditions in droplets
traveling at different velocities is plotted in figure 8b for [F]/[T] ¼
2. Images can once again be analyzed using the standard devia-
tion of intensities normalized by the mean intensity. In this case,
it is appropriate to use the relation s0¼ (s/I)narrow/(s/I)chamber,
where (s/I)narrowis the value for a droplet undergoing its first
deformation in the narrow region of the microchannel and (s/
I)chamberis the ‘‘original’’ value of the isotropic network state of
0.6 s?1. Network points (lower images) are determined from original data (upper images). Droplet radii and axes are marked with vertical gray lines.
Compared to the round droplet in the reaction chamber prior to any compression (black squares), circular droplets show a slight shift in peak frequency
(arrows). The first, second, and third repetitions through the microchip are plotted with circles, triangles, and diamonds, respectively. For comparison,
the histograms for beads in aqueous solution in a similar device are plotted in solid lines for both circles and plugs.
Histogramsof networkpoint distancesfromcenter of massof (a) circularor (b) compressed dropletswith [F]/[T] ¼ 2 andrate of deformation3 ¼
which is reflected by their higher values of s0(black data points in panel b, also see panel c). No separation is visually observed for 3 less than one, where
s0< 2 (white data points in panel b, also see panel a). The flow fields within droplets in the two conditions are also sketched (below the images in panels
a and b), where the network, drawn in gray, either extends along the entire droplet (a) or resides in the back half of the droplet (b). For comparison, the
horizontal gray line is the average s* value in stopped flow from figure 4. Dotted line is guide to the eye.
Two ‘‘phases’’ of fibrin-rich and fibrin-poor regions are found for droplets experiencing rates of deformation 3 approximately one or greater,
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1933–1941 | 1939
the same droplet while it resided in the reaction chamber.
Comparing the calculated values of s0with visual inspection of
fluorescence images, it appears that s0> 2 correlates with the
segregation of fibrin-rich and fibrin-poor regions within a single
droplet. Moreover, a critical rate of deformation 3 z 1 s?1can be
identified as a threshold for this coacervation behavior.
This behavior can be interpreted as a shape transition in fibrin
networks owing to hydrodynamic flow within a moving and
deformed droplet. As long as the liquid (buffer) flows slowly
through the fibrin network, a small pressure gradient is needed to
accomplish flow and thus the network remains unaffected, aside
from a small elastic perturbation. This corresponds roughly to
the situation sketched in figure 8a. As the flow becomes stronger,
the pressure gradient along the protein network may become
large enough to overcome the elastic forces maintaining its
shape. As it is compressed, the fiber network inside becomes
denser, and the flow resistance increases further. This may lead to
a runaway instability, which finally results in a situation which is
sketched in figure 8c. The flow is now almost entirely expelled
from the protein network.
In order to gain some quantitative understanding of this effect,
we start by assuming the network forms at some initial axial
length l0, which is given, e.g., by the length of the water droplet.
As the flow increases, this length may change and will be called l.
Weintroduce the dimensionless quantity x ¼ l/l0for convenience.
The force (per area) that is needed to elongate or compress the
network longitudinally is Fl¼ E(x ? 1) given by Hooke’s law (E
is Young’s modulus). The flow rate within the fiber network is
given by VpP/h, where Vp is the pressure gradient, h is the
viscosity of the liquid, and P is the permeability of the network.
Since the permeability of a mesh is roughly (up to a prefactor of
order unity) given by the square of the mesh width, we can obtain
an expression for P as a function of the network density. If d0is
the average diameter of the fibers, we find
where r0< 1 is the initial density of the network directly after
formation. The force (per area) exerted onto the network
longitudinally owing to the hydrodynamic flow is
where_V is the flow rate, which is related to the deformation rate
by_V ¼ 3a3(a is the radius of the droplet). The network’s ‘‘bulk’’
shape transition takes place where Fl? Ffhas a degenerate zero.
We define a dimensionless scales deformation rate
~ 3 ¼
which is of order unity at the transition. This corresponds to
a Young’s modulus of the fibrin network within a plug in an
elastic moduli range that was reported by Wen et al.15Therefore,
we find that this simple physical model is both compatible with
existing descriptions of fibrin networks and is able to explain our
new experimental evidence of a transition at a critical rate of
deformation 3 z 1 s?1.
The effect of this threshold deformation rate is manifested as
irreversible clotting into a dense fibrin network. In the reaction
chamber, the droplets initially contained an isotropic, fibrous
ripened network. This is considerably different to the homoge-
nous mixture used as the starting point in measurements of s* in
the serpentine device, where the continuously moving droplet did
not allow droplet spanning network formation. The fact that
a similar s parameter describes the aggregation phenomena in
both cases demonstrates the generality of hydrodynamically
induced aggregation. Therefore, these hydrodynamic effects may
be a crucial consideration in droplet-based experiments, where
multiple components experience microfluidic flow conditions.
This effect is especially important in bionetwork formation, as
shown here, and other processes such as polymerization (e.g.
online particle or capsule synthesis). Furthermore, the complex
effects of hydrodynamics on network aggregation under micro-
fluidic flow conditions might suggest a general tendency of fibrin
networks to aggregate in vitro. The preferred aggregation states
also serve as a reminder that the networks are delicate and non-
equilibrium in their nature and therefore access to them, via
microfluidic tools, is particularly novel.
The protein fibrin was selected to explore the influence of
microfluidic flow on the formation and aggregation of a bionet-
work. By its nature as a primary component of blood clots, fibrin
exhibits strong adhesion to surfaces including microchannel
walls. Therefore, the networks in this study were confined within
droplets. This was achieved using tailored microfluidic chips,
containing sites for droplet formation, electrocoalescence, in situ
networks. The formation of fibrin was initiated in a droplet by
coalescing two droplets, one with the fibrinogen monomer and
the other containing the enzyme thrombin. The development of
fibrin fibers within the resulting droplet required time for acti-
vation and fibril growth, and eventually neighboring fibers
interacted to form a three-dimensional network.
Hydrodynamic flow within the droplets was found to
dramatically accelerate the aggregation of fibrin into small and,
eventually, very large particles, i.e. clots. However, in stopped
flow environments, isotropic three-dimensional networks were
produced. In this case, the fibrin formation was not influenced by
extreme flow patterns because it remained ‘‘parked’’ in
a chamber. These isotropic networks were then driven through
repeated geometric confinements to study the elasticity of the
networks. Fluorescence analysis of the fibrin network at slow
velocities did not show any evidence of inelastic deformation,
despite undergoing repeated cycles of compression. However,
a critical rate of deformation 3 was identified, above which fibrin-
rich and fibrin-poor regions appear within the droplet. In this
case, the clotting behavior of fibrin into a dense pellet within the
droplet occurs as a result of surpassing a critical hydrodynamic
force on the network. In both microchips, the standard deviation
of intensities normalized by the mean intensity, i.e. s* or s0, is an
appropriate parameter to quantify the extent of aggregation.
manipulation of the fibrin
1940 | Lab Chip, 2009, 9, 1933–1941This journal is ª The Royal Society of Chemistry 2009
Our microfluidic approach allows the visualization of fibrin
coacervation under a variety of experimental conditions.
Therefore, the tendency of fibrin networks to aggregate can be
approached in multiple ways to understand some general prin-
ciples of the protein’s behavior. In the case of fibrin, its naturally
‘‘sticky’’ properties are conserved in the microchips, which
manifests as a form of ‘‘clotting’’. High quality fluorescence
images allow the direct visualization of the networks and
aggregates. This, in concert with precise mixing of small volumes
in microfluidic devices, enables detailed studies of network
behavior and conformation under confinement and well-
controlled experimental conditions. Generally speaking, the
powerful experimental approach demonstrated here is broadly
relevant to a wide variety of physical systems where a greater
understanding of the delicate interplay between material prop-
erties and flow is desired.
The authors thank Udo Krafft for technical assistance and
Audrey Steinberger for sharing her preliminary experimental
data on droplet microflow environments. H.E. acknowledges
a fellowship from the Alexander von Humboldt foundation.
Funding was provided by the Deutsche Forschungsgemeinschaft
in the framework of SPP 1164 ‘‘Micro- and Nanofluidics’’ (PF
375/4, SE 1118/2, SE 1118/4) and SFB 755 ‘‘Nanoscale Photonic
1 C. Hansen and S. R. Quake, Current Opinion in Structural Biology,
2003, 13, 538.
2 S. K. Sia and G. M. Whitesides, Electrophoresis, 2003, 24, 3563–3576.
3 H. Song, D. L. Chen and R. F. Ismagilov, Angewandte Chemie
International Edition, 2006, 45, 7336–7356.
4 S.-Y. Teh, R. Lin, L.-H.Hungand A. P. Lee, LabChip, 2008,8, 1087–
5 E. Surenjav, C. Priest, S. Herminghaus and R. Seemann, Lab Chip,
2009, 9, 325–330.
6 L. Pollack, M. W. Tate, A. C. Finnefrock, C. Kalidas, S. Trotter,
N. C. Darnton, L. Lurio, R. H. Austin, C. A. Batt, S. M. Gruner
and S. G. J. Mochrie, Physical Review Letters, 2001, 86, 4962.
7 S. K€ oster, J. Kierfeld and T. Pfohl, Eur. Phys. J. E, 2008, 25, 439–449.
8 T. Pfohl, A. Otten, S. K€ oster, R. Dootz, B. Struth and H. M. Evans,
Biomacromolecules, 2007, 8, 2167–2172.
9 B. Ratner, Biomaterials Science, an Introduction to Materials in
Medicine, Academic Press, 1996.
10 H. Stormorken and K. S. Sakariassen,Thrombosis Research, 1997, 88,
11 M. Levi, T. T. Keller, E. van Gorp and H. ten Cate, Cardiovascular
Research, 2003, 60, 26.
12 R. A. Adams, M. Passino, B. D. Sachs, T. Nuriel and K. Akassoglou,
Mol. Interv., 2004, 4, 163–176.
13 F. R. Rickles, S. Patierno and P. M. Fernandez, Chest, 2003, 124,
14 W. Liu, L. M. Jawerth, E. A. Sparks, M. R. Falvo, R. R. Hantgan,
R. Superfine, S. T. Lord and M. Guthold, Science, 2006, 313, 634.
15 Q. Wen, A. Basu, J. P. Winer, A. Yodh and P. A. Janmey, New
Journal of Physics, 2007, 9, 428.
16 J. W. Weisel, C. Nagaswami and L. Makowski, PNAS, 1987, 84,
17 E. Di Stasio, C. Nagaswami, J. W. Weisel and E. Di Cera, Biophys. J.,
1998, 75, 1973–1979.
18 F. Ferri, M. Greco, G. Arc? ovito, M. De Spirito and M. Rocco,
Physical Review E, 2002, 66, 011913.
19 F. Ferri, M. Greco, G. Arcovito, F. A. Bassi, M. De Spirito,
E. Paganini and M. Rocco, Physical Review E, 2001, 63, 031401.
20 J. W. Weisel, in Advances in Protein Chemistry, eds. D. A. D. Parry
and J. Squire, Elsevier, San Diego, 1st edn, vol. 70, 2007, pp. 247–299.
21 M. E. Carr and J. Hermans, Macromolecules, 1978, 11, 46–50.
22 E. A. Ryan, L. F. Mockros, J. W. Weisel and L. Lorand, Biophys. J.,
1999, 77, 2813–2826.
23 W. A. Voter, C. Lucaveche, A. E. Blaurock and H. P. Erickson,
Biopolymers, 1986, 25, 2359–2373.
24 J. W. Weisel, Biophysical Chemistry, 2004, 112, 267.
25 I. N. Chemysh and J. W. Wiesel, Blood, 2008, 111, 4854–4861.
26 H. Song, H. W. Li, M. S. Munson, T. G. VanHa and R. F. Ismagilov,
Anal. Chem., 2006, 78, 4839–4849.
27 Available online, http://rsbweb.nih.gov/ij/.
28 T. Thorsen, R. W. Roberts, F. H. Arnold and S. R. Quake, Physical
Review Letters, 2001, 86, 4163.
29 T. Nisisako, T. Torii and T. Higuchi, Lab Chip, 2002, 2, 24–26.
30 C. Priest, S. Herminghaus and R. Seemann, Applied Physics Letters,
2006, 89, 134101.
31 H. Song, M. R. Bringer, J. D. Tice, C. J. Gerdts and R. F. Ismagilov,
Applied Physics Letters, 2003, 83(12), 4664–4666.
32 H. Kinoshita, S. Kaneda, T. Fujii and M. Oshima, Lab Chip, 2007, 7,
33 D. Malsch, M. Kielpinski, R. Merthan, J. Albert, G. Mayer,
J. M. K€ ohler, H. S€ uße, M. Stahl and T. Henkel, Chemical
Engineering Journal, 2008, 135S, S166–S172.
34 W. Tanthapanichakoon, N. Aoki, K. Matsuyama and K. Mae,
Chemical Engineering Science, 2006, 61, 4220–4232.
35 C. G. deKruif, F. Weinbreck and R. De Vries, Current Opinion in
Colloid and Interface Science, 2004, 9(5), 340–349.
36 S. K€ oster, F. E. Angile, H. Duan, J. J. Agresti, A. Wintner,
C. Schmitz,A. C.Rowat,
A. D. Griffiths and D. A. Weitz, Lab Chip, 2008, 8, 1110–1115.
C. A. Merten,D. Pisignano,
This journal is ª The Royal Society of Chemistry 2009Lab Chip, 2009, 9, 1933–1941 | 1941