Chemical Engineering Science 160 (2017) 200–209

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

The wound healing assay revisited: A transport phenomena approach MARK

a a,b,c,⁎ a,b,c
Flora Ascione , Sergio Caserta , Stefano Guido
a
Dipartimento di Ingegneria Chimica dei Materiali e della Produzione Industriale (DICMAPI) Università di Napoli Federico II, P.le Tecchio, 80, 80125
Napoli, Italy
b
CEINGE Biotecnologie Avanzate, Via Sergio Pansini, 5, 80131 Naples, Italy
c
Consorzio Interuniversitario Nazionale per la Scienza e Tecnologia dei Materiali (INSTM), UdR INSTM Napoli Federico II, P.le Tecchio, 80, 80125 Napoli,
Italy

A R T I C L E I N F O A BS T RAC T

Keywords: The Wound Healing (WH) assay is one of the most popular methods for the analysis of cell migration in vitro,
Cell motility widely used to investigate physiological and pathological processes. Several experimental factors of difficult
Wound Healing control, such as variation of cell density, hinder a reliable and reproducible application of this assay. We
Transport processes investigate the effect of cell density (from very low values to complete surface occupation) on WH assays on
Time-lapse microscopy
human fibrosarcoma cells, by using in vitro time-lapse microscopy. We found that wound closure velocity is
Image analysis
linear with cell density, and explained this dependence by analyzing wound closure as a diffusion-reaction
Biomedical engineering
process, according to available models. This finding leads to a simple scaling of the experimental data to account
for cell density differences, obtaining a significant improvement in the quantitative assessment of results. We
also suggest a simple way to evaluate whether cell motility or proliferation drive the process, based on a non-
dimensional parameter.

1. Introduction
The dynamic behavior of the cells, driven by proliferation and
migration mechanisms (Lauffenburger, 1989; Lauffenburger and
Horwitz, 1996), is essential for a wide spectrum of physiological and
pathological processes, including morphogenesis, angiogenesis, im-
mune response, tissue repair, tumor growth and invasion (Horwitz
and Webb, 2003; Ziebert and Aranson, 2013). In order to achieve a
better comprehension of these complex processes (Lauffenburger,
1989; Veltman, 2014) a rigorous approach, based on the quantitative
measurement of well-defined cell movement and proliferation indices,
is required. For this reason, the development of quantitative analyses is
nowadays within the core business of Chemical Engineering (Ottino,
2011), which can contribute to the building of mathematical models,
based on a transport phenomena approach, useful to describe and
predict the mechanisms driving cell dynamics (Love, 2010).
The wound healing (WH) assay is widely used to quantitatively
investigate the dynamical aspects of cell behavior in vitro (Liang et al.,
2007; Kramer et al., 2013). In the classical WH assay, also referred as
scratch test, the cells are grown on a two-dimensional surface; in
principle cell growth can reach a condition defined as confluency,
where the available space is completely covered by cells (Sheardown
and Cheng, 1996). An artificial scratch is then created on the confluent
cell layer by mechanically scraping off an area of cells with a sharp-
tipped instrument (Rodriguez et al., 2005). Then the cells are washed
with medium to remove cell debris and floating cells. Several alter-
native wounding techniques, including laser ablation (Kiehart et al.,
2000; Zordan et al., 2011), electric field application (Keese et al., 2004)
and microfluidic approaches (Murrell et al., 2011; Nie et al., 2007),
have been developed to implement the WH assay in a more controllable
way (Riahi et al., 2012).
In response to the stimulus arising from the availability of free
space (Block et al., 2004), the cells at the wound edges, which are no
longer contact-inhibited, proliferate and move toward the center of the
denuded region to cover the wound area (Tremel et al., 2009; Poujade
et al., 2007). Depending on the type of cells involved, two main
mechanisms of wound healing have been identified (Kramer et al.,
2013; Liang et al., 2007; Rodriguez et al., 2005; Rørth, 2009). In fact,
epithelial-like cells repopulate the wound area in a collective mode,
ensured by strong cell-cell interactions, moving as two coherent sheets
(Nikolic et al., 2006; Poujade et al., 2007; Vitorino and Meyer, 2008).
Fibroblast-like cells cover the wound region moving isotropically as
dispersed individual units (Liang et al., 2007), due to the lack of strong
cell-cell contacts.

⁎
Corresponding author at: Dipartimento di Ingegneria Chimica dei Materiali e della Produzione Industriale (DICMAPI) Università di Napoli Federico II, P.le Tecchio, 80, 80125
Napoli, Italy.
E-mail address: sergio.caserta@unina.it (S. Caserta).

http://dx.doi.org/10.1016/j.ces.2016.11.014
Received 25 March 2016; Received in revised form 28 October 2016; Accepted 5 November 2016
Available online 09 November 2016
0009-2509/ © 2016 Elsevier Ltd. All rights reserved.
F. Ascione et al.
The wound healing process can be mathematically described using
the Fisher-Kolmorgoroff equation (Arnold and Adam, 1999; Cai et al.,
2007; Johnston et al., 2014; Sherratt and Murray, 1990), which
includes terms for modeling cell motility and proliferation. Both these
mechanisms are involved in the spatial spreading of the cells invading
the wound area. The Fisher-Kolmorgoroff equation (Eq. (1)) describes
the evolution in space and time of the cell density u(x,t) as a reaction-
diffusion system, function of two contributions. In Eq. (1) u is cell
density at time t at a given distance x, measured from the wound edge,
D is the constant diffusivity (also referred as random motility coeffi-
cient), k is the proliferation rate and û is cell density at confluency
(Maini et al., 2004b):
∂u ∂ 2u ⎛ u⎞
=D 2 +ku ⎜1 − ⎟
∂t ∂x ⎝ û ⎠ (1)
The diffusion term of the equation models cell motility as a Fickian
diffusion with a cell flux proportional to the concentration gradients, in
which the driving force that determines the spreading of the two cell
sheets in the denuded area is the difference between cell density on the
wound edge and in the wound region. The reaction term of the
equation, that is a parabolic function of cell density, mathematically
describes cell proliferation as a logistic growth, and includes crowding
effects by reducing the growth rate as the cell density approaches
confluency (û). Overall, the Fisher-Kolmorgoroff equation predicts that
after a short transient, the movement of the invading cell front can be
observed in terms of a traveling wave, that propagates with constant
speed s = 4ûkD in the direction perpendicular to the wound (Maini
et al., 2004a; Tremel et al., 2009).
As the measurement of cell density is a challenging task from an
experimental point of view, the wound closure process is often
quantified in terms of change in the wound size over time (Zahm
et al., 1997). In order to gain this quantitative information over long
periods of time, two methodological approaches are typically used. A
popular approach is based on the manual acquisition of microscope
images within the sample, in almost random positions that can vary
along the wound from time to time, at the beginning and at fixed time
intervals (for example every 6 h) until the gap is closed (Kanazawa
et al., 2010). This approach is approximate, since it doesn’t allow
investigating the dynamic of cell spreading in the wound area. An
alternative, and more reliable, approach is based on direct visualization
by time-lapse microscopy (TLM), which enables to dynamically moni-
tor specific regions along the wound, rather than compare randomly
acquired images along the wound, iteratively acquiring sample images
with a defined time frequency (Terryn et al., 2009), while controlling
the environmental parameters to ensure cell viability (Silano et al.,
2012; Wong et al., 2013). Both methodological approaches allow to
determine the number of cells in the wound region (Andújar et al.,
2013; Tsai et al., 2014) or the percentage of wound closure (Lucena
et al., 2011; Sumagin et al., 2013; Yue et al., 2010). However, TLM
should be used to obtain accurate quantitative measurements, such as
wound closure velocity, calculated by measuring the reduction of
wound area over time, or the cell front propagation speed, by
quantifying the position of the wound front over time (Bindschadler
and McGrath, 2007; Poujade et al., 2007). WH assays are typically
performed in multi-well culture plates in order to carry out several tests
in parallel. This allows to investigate simultaneously the effect of
various biological treatments on cell dynamics. However, as several
common biological factors, including contaminations of the samples
and environmental variations, can skew the experimental results, a
control cell sample where no treatment has been done is necessary to
provide a baseline for comparison.
Although the WH assay provides a valuable experimental approach
for studying cell dynamics in vitro, the outcomes of the WH assay are
somewhat influenced by several factors, that may represent a limit in
accomplishing reproducible and reliable quantitative results. For
201
Chemical Engineering Science 160 (2017) 200–209
example, the wound width can vary along its length and among
different experiments (Grasso et al., 2007) (Riahi et al., 2012).
Moreover, the scratching process may involve mechanical injures to
the cells at the wound edges (Zhang et al., 2013). Some damaged cells
and cell debris can also keep attached to the wound margins, perturb-
ing the motility of other cells (Ascione et al., 2016; Cochet-Escartin
et al., 2014). Additionally, the migrating surface can be damaged in the
scraping process, leading to preferential paths in cell movements (Chou
u
et al., 1995). Furthermore, the relative cell confluence ( û ) is challenging
to control and reproduce even within the same culture plate and among
different cell samples (Doran et al., 2009; Jin et al., 2016). The
difficulty to obtain the same cell density in the samples can arise from
inaccuracy in cell counting, as well as from anisotropies in the spatial
spreading of the cells, mainly due to uneven cell adhesion in the culture
plate. Several biological treatments, including gene silencing, transfec-
tion, treatment with drugs at high concentration, require aggressive
techniques, which may result in undesired detachment of cells from the
plate. The final result is a not negligible variability in cell densities
among the samples. Overall, these sources of variability might influ-
ence the outcome of the WH assay making it difficult to compare
independent experiments. Furthermore, the sensing of chemotactic
cues released by cells lying on the other wound edge may also play a key
role in the healing process, the related signaling being also dependent
on the number of cells in the sample.
In this work, we revisited the methodological approach typically
used to quantify the wound closure dynamic in a WH assay, by using a
novel analysis approach based on transport phenomena concepts. In
particular, we explored the influence of cell density on the wound
closure dynamics, and show that it strongly affects the reproducibility
of the WH assay. We performed in vitro WH assays on HT1080 human
fibrosarcoma cells, which exhibit fibroblast-like dynamic behavior, by
using TLM image acquisition. Overall, our work is addressed to
overcome, at least in part, the limitations related to the conventional
quantitative analysis of the WH assay. In particular, we propose a
phenomenological scaling of the experimental data, based on a trans-
port phenomena approach, in order to account for the effect of cell
density on wound closure velocity. The approach we propose suggests
an easy way to describe complex phenomena, such as WH, where
several biological (Matsubayashi et al., 2004) as well as physical
mechanisms (Cochet-Escartin et al., 2014; Ladoux, 2009; Salm and
Pismen, 2012; Tambe et al., 2011) are involved.
2. Materials and methods
2.1. Cell cultures
HT1080 human fibrosarcoma cells were maintained at 37 °C in
Dulbecco's Modified Eagle Medium (Lonza, Switzerland) supplemented
with 10% fetal bovine serum (Lonza, Switzerland) and antibiotics (50
units/mL penicillin and 50 µg/mL streptomycin) (Lonza, Switzerland)
in a humidified atmosphere containing 5% CO2 in air. Cultures between
passages 18 and 22 were used for WH experiments.
2.2. Time-lapse microscopy
Time-lapse Microscopy (TLM) acquisitions were carried out using
an inverted microscope (Zeiss Axiovert 200; Carl Zeiss, Jena, Germany)
and a 10× objective (CP Achromat Ph1) in phase contrast (Caserta
et al., 2013; Vasaturo et al., 2012). The microscope was enclosed in a
homemade incubator keeping the sample at 37 °C and under 5% CO2,
humidified atmosphere. Live-cell imaging was performed by a high-
sensitivity CCD camera (Orca AG; Hamamatsu, Japan) driven by
homemade control software in Labview. The images were iteratively
acquired at several locations (fields of view) within the samples using a
motorized x–y stage and focus control. For each cell concentration
investigated, at least three independent fields of view were selected in
F. Ascione et al. Chemical Engineering Science 160 (2017) 200–209

two wells of the multiwell culture plate. The delay between two
consecutive images of the same field of view was set to 10 min; the
overall experiment duration was about 60 h. For each field of view, two
images visualizing a region of the cell monolayer up to 700 µm from the
cell front on the left and the right wound edge were acquired before
starting TLM acquisitions.

2.3. WH assays

HT1080 fibrosarcoma cells were plated in duplicate on uncoated

24-well culture dishes at different cell densities: 34*10−4, 17*10−4,
8.5*10−4, 4.3*10−4 and 1.7*10−4 cells/μm2. In order to allow the cells
to adhere and grow, they were incubated at 37 °C for about 24 h. The
samples were then scratched manually with a p200 pipette tip to scrape
away an area of cells. Cell debris were then removed washing the
samples twice with phosphate-buffered saline (PBS) and the wounded
cell samples were covered with fresh culture medium.
In one of the experimental sets, 5 µg/ml of mitomycin-C (Sigma-
Aldrich, Australia) were added to the culture medium after the
scratching, to inhibit cell proliferation (Schreier et al., 1993). In order
to determine the optimal concentration of mitomycin-C to block cell
proliferation without affecting cell viability, we investigated the
changes in the number of cells 24 h after the treatment with various
concentrations of mitomycin-C (0, 5, 10 and 15 µg/ml). Mitomycin-C
at 5 µg/ml blocked cell proliferation without decreasing cell number,
while mitomycin-C at 10 and 15 µg/ml was cytotoxic. This finding is in
agreement with previous works (Kanazawa et al., 2010).

2.4. Image analysis

The size of the cell-free area (A), representing the wound, was
measured by using a homemade automated image analysis algorithm
that performed image segmentation for each time step. The algorithm
was able to identify the contour of the wound edges, which showed
finger-like shape due to the presence of scattering cells on the
advancing front. For some cell samples showing very low density, the
size of the wound area was measured by drawing manually the contour
of the scratch region for 8 time values, equally spaced along the entire
experiment length. We considered the wound closed when the area
approached values negligible, i.e. comparable to the size of a few cells.
The wound width (b) was estimated considering the area of an
equivalent rectangle: for each time point, b was calculated as the ratio
between measured cell-free area A and the height of the image
(h=628 µm in all the experiments). The position of the invading cell
b −b
front was calculated as x= 02 , where b0 is the wound width at time 0.
A schematic representation of the measured parameters is reported in
Fig. 1, which shows two images of a WH experiment corresponding to
t=0 h and t > 0 h. The dotted lines indicate the wound edge on the right
and left side of the scratch, whereas the double arrowed line shows the
wound width. x=0 is defined as the cell front on the left edge of the
wound at time 0.
Cell density at time 0 (um) was measured by manually counting the
cells in the images acquired on the left and right wound edge before
starting TLM image acquisition (see above). The density value was
obtained by dividing the number of cells by the occupied area, as
manually measured (Image Pro Plus). At least 50 cells were counted on
each side, typical value being 150 cells (per side) over an area of about
1.8*105 µm2.
For three samples, cell density profile in space over time during the
wound closure experiment was analyzed with major details, by
measuring cell density at various distances x behind the cell front.
Specifically, the image area was divided into rectangular regions with
height h (image height, Fig. 1) and length 30 µm over the x direction.
For each region the cells were counted and the density calculated as the
ratio between the cell number and the area. This measurement was
Fig. 1. Images acquired during a WH experiment at the time t=0 and for a generic t > 0.
The area of the wound at t=0 (A0) and at t > 0 (A) is shown in black. The height of the
image (h) and the wound width (b0 and b for t=0 and t > 0, respectively) are also
indicated. x (t) is the position of the invading cell front, which changes in time along the x
direction. Scale bar=100 µm.
taken at three time points from the scratching, corresponding to a
wound closure of 0%, 50% and 100%.
3. Results and discussion
HT1080 human fibrosarcoma cells were plated in duplicate (two
independent wells) on a 24-well culture dish at different densities up:
34*10−4, 17*10−4, 8.5*10−4, 4.3*10−4 and 1.7*10−4 cells2 (Fig. 2). The
μm
cells were allowed to spread and grow for 24 h before the scratching.
For each cell density, at least three independent fields of view were
selected in TLM experiments. The cell density um at time 0, measured
according to the procedure described in the Materials and Methods
section, can be considered as an estimate of the local cell density in the
wound region; this data is only minimally affected by the damage
caused by the scratching. Moreover, we estimated fluctuations in cell
density over the monolayer in each sample by measuring the difference
in the density on the left and the right edge of the wound. The
discrepancy we measured between the two edges was about 18%. It's
worth mentioning that a general trend can be observed; in fact, at high
density the discrepancy is lower than 10%, it increases in correspon-
dence of intermediate cell density values, reaching in some cases a
value of about 50% for samples showing low cell density.
In Fig. 2, the images of the samples at t=0, acquired within 45 min
after the scratching process, are compared. On each line of the table,
images of 3 samples with the same value of the nominal plated cell
density up are shown; up and the average value of the three measure-

202
F. Ascione et al. Chemical Engineering Science 160 (2017) 200–209

Fig. 2. Images of cell samples acquired at t=0, corresponding to different initial cell densities. up is the plated cell density, um is the cell density of the sample at time 0 measured by
image analysis, according to the protocol described in the methods section. um is the average value of the measured cell density for samples corresponding to the same up. Scale
bar=100 µm.

ments of the density um are reported in the second column of the table,
while the individual value of the measured density um is overlaid on
each image of the table. It is evident that the same nominal value of
plated cell density up corresponds to different values of cell density um.
For example, for up=4.3*10−4 cells2 we obtained three different values of
μm
um: 5.2*10−4, 6.3*10−4 and 7.2*10−4 cells
, that are relative to a mean
μm2
−4 cells
value um =6.2 ± 1.0*10 ; the fluctuation among data was indicated
μm2
as standard deviation. It's worth mentioning that um was determined
after 24 h from the cell plating. Since the doubling time for HT1080
fibrosarcoma cells is 20–24 h (Hoffman et al., 2006; Ohizumi et al.,
1995), it should be expected that the cell density before the scratching,
i.e. after 24 h from cell plating operation, was almost doubled with
respect to the plated cell density up, especially for cell samples with low
density. In Fig. 3, the measured cell density um relative to the same
data as in Fig. 2, is reported as a function of the plated cell density up.
Two limit cases can be considered; specifically, if all the cells plated in
each well 24 h before starting the assay were adhered on the surface
and no proliferation occurred, we would expect all data points to lie on
the bisector (continuous line), i.e. um=up. Otherwise, if all the cells were
seeded on the culture dish and underwent cell division, the data points
in Fig. 3 should be located on the dashed line, corresponding to
um=2 up. Fig. 3 shows that the efficiency of cell adhesion and duplica-
tion process is not uniform for the cell samples under investigation: the

203
F. Ascione et al. Chemical Engineering Science 160 (2017) 200–209

and 3.6*10−4 cells

Fig. 3. Cell density (um) measured by image analysis, reported as a function of the plated
cell density (up). The continuous line corresponds to um=up, while the dashed line
corresponds to um=2 up.
higher the cell density is, the bigger the distance of the corresponding
, with um = 2.9 ± 0.8*10−4 cells
) are almost doubled with
μm2 μm2
respect to the plated cell density. This different behavior can be due to
the fact that at low density the cells are freely proliferating, because
they are far from confluency, while in cell samples with higher density,
cell proliferation is decreased by contact inhibition (Poujade et al.,
2007) caused by the strong physical constraints. The graph in Fig. 3
also shows that plating the cells at the same cell density in different
independent wells, we obtained different values of cell density um after
24 h, as shown in Fig. 2. The fluctuation we measured for um can lead
to an error up to 30%. The discrepancy we found among the values of
measured cell densities can be due to heterogeneities in the cell density
over the plate surface (where we estimated an average error of about
18%), or to difficulties in controlling cell counting and seeding. This
suggests that the cell density in the region where the scratch is made
can be hard to control. Moreover, up resulted to be an unreliable
parameter, as we found a low correspondence with the effectively
measured cell density. For this reason, henceforward we will use the
cell density measured at time 0 (um) to identify the samples.
During WH assays, cell samples were imaged by TLM, in order to
capture the wound closure process. The cell density profile in space
(along the direction of wound closure x) and time was analyzed for
three samples corresponding to different cell densities um: 27*10−4 cells2 ,
μm
12*10−4 cells
, and 3*10−4 cells
. In Fig. 4(A), (B), and (C) cell density
μm 2 μm 2

data points from the dashed line, which represents the ideal case. For evolution in space was reported for three increasing time steps from the
example, for up =34*10−4 , cells2 we obtained three different values of um scratching, corresponding to a wound closure of 0%, 50% and 100%. In
μm
the first and second time step, the curves describing cell density
(27*10−4, 29*10−4 and 44*10−4 , with um = 33 ± 9.3*10−4
cells cells
), that evolution in space exhibit the form of a wave, due to the presence of
μm2 μm2
are far from the double of the plated cell density. Conversely, for a cell density gradient on the wound edge. Figs. 4(A) and (B) show that
up=1.7*10−4 cells2 , the values of um we measured (2.0*10−4, 3.0*10−4 cell density on the edges of the wound does not change significantly
μm

Fig. 4. Cell density profile in space (along the x direction) and time, for cell samples showing different cell density at time 0, as measured by image analysis: 27*10−4 cells
(A), 12*10−4
μm2
cells −4 cells
(B) and 3*10 (C). For each sample, cell density profile in space is reported for three different times from the scratching, corresponding to 0%, 50% and 100% of wound closure,
μm2 μm2
respectively.

204
F. Ascione et al. Chemical Engineering Science 160 (2017) 200–209

Fig. 5. Dynamic evolution of wound closure. Evolution in time of the wound area A,
normalized with respect to the initial value A0 (A), and of the position of the invading cell
front x (B) for cell samples with different measured densities at time 0. In the legend the
value of the cell density at time 0 is reported for each data series. Fig. 6. The wound closure velocity (α) and the velocity of cell front progression (v) are
plotted as a function of the measured cell density at time 0. Both trends are approximated
as linear, R2=0.94 and 0.85, respectively.
over time, the steady state density profile tending to a constant, close to
the density at the edge at t=0. In Fig. 4(C) cell density on wound edges
been here introduced to assess the validity of the analysis and verify the
increases significantly over time, and the density profile at the final
proposed results over an extended range of cell densities. This
time, corresponding to 100% wound closure, is higher than the edge
enhances the generality of our finding, that can be applied also in
value at t=0. The different behavior we observed can be related to the
extreme experimental conditions, where invasive protocols, or unpre-
different cell density of the samples at time 0, respect to the confluency
dictable sources of error, make it difficult and ineffective to achieve the
(û), which results in different proliferation rate, as we discussed above.
goal of having cells at confluency before scratching.
It is worth mentioning that the presence of isolated cells in the wound
The curves describing the evolution of A/A0 in time show a linear
area can influence the measurement of the local cell density um(x), but
trend (Eq. (2)); the slope of the lines represents the wound closure
only marginally the measurement of the cell nude area, that we used to
velocity (α):
measure the wound closure kinetic.
The wound closure kinetics was determined for 12 cell samples A
=1 − α*t
corresponding to different measured densities um in the range 2*10−4 - A0 (2)
29*10−4 cells2 , by measuring for each sample the reduction of the cell-
μm Analogously, the curves showing the progression of x over time also
free area (A/A0) over time (Fig. 5(A)). We also quantified the wound
exhibit a linear trend, with the slope corresponding to the velocity of
closure process determining the evolution over time of the invading cell
cell front propagation (v) (Eq. (3)):
front position (x ) (Fig. 5(B)), calculated according to the procedure
described in the Matherial and Methods section (Fig. 1). It's worth x=v*t (3)
noting that the curves in Fig. 5(B) reach different final values of x , that
This suggests that the wound clusure process occurs at constant
correspond to differences in the initial wound size b0 . Overall, the
2 speed. It's worth mentioning that the two velocities are linked to each
graphs in Fig. 5 highlight a strong discrepancy in the wound closure
other by a simple geometrical relationship: v=(α*b0)/2 . Looking at the
time for the different cell samples, ranging from 10 h to about 60 h.
data more in detail, the speed appears to be constant after a short
This relevant variation can be attributed to the differences in the cell
latency time, of about 1 h, that we are here neglecting. This result is in
density of the samples before the scratching, all the other conditions
agreement with the Fisher-Kolmorgoroff model, which predicts that the
being the same. It can be noticed that the curves in Fig. 5 correspond-
progression of the invading cell front can be observed in terms of a
ing to cell samples with very low initial density, show only few data
traveling wave that propagates with constant speed after a short latency
points, because in these cases the measure of the wound size was taken
time (Maini et al., 2004b).
manually for a limited number of time steps, as described in the
In Fig. 6(A) and (B), the wound closure velocity (α) and the velocity
Material and Methods section. It's worth mentioning that these
of cell front propagation (v) respectively, measured by linear fit of each
samples with extremely low initial density (um < 3.6*10−4 cells2 ) have
μm data set reported in Fig. 5(A) and (B), are reported as a function of the

205
F. Ascione et al.
Fig. 7. Dynamic evolution of wound closure after data scaling. Evolution in time of the
wound area A, normalized with respect to the initial value A0 (A), and of the position of
the invading cell front x (B) for cell samples with different measured densities at time 0.
Symbols are in agreement with the legend in Fig. 5.
cell density um measured at time 0. A linear relationship between both
α and v and the cell density um can be observed (Eq. (4), Eq. (5)), β and
γ being the coefficients of proportionality:
α=β*um (4)
v=γ*um (5)
We provided a phenomenological and a rational explanations of the
proportional relationship we found. From a phenomenological point of
view, the wound closure process is depending on two mechanisms, i.e.,
cell motility and proliferation. They are both intimately linked to cell
density. In fact, the higher the cell density is, the higher the number of
motile cells that can contribute to cover the cell-free area. On the other
hand, a higher cell density corresponds to a higher number of cells
which can undergo proliferation under the stimulus of empty space
creation. Moreover, the data in Fig. 6 show a sigmoidal trend, high-
lighting two saturation effects, due to two extreme situations we have
investigated. The first one is at high values of um, a condition of
overconfluence, where cell density is very high, and both cell motility
and proliferation are reduced by contact inhibition and crowding
effects. This cell density value can be roughly estimated calculating
the maximum density we can expect, considering uncompressed round
cells with a diameter of 20 µm, that is the average value we measured in
our samples. In this case, neglecting any shape factor, or possible cell
compression, the estimated packing density, calculated as the recipro-
cal of cell area, is about 3*10−3 cells2 . As cell density approaches this
μm
value, interactions among cells become relevant, and the motility of
each cell can be expected to deviate from the pure diffusional trend
typical of isolated cells. The second saturation condition is relative to
very low values of cell density, where the samples are very far from the
206
Chemical Engineering Science 160 (2017) 200–209
confluency condition necessary to perform a WH assay. In this case the
process kinetic is very slow, and also appears to be only minimally
influenced by reductions of um below a critical value. The value of cell
density where this saturation occurs (about 3*10−4 cells2 ), corresponds to
μm
the condition where only one cell is counted in a region 10 times the
cell area. However, in our analysis we neglected these deviations,
approximating the dependence to a simple linearity.
On top of the above phenomenological explanation of the propor-
tional relationship, we found a rationale description of the dependence
of the wound closure velocity from the cell density, based on transport
phenomena concepts. The Fisher-Kolmorgoroff model mathematically
describes the wound closure process as sum of two contributions, cell
motility and proliferation. Cell motility, mathematically described as
diffusion, is induced by a flow of cells that according to the Fickian
approach diffuse from areas with high cell density to areas with low
density. As first approximation the cell density gradient can be
estimated from the difference between the density at the wound edge
and the density in the cell-free area, that is 0. In this scenario, the
driving force for the diffusive term is substantially proportional to the
cell density within the monolayer um. The proliferation contribution,
described by the last term of the Fisher-Kolmorgoroff model, includes a
proliferation rate not monotonic with cell density. The proliferation
rate active for the wound closure depends on the local cell density on
the wound edge. Cells on the sheet front face the cell monolayer on one
side, but the nude area on the other side; so their behavior is not driven
by um, that is the value measured in the bulk of the monolayer, but to
some average between um and the density in the scratch region (that is
0). As a consequence, the effective cell density the cells feel on the
wound edge can be estimated as about half of um. This suggests that the
proliferation rate grows with um in an almost linear way, for every value
of um that is not higher than cell density at confluency, û . In this
scenario, both the motility and the proliferation term of the Fisher-
Kolmorgoroff model are proportional to the cell density in the
monolayer. The analysis we proposed suggests why the wound closure
velocity (α) and the velocity of cell front propagation (v) are linearly
dependent with the cell density. A rigorous mathematical demonstra-
tion of the effective dependency should include the influence of the
saturation effects, and in particular the influence of cell interactions on
motility, that is expected to deviate from a simple diffusion model as
cells approach confluence.
The linear relationship between A/A0 (or x ) and the time (Eq. (2),
Eq. (3)), and between the wound closure velocity α (or the velocity of
cell front progression v) and the cell density um (Eq. (4), Eq. (5)),
suggests a phenomenological scaling of the experimental data. From
Fig. 5(A) we can observe that the same generic value of the closure
parameter (A/A0)* is reached at two different time values (t1 and t2 ) for
two different data sets corresponding to two different cell densities (u1
and u2). Considering that for both samples A/A0 linearly decreases with
the time t ( Eq. (2)), we obtain that (A/A 0)*=1 − α1 *t1=1 − α2 *t2 , α1 and
α being the wound closure velocity for the two samples. Considering
that α is directly proportional to cell density u (Eq. (4)), we obtain
(A/A 0)*=1 − β*um1*t1=1 − β*um2 *t2 . Similar calculations can be made
t u
also in terms of x and v. From this relationship, we find that t1 = um2 . To
2 m1
scale the curves with respect to a reference one, it is possible to apply a
scaling on time. The scale factor is the ratio between the cell density at
time 0 of each given sample (umi) and the cell density at time 0 of a
reference sample (umR). It's worth mentioning that the value of the
wound closure time, calculated after data scaling, is determined by the
reference cell density (umR), that can in principle be any value, also
arbitrarily chosen. As a consequence, in order to compare different
samples or biological treatments it is important to scale all the data
respect to the same reference cell density. The scaled time tc can be
calculated with the following equation (Eq. (6)):
um i
tc = *t
um R (6)
F. Ascione et al.
Fig. 8. Direct comparison of wound closure dynamics before (A and C) and after (B and D) data scaling. Evolution in time of the wound area A, normalized with respect to the initial
value A0, and of the position of the invading cell front x for cell samples with different measured densities at time 0. Each data point represents the average of all the measurements
reported in Figs. 5 and 7, respectively, and the standard deviation is reported as error bars. The black point indicates 50% of wound closure.
In other words, the wound closure time of a sample with a given
density (umi) can be scaled to the closure time relative to a reference
density (umR) using Eq. (6).
In Fig. 7 we report the evolution in time of the cell-free area
reduction (A/A0) (Fig. 7(A)) and the position of the invading cell front
(x ) (Fig. 7(B)) after data scaling. Comparing Fig. 5 and Fig. 7, it can be
observed that after the scaling the curves are significantly closer to each
other.
In Fig. 8 a direct comparison between the experimental data before
and after data scaling is reported. Each data point represents the
average of all the measurements taken for each sample. As a measure of
the reproducibility error, related to biological fluctuations, the standard
deviation is reported as error bars. In Fig. 8(A) and (B) the mean value
of A/A0 is reported as a function of time before and after data scaling,
respectively. In Fig. 8(C) and (D) the mean value of the position of the
invading cell front (x ) is reported as a function of time before and after
data scaling, respectively. The large error bars we observe in Fig. 8(A)
and (C) indicate the high discrepancy in the time observed in the raw
data, due to the different cell densities we measured. For instance,
considering a wound closure of 50%, indicated with a black point in
Fig. 8(A) and (B), raw data show A/A0=0.5 ± 0.3. Such an error bar
(60% of the measured value) makes almost impossible to compare
different data sets, and strongly limits the reliability of the WH assay as
a quantitative and comparative research tool. The same condition,
relative to a wound closure of 50%, after data scaling reduces the error
bar to 0.1, i.e., 3 times less compared to the raw data (corresponding to
an error about 20% of the measured value). In this case, the reduced
size of the error bars and the improved experimental reproducibility
make much more reliable the comparison between different experi-
mental conditions or different sample types. The value of the residual
error we measured (about 20% for a closure of half the wound size) is
207
Chemical Engineering Science 160 (2017) 200–209
close to the fluctuation we measured in the cell density over each side
of the wound (18%), suggesting that a fine tuning of the proposed
analysis could still provide further improvement of data correction. We
should also consider that in our experiments we stressed the density
dependence probably above the standard variability usually observed
during standard experiments. In general, the residual error we
observed after data scaling can be related to fluctuations in the local
cell density within the monolayer, but also to other possible biological
sources of error we mentioned in the Introduction.
Our analysis can be considered as a proof that the diffusion-
reaction approach is a good representation of the real process. This
suggests also a way to estimate which of the two terms in the model is
dominant, by calculating the Thiele modulus (ϕ), defined as the ratio
between the characteristic time of the migration and proliferation
processes. As first approximation the characteristic time of the
proliferation process can be estimated from the duplication time τ of
HT1080 cells, which is known to be about 24 h (Rasheed et al., 1974).
This approximation corresponds to neglect the values of the plated cell
density (um) and of the cell density at confluency (û), and approximate
the proliferation rate as a simple first order kinetic with a constant
k=ln2(τ). The resulting expression of the Thiele modulus is:
b0 k
ϕ=
2 D (7)
In this equation, b0 is the wound width, representing the character-
istic length along which cell diffusion occurs, D is cell diffusivity, that
can be estimated from random cell motility experiments (Ascione et al.,
2014). In our case ϕ≅3, suggesting that cell motility is the dominant
mechanism, thus meaning that the contribution of cell proliferation to
the wound healing process can be, as first approximation, neglected.
F. Ascione et al.
Fig. 9. Dynamic evolution of wound closure before (A) and after (B) data scaling for
samples (with different measured cell density at time 0) in the WH experiment
performed in presence of mitomycin-C. Each data point in the inset of the graphs
represents the average of all the measurements; the standard deviation is reported as
error bar.
Fig. 10. The wound closure velocity (α) is plotted as a function of the cell density at time
0, for the WH experiment performed in presence of mitomycin-C.
It's worth mentioning that information about which is the dominating
mechanism in the wound healing process, and how to determine it, is
still lacking in the literature. Some methods proposed to identify the
leading mechanism are based on numerical simulation (Sengers et al.,
2007). The approach here proposed is based on the simple evaluation
of a non-dimensional parameter.
In order to verify that the wound closure process is dominated by
the cell motility mechanism, we performed the same WH experiment
treating the cells with mitomycin-C to inhibit cell proliferation, and
208
Chemical Engineering Science 160 (2017) 200–209
repeated the scaling analysis. From the model point of view, in this
experiment the Thiele modulus is 0. Fig. 9 shows the evolution in time
of the cell-free area reduction (A/A0) before (Fig. 9(A)) and after
(Fig. 9(B)) data scaling. In the inset, the average values of A/A0 are
shown as a function of time and the standard deviation is reported as
error bar. In Fig. 10, the wound closure velocity (α) is shown as a
function of the cell density at time 0, in a range 0.7*10−4–20*10−4 cells2 .
μm
Analogously to Fig. 6, the two saturation effects are also visible, due to
the two extreme situations we have investigated. Also in this case, after
the scaling the curves are closer to each other, with a relevant
improvement of the experimental reproducibility, as suggested by the
shorter error bars in the insets in Fig. 9(B). Even so, also in this case a
residual error can be observed after the scaling. We can state that the
proposed scaling successfully applies independently whether the
wound closure is driven by cell motility or cell proliferation.
4. Conclusions
In this work we revisited the methodological approach typically
used to quantitatively investigate the wound closure dynamics in the in
vitro WH assay. We focused on the influence of cell density on the
wound healing process. The difficulty to obtain homogeneous cell
density among the samples can be due to inaccuracy in cell counting
and seeding, uneven cell adhesion to the culture dish, anisotropies in
the spatial spreading of the cells, as well as to aggressive biological
treatments, that may result in unwanted detachment of an amount of
cells from the culture plate. This may represent a limit in accomplish-
ing reproducible and reliable quantitative results, thus making it
difficult to compare independent experiments.
Our work is addressed to overcome, at least in part, the limitations
related to the conventional quantitative analysis of the WH assay. We
performed in vitro WH assays on HT1080 human fibrosarcoma cells by
using time-lapse microscopy image acquisition. In order to investigate
the effect of cell density on the wound closure process, we considered
samples with different densities (from very low values to complete
surface occupation). We measured the wound closure velocity and
found a linear relationship with cell density. This correlation was
explained by analyzing wound closure as a diffusion-reaction process,
according to the Fisher-Kolmorgoroff model, which describes the
wound closure process as the result of cell motility and proliferation
mechanisms. Following this scenario, we suggested a phenomenologi-
cal scaling of the experimental data, to account for the effect of cell
density on wound closure velocity. In particular, the wound closure
time of a sample with a given density can be scaled to the closure time
relative to a reference (arbitrarily chosen) density. We also proposed a
simple way to estimate whether the wound closure process is driven by
cell motility or proliferation, by the simple calculation of a non-
dimensional parameter.
The proposed approach can improve experimental reproducibility
and make much more reliable the comparison between different
experimental conditions or different sample types. We demonstrated
that a simple transport model efficiently describes cell dynamic
behavior during a high complex preocess, such as wound healing,
where several biological as well as physical mechanisms are involved.
The interpretation of cell dynamic processes according to transport
phenomena concepts can represent a new fronteer for the application
of chemical engineering approach to several biological and biomedical
fields, including drug delivery, cancer invasion, and tissue regenera-
tion.
Acknowledgements
This work has been done under the umbrella of COST MP1106
Smart and Green interfaces for biomedical and industrial applications.
F. Ascione et al.
References
Andújar, I., Ríos, J.L., Giner, R.M., Recio, M.C., 2013. Shikonin promotes intestinal
wound healing in vitro via induction of TGF-β release in IEC-18 cells. Eur. J. Pharm.
Sci. 49, 637–641.
Arnold, J., Adam, J., 1999. A simplified model of wound healing - II: the critical size
defect in two dimensions. Math. Comput. Model. 30, 47–60.
Ascione, F., Caserta, S., Perris, R., Guido, S., 2014. Investigation of cell dynamics in vitro
by time LapseMicroscopy and image analysis. Chem. Eng. Trans. 38, 517–522.
Ascione, F., Vasaturo, A., Caserta, S., D'Esposito, V., Formisano, P., Guido, S., 2016.
Comparison between fibroblast wound healing and cell random migration assays in
vitro. Exp. Cell Res.
Bindschadler, M., McGrath, J.L., 2007. Sheet migration by wounded monolayers as an
emergent property of single-cell dynamics. J. Cell Sci. 120, 876–884.
Block, E.R., Matela, A.R., SundarRaj, N., Iszkula, E.R., Klarlund, J.K., 2004. Wounding
induces motility in sheets of corneal epithelial cells through loss of spatial
constraints: role of heparin-binding epidermal growth factor-like growth factor
signaling. J. Biol. Chem. 279, 24307–24312.
Cai, A.Q., Landman, K.A., Hughes, B.D., 2007. Multi-scale modeling of a wound-healing
cell migration assay. J. Theor. Biol. 245, 576–594.
Caserta, S., Campello, S., Tomaiuolo, G., Sabetta, L., Guido, S., 2013. A methodology to
study chemotaxis in 3-D collagen gels. Aiche J..
Chou, L., Firth, J., Uitto, V., Brunette, D., 1995. Substratum surface-topography alters
cell-shape and regulates fibronectin messenger-RNA level, messenger-RNA stability,
secretion and assembly in human fibroblasts. J. Cell Sci. 108, 1563–1573.
Cochet-Escartin, O., Ranft, J., Silberzan, P., Marcq, P., 2014. Border forces and friction
control epithelial closure dynamics. Biophys. J. 106, 65–73.
Doran, M.R., Mills, R.J., Parker, A.J., Landman, K.A., Cooper-White, J.J., 2009. A cell
migration device that maintains a defined surface with no cellular damage during
wound edge generation. Lab Chip 9, 2364–2369.
Grasso, S., Hernandez, J., Chifflet, S., 2007. Roles of wound geometry, wound size, and
extracellular matrix in the healing response of bovine corneal endothelial cells in
culture. Am. J. Physiol.-Cell Physiol. 293, C1327–C1337.
Hoffman, A., Greene, J.J., Spetner, L.M., Burke, M., 2006. Redox-mediated bypass of
restriction point via skipping of G1pm. Theor. Biol. Med Model 3, 26.
Horwitz, R., Webb, D., 2003. Cell migration. Curr. Biol. 13, R756–R759.
Jin, W., Shah, E.T., Penington, C.J., McCue, S.W., Chopin, L.K., Simpson, M.J., 2016.
Reproducibility of scratch assays is affected by the initial degree of confluence:
experiments, modelling and model selection. J. Theor. Biol. 390, 136–145.
Johnston, S.T., Simpson, M.J., McElwain, D.L., 2014. How much information can be
obtained from tracking the position of the leading edge in a scratch assay? J. R. Soc.
Interface 11, 20140325.
Kanazawa, S., Fujiwara, T., Matsuzaki, S., Shingaki, K., Taniguchi, M., Miyata, S.,
Tohyama, M., Sakai, Y., Yano, K., Hosokawa, K., Kubo, T., 2010. bFGF regulates PI3-
kinase-Rac1-JNK pathway and promotes fibroblast migration in wound healing.
PLoS One 5, e12228.
Keese, C., Wegener, J., Walker, S., Giaever, L., 2004. Electrical wound-healing assay for
cells in vitro. Proc. Natl. Acad. Sci. USA 101, 1554–1559.
Kiehart, D., Galbraith, C., Edwards, K., Rickoll, W., Montague, R., 2000. Multiple forces
contribute to cell sheet morphogenesis for dorsal closure in Drosophila. J. Cell Biol.
149, 471–490.
Kramer, N., Walzl, A., Unger, C., Rosner, M., Krupitza, G., Hengstschlaeger, M., Dolznig,
H., 2013. In vitro cell migration and invasion assays. Mutat. Res.-Rev. Mutat. Res.
752, 10–24.
Ladoux, B., 2009. Biophysics Cells guided on their journey. Nat. Phys. 5, 377–378.
Lauffenburger, D., 1989. A simple-model FOR THE effects OF receptor-mediated cell-
substratum adhesion ON cell-migration. Chem. Eng. Sci. 44, 1903–1914.
Lauffenburger, D.A., Horwitz, A.F., 1996. Cell migration: A physically integrated
molecular process. Cell 84, 359–369.
Liang, C.-C., Park, A.Y., Guan, J.-L., 2007. In vitro scratch assay: a convenient and
inexpensive method for analysis of cell migration in vitro. Nat. Protoc. 2, 329–333.
Love, J.C., 2010. Integrated process design for single-cell analytical technologies. Aiche J.
56, 2496–2502.
Lucena, S., Sanchez, E.E., Perez, J.C., 2011. Anti-metastatic activity of the recombinant
disintegrin, r-mojastin 1, from the Mohave rattlesnake. Toxicon 57, 794–802.
Maini, P.K., McElwain, D.L.S., Leavesley, D., 2004a. Travelling waves in a wound healing
assay. Appl. Math. Lett. 17, 575–580.
Maini, P.K., McElwain, D.L.S., Leavesley, D.I., 2004b. Traveling wave model to interpret
a wound-healing cell migration assay for human peritoneal mesothelial cells. Tissue
Eng. 10, 475–482.
Matsubayashi, Y., Ebisuya, M., Honjoh, S., Nishida, E., 2004. ERK activation propagates
in epithelial cell sheets and regulates their migration during wound healing. Curr.
Biol. 14, 731–735.
Murrell, M., Kamm, R., Matsudaira, P., 2011. Tension, free space, and cell damage in a
microfluidic wound healing assay. PLoS One 6, e24283.
209
Chemical Engineering Science 160 (2017) 200–209
Nie, F., Yamada, M., Kobayashi, J., Yamato, M., Kikuchi, A., Okano, T., 2007. On-chip
cell migration assay using microfluidic channels. Biomaterials 28, 4017–4022.
Nikolic, D., Boettiger, A., Bar-Sagi, D., Carbeck, J., Shvartsman, S., 2006. Role of
boundary conditions in an experimental model of epithelial wound healing. Am. J.
Physiol.-Cell Physiol. 291, C68–C75.
Ohizumi, I., Tanemura, M., Kaihoh, S., 1995. A novel styryl diphenylamine derivative
reverts the transformed phenotype of human fibrosarcoma HT1080 cells. Br. J.
Cancer 72, 1219–1223.
Ottino, J.M., 2011. Chemical engineering in a complex world: grand challenges, vast
opportunities. AiChE J. 57, 1654–1668.
Poujade, M., Grasland-Mongrain, E., Hertzog, A., Jouanneau, J., Chavrier, P., Ladoux, B.,
Buguin, A., Silberzan, P., 2007. Collective migration of an epithelial monolayer in
response to a model wound. Proc. Natl. Acad. Sci. USA 104, 15988–15993.
Rasheed, S., Nelson-Rees, W.A., Toth, E.M., Arnstein, P., Gardner, M.B., 1974.
Characterization of a newly derived human sarcoma cell line (HT-1080). Cancer 33,
1027–1033.
Riahi, R., Yang, Y., Zhang, D., Wong, P., 2012. Advances in wound-healing assays for
probing collective cell migration. Jala 17, 59–65.
Rodriguez, L.G., Wu, X., Guan, J.L., 2005. Wound-healing assay. Methods Mol. Biol. 294,
23–29.
Rørth, P., 2009. collective cell migration. Annu Rev. Cell Dev. Biol. 25, 407–429.
Salm, M., Pismen, L., 2012. Chemical and mechanical signaling in epithelial spreading.
Phys. Biol., 9.
Schreier, T., DEGEN, E., BASCHONG, W., 1993. Fibroblast migration and proliferation
during in-vitro wound-healing - a quantitative comparison between various growth-
factors and a low-molecular-weight blood dialysate used in the clinic to normalize
impaired wound-healing. Res. Exp. Med. 193, 195–205.
Sengers, B.G., Please, C.P., Oreffo, R.O., 2007. Experimental characterization and
computational modelling of two-dimensional cell spreading for skeletal regeneration.
J. R. Soc. Interface 4, 1107–1117.
Sheardown, H., Cheng, Y.L., 1996. Mechanisms of corneal epithelial wound healing.
Chem. Eng. Sci. 51, 4517–4529.
Sherratt, J.A., Murray, J.D., 1990. Models of epidermal wound healing. Proc. Biol. Sci.
241, 29–36.
Silano, M., Vincentini, O., Luciani, A., Felli, C., Caserta, S., Esposito, S., Villella, V.R.,
Pettoello-Mantovani, M., Guido, S., Maiuri, L., 2012. Early tissue transglutaminase-
mediated response underlies K562(S)-cell gliadin-dependent agglutination. Pediatr.
Res. 71, 532–538.
Sumagin, R., Robin, A.Z., Nusrat, A., Parkos, C.A., 2013. Activation of PKCβII by PMA
facilitates enhanced epithelial wound repair through increased cell spreading and
migration. PLoS One 8, e55775.
Tambe, D.T., Hardin, C.C., Angelini, T.E., Rajendran, K., Park, C.Y., Serra-Picamal, X.,
Zhou, E.H., Zaman, M.H., Butler, J.P., Weitz, D.A., Fredberg, J.J., Trepat, X., 2011.
Collective cell guidance by cooperative intercellular forces. Nat. Mater. 10, 469–475.
Terryn, C., Bonnomet, A., Cutrona, J., Coraux, C., Tournier, J.-M., Nawrocki-Raby, B.,
Polette, M., Birembaut, P., Zahm, J.-M., 2009. Video-microscopic imaging of cell
spatio-temporal dispersion and migration. Crit. Rev. Oncol. Hematol. 69, 144–152.
Tremel, A., Cai, A., Tirtaatmadja, N., Hughes, B.D., Stevens, G.W., Landman, K.A.,
O'Connor, A.J., 2009. Cell migration and proliferation during monolayer formation
and wound healing. Chem. Eng. Sci. 64, 247–253.
Tsai, M., Huang, H., Hsu, J., Lai, Y., Hsiao, Y., Lu, F., Chang, H., 2014. Topical N-
acetylcysteine accelerates wound healing in vitro and in vivo via the PKC/Stat3
pathway. Int. J. Mol. Sci. 15, 7563–7578.
Vasaturo, A., Caserta, S., Russo, I., Preziosi, V., Ciacci, C., Guido, S., 2012. A novel
chemotaxis assay in 3-d collagen gels by time-lapse microscopy. PLoS One 7, e52251
.
Veltman, D., 2014. Actin dynamics: cell migration takes a new turn with arpin. Curr. Biol.
24, R31–R33.
Vitorino, P., Meyer, T., 2008. Modular control of endothelial sheet migration. Genes Dev.
22, 3268–3281.
Wong, C., Chen, A.A., Behr, B., Shen, S., 2013. Time-lapse microscopy and image
analysis in basic and clinical embryo development research. Reprod. Biomed. Online
26, 120–129.
Yue, P., Leung, E., Mak, N., Wong, R., 2010. A simplified method for quantifying cell
migration/wound healing in 96-well plates. J. Biomol. Screen. 15, 427–433.
Zahm, J.M., Kaplan, H., Hérard, A.L., Doriot, F., Pierrot, D., Somelette, P., Puchelle, E.,
1997. Cell migration and proliferation during the in vitro wound repair of the
respiratory epithelium. Cell Motil. Cytoskelet. 37, 33–43.
Zhang, M., Li, H., Ma, H., Qin, J., 2013. A simple microfluidic strategy for cell migration
assay in an in vitro wound-healing model. Wound Repair Regen. 21, 897–903.
Ziebert, F., Aranson, I.S., 2013. Effects of adhesion dynamics and substrate compliance
on the shape and motility of crawling cells. PLoS One 8, e64511.
Zordan, M., Mill, C., Riese, D., Leary, J., 2011. A high throughput, interactive imaging,
bright-field wound healing assay. Cytom. Part A 79A, 227–232.