modeling the effect of exogenous calcium on keratinocyte ...staff · bonds, number of cells in...
TRANSCRIPT
Modeling the Effect of Exogenous Calcium on Keratinocyte
and HaCat Cell Proliferation and Differentiation
Using an Agent-Based Computational Paradigm
DAWN WALKER, Ph.D.,1 TAO SUN, Ph.D.,2 SHEILA MACNEIL, Ph.D.,2
and ROD SMALLWOOD, Ph.D.1
ABSTRACT
In this study we sought to develop a computational modeling paradigm in order to describe the influenceof calcium on normal and transformed keratinocyte proliferation and differentiation. Keratinocytes andHaCat cells were grown in monolayer cultures with low and physiologic calcium concentrations, andlevels of proliferation and involucrin expression were assessed. Both types of cells grew as monolayersunder a low-calcium environment, and stratified in media with physiologic levels of calcium. However,keratinocytes were more proliferative in low rather than physiologic levels of calcium, whereas theopposite was true for HaCat cells. Normal keratinocytes differentiated as calcium levels increased. HaCatcells showed little differentiation at any calcium concentration. However, while the computer simulationcould be modified to describe the effect of calcium on the growth of normal keratinocytes, our findingsdid not support the hypothesis that simply ‘‘turning off’’ the ability of HaCat cells to differentiate wouldaccount for the growth characteristics of these transformed cells. This demonstrates the application ofcomputational modeling to hypothesis testing in biological systems.
INTRODUCTION
C OMPUTATIONAL MODELS that attempt to describe the
behavior of biological tissue using a set of mathe-
matical equations fail to capture the granularity of the tis-
sue and the behavior of individual cells in the tissue. Our
primary interest is in the interactions of individual cells,
and the resulting emergent tissue structure and function.
We have developed a cell-based model of the growth
characteristics of urothelial cells in monolayer culture in
low or physiologic levels of calcium,1 which gives a qua-
litative fit to in vitro cell behavior. The agent-based model
was also used to predict wound healing in urothelial cell
cultures maintained in low and physiologic exogenous
calcium [Ca2þ]e. This simple rule-based model of cell
behavior, incorporating rules relating to contact inhibition
of proliferation and migration, was sufficient to qualita-
tively predict the calcium-dependent pattern of wound
closure observed in vitro.2
We now extend this approach to normal human kerati-
nocytes, which, unlike urothelial cells, will differentiate
in vitro if conditions permit. In the absence of permissive
levels of calcium ([Ca2þ]e levels lower than 0.5mM),
keratinocytes are unable to stratify and differentiate; rather,
they proliferate rapidly with a high growth fraction. They
can be rapidly induced to terminally differentiate by re-
storing calcium to the level normally found in standard
culture medium (1.2mM).3 When calcium is elevated to
physiologic levels in culture, cell-to-cell contact occurs
rapidly and desmosomes form within 2 h. The cells stratify
1Department of Computer Science, Kroto Research Institute, North Campus, University of Sheffield, Broad Lane, Sheffield, United
Kingdom.2Department of Engineering Materials, Kroto Research Institute, North Campus, University of Sheffield, Broad Lane, Sheffield,
United Kingdom.
TISSUE ENGINEERINGVolume 12, Number 8, 2006# Mary Ann Liebert, Inc.
1
by l–2 days and terminally differentiate with cell sloughing
by 3–4 days.3,4
The HaCat cell line is a spontaneously transformed
(following overheating of an incubator) human epithelial
cell line derived from adult skin. It is immortal (>140
passages), and has a transformed phenotype in vitro (clono-
genic on plastic and in agar) but remains nontumorigenic.
Despite their altered and unlimited growth potential, HaCat
cells have been reported to form an orderly structured and
differentiated epidermal tissue when transplanted onto nude
mice.5
The primary aim of this work was to further develop our
existing computational model in order to describe the in-
fluence of extracellular calcium on the growth and differ-
entiation of normal keratinocytes. The second aim was to
explore the potential of the model to explore a simple hypo-
thesis that we proposed to explain the very different response
to calcium of a spontaneously transformed keratinocyte cell
line (HaCats) compared to normal keratinocytes.
MATERIALS AND METHODS
Cell culture
Normal human keratinocytes (NHKs) were isolated as
described elsewhere.6–8 Defined keratinocyte-serum free
medium (SFM) (Invitrogen, UK; calcium concentration
0.09mM) supplemented with 100 IU/mL penicillin and
100 mg/mL streptomycin was used for both NHK and Ha-
Cat cell cultures. The concentration of calcium in the
medium was increased from 0.09mM to 2mM by the ad-
dition of calcium chloride.
Measurement of cell proliferation
Cell number was assessed using the MTT (3-(4,5-
Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide)
assay (an indirect assay of total cell viability) as described
by Marcus et al.;9 total DNA was measured using Hoechst
fluorescent stain (33258; Sigma-Aldrich, UK) as done by
Higham et al.10 (this measures all cellular DNA regardless
of whether cells are alive or dead) and by using bromo-
deoxyuridine (BrdU) immunostaining, which measures ac-
tive DNA synthesis, as described in a following section.
Cells were incubated in BrdU-labeling medium (Roche
Applied Science, UK) for 4 h, fixed in ethanol:glycine fixa-
tive (3 parts 50mM glycine solution:7 parts absolute etha-
nol) and stored at �208C for 20–30min. After a thorough
wash with phosphate-buffered saline to remove all traces of
fixative, anti-BrdU monoclonal antibody (Roche Applied
Science) was added for 30min at 378C. The cells then were
incubated with fluorescein-conjugated anti-mouse second-
ary antibody for 30min. After counterstaining with 40,6-diamidino-2-phenylindole dihydrochloride (DAPI) for
5min,5 the cells were visualized using an Axon image ex-
press system (Molecular Devices, Sunnydale, CA).
Identification of keratinocytes and assessment
of differentiation
NHK or HaCat cells were immunolabeled with pancy-
tokeratin (which stains all keratinocytes) or involucrin
(which is an indicator of cornified envelope formation and
identifies differentiating cells) and visualized as described
elsewhere.11,12 The extent of involucrin expression was
then related to total DNA to determine the extent of dif-
ferentiation in a population of cells.
All the fluorescent stains were visualized using an Axon
image express system (Molecular Devices). Briefly, fluor-
escence micrographs of immunolabeled samples were taken
using epifluorescent illumination at lex¼ 495 nm, lem¼515 nm (for fluorescein isothiocyanate/pancytokeratin, in-
volucrin, BrdU visualization) and lex¼ 358 nm, lem¼461 nm (for DAPI/nuclei visualization).
The stain was then eluted by incubating the cells in 0.1M
sodium hydroxide for 1 h at 378C, transferred to a cuvette,
and quantified using a fluorescence plate reader (lex¼495 nm, lem¼ 515 nm for fluorescein isothiocyanate/
pancytokeratin, involucrin, BrdU; lex¼ 358 nm, lem¼461 nm for DNA).
Computational modeling
We have developed an ‘‘agent-based’’ paradigm using
Matlab (The Mathworks, Inc., Natick, MA), in which in-
dividual biological cells are represented as computational
entities or ‘‘software agents.’’ These agents live on a 2-
dimensional, square substrate with user-defined dimensions
and modifiable exogenous [Ca2þ]e. The user can select the
number of cells placed either randomly or in specified lo-
cations. Mean cell cycle time and migration rate are scaled
so that each time step in the model represents approxi-
mately 30min in real time. At each time step, agents are
interrogated in turn. The internal parameter state (e.g.,
position in cell cycle, flag indicating attachment to sub-
strate) and environment (number and proximity of neigh-
boring cells, [Ca2þ]e) control the execution of rules that
may change the internal parameters. These rules represent,
for instance, intercellular bonding, changes in morphology,
cell cycle progression, and arrest. If a cell divides, one
daughter cell overwrites the data structure of the parent,
and a second daughter cell is added to the simulation. A
simple ‘‘physical model’’ resolves the intercellular forces
resulting from cell movement, growth, and division in the
plane of the substrate. The general architecture of the
model is shown in Fig. 1.
The current model simulates the behavior of cells in
differing exogenous calcium ion concentrations. In parti-
cular, we have studied the effect [Ca2þ]e on proliferation.
The probability of an intercellular bond being formed be-
tween a pair of modeled cells is inversely proportional to
the separation between the edges and related to [Ca2þ]e via
a sigmoid function with the inflection point at 1.0mM,
based on the relationship between the binding activity of
2 WALKER ET AL.
E-cadherin–mediated bonds and [Ca2þ]e.13 Cells with 4 or
more intercellular bonds enter the G0 phase of the cell
cycle and become quiescent, which results in a lower
growth rate in models with higher [Ca2þ]e. Simulations in
low and physiologic [Ca2þ]e agree qualitatively with
in vitro experiments on urothelial cells.1
The first objective of the current study was to determine
what changes in the model rules and parameters for ur-
othelial cells were needed in order to represent the char-
acteristics of normal human keratinocytes; the second
was to explore the hypothesis that the aberrant behavior
of transformed human keratinocytes (HaCat cells) can
be explained by the inactivation of the differentiation
pathway.
Modeling NHKs. Table 1 summarizes the changes to the
rule sets that were necessary to represent keratinocyte, in
contrast to urothelial cell behavior. In particular, new rule
sets governing differentiation (defined in terms of in-
volucrin expression and cell cycle exit) and stratification
were needed. The general form of these rules (i.e., in-
creased involucrin expression/stratification with increased
exogenous calcium) was constructed on the basis of known
aspects of cell behavior, but the actual parameters defining
the rule sets were tuned to give population level results that
agreed with experimental data. A simpler ‘‘physical’’ al-
gorithm than that described elsewhere1,2 was used, with
cells sequentially shifted a maximum of 5 times per agent
iteration to minimize overlap with their neighbors. This
reduced the solution time without significantly affecting the
results (data not shown).
All model parameters were set to replicate those used in
the in vitro experimental work as closely as possible.
Model simulations were run with exogenous calcium con-
centrations set to 9 values between 0.05mM and 2.5mM.
In the case of the normal human keratinocyte population
growth simulations, or ‘‘virtual MTT assays,’’ the model
substrate size was set at 1�1mm and the number of cells
seeded at the start of the simulation was 400, giving an
initial seeding density of 4�104 cells per cm2.
The minimum cell cycle times in the model were set to
27 h (54 model iterations), and cell migration speed was set
to 20 mm/iteration (equivalent to 0.67 mm/min, a value de-
rived from several observations of cell migration published
in the literature). Simulations applied the agent rules and
simple physics interactions to the seeded cell populations.
At every iteration, statistics relating to the state of the cell
population (e.g., total cell number, number of cells with
bonds, number of cells in different phases of the cell cycle)
were saved.
In the case of the ‘‘virtual involucrin assays,’’ the model
protocol was matched to the in vitro procedure as closely
as possible (i.e., an initial cell population with a density
of 2�105 cells per cm2 was ‘‘expanded’’ for 7 days [350
iterations] in 0.05mM calcium, the model was saved, the
calcium level switched to 1 of the 9 values between 0.05mM
and 2.5mM, and the simulations resumed for a further 5
days [250 iterations]).
To ensure that trends predicted by the model were due to
the cellular level rules and were not significantly affected
by starting conditions or stochastic events based on random
numbers generated during the simulations, each set of mod-
els was run 3 times using different starting conditions and
random starting positions of seeded cells.
Modeling HaCat cells. This transformed cell line shows
increased proliferation compared to normal human kerati-
nocytes. It has been hypothesized that the differentiation
pathway for these cells has somehow been disabled, al-
though the precise mechanism is not known.14 We used our
computational model to examine this hypothesis. In the
HaCat simulations, cells were modeled to not express in-
volucrin, or to permanently leave the cell cycle. All other
rules remain the same as for untransformed cells.
In the case of the HaCat models, both a 0.5�0.5-mm
substrate with 100 seeded cells and a 1�1-mm substrate
with 400 seeded cells were used, giving identical initial
seeding densities to the normal human keratinocyte simu-
lations.
COMMUNICATIONRULES
For everycell agent
APOPTOSISRULES
MOTILITYRULES
BONDINGRULES
MORPHOLOGYRULES
CELL CYCLERULES
Incrementtime, t
DIFFERENTIATIONRULES
STRATIFICATIONRULES
EQUILIBRIATE CELL POSITIONSDUE TO GROWTH, MIGRATION, ETC.
FIG. 1. Flow chart illustrating model structure. Boxes with bold
outline indicate new behavior introduced to model keratinocytes.
EXOGENOUS CALCIUM ON KERATINOCYTE AND HaCat CELL PROLIFERATION 3
RESULTS
In vitro results
Effect of calcium on cell proliferation. Fig. 2 shows
DAPI (part A) and pancytokeratin (part B) immuno-
fluorescence micrographs of keratinocyte distribution
approximately 12 days after plating in defined keratinocyte-
SFM medium with low calcium (0.09mM). In physiologic
calcium (2mM), most of the cells stratified (as shown in
Fig. 2C and D) and the formation of colonies is apparent,
whereas cells are more evenly distributed in the low-
calcium medium.
Similar experiments with HaCat cells are shown in Fig.
2E–H. In the low-calcium media (Fig. 2E and F) cells are
very clearly separated, with no evidence of cell-to-cell
contact. In media with physiologic levels of calcium, there
is more evidence of cell-to-cell contact but the degree of
aggregation or stratification is clearly less than for the
normal keratinocytes if one contrasts Fig. 2G and H with
Fig. 2C and D.
TABLE 1. SUMMARY OF MODEL RULES
Rule type
Variable Urothelium rule Keratinocyte rule Justification
Bonding Cell has 1 in 3 chance of binding
to substrate per iteration.
Probability of bonding to substrate
related to involucrin expression
level of cell via a sigmoidal
relationship.
Keratinocytes are harder to seed
than urothelial cells—
approximately 50% do not
attach after 18 h and die. It is
probably the more differentiated
cells that do not attach.
Apoptosis Cells that have failed to bond to
the substrate after approximately
7 h have 1/50 chance of under-
going apoptosis per iteration.
No change. Cells that fail to attach undergo
apoptosis.
Cell cycle Cells enter cycle immediately
after bonding to substrate.
Cells start to cycle 300 iterations
after bonding to substrate.
Keratinocytes undergo a lag phase
of several days before
commencing exponential growth.
Cell cycle Minimum cell cycle time 15 h. Minimum cell cycle time increased
to 27 h.
Keratinocytes in culture proliferate
more slowly than urothelial cells.
Cell cycle Cells require 4 intercellular
bonds before becoming contact-
inhibited and entering G0.
Cells require 6 intercellular bonds
before becoming contact-
inhibited and entering G0.
Keratinocytes in higher calcium
levels stratify, and thus form
more bonds than urothelial cells.
Cells can continue to cycle even
in stratified colonies.
Differentiation None. Cells are seeded with an initial
involucrin level randomly
distributed between 0 and 1. Once
the cell has bonded to the
substrate and entered the cell
cycle, this expression level is
increased by [Ca2þ]e/500
per iteration until a maximum
level of 1.
Keratinocytes obtained from
patients will be at various stages
of differentiation. Once in cul-
ture, they express involucrin in
a calcium-dependent fashion.
Differentiation None. A cell with an involucrin expression
level of �0.5 will differentiate
and permanently leave the cell
cycle.
Unlike urothelial cells,
keratinocytes differentiate in
cell culture.
Stratification None. Cells subject to compressive forces
from at least 3 neighbors can lose
contact with the substrate, then
ascend to form a suprabasal layer.
The probability of any cell stra-
tifying is related to exogenous
calcium via a sigmoidal function
with the inflection point
at 1.5mM.
In contrast to urothelial cells, which
grow as a monolayer regardless
of exogenous calcium levels,
keratinocytes in higher calcium
stratify before confluence.
Stratification tends to occur only
in exogenous calcium levels>1mM.
4 WALKER ET AL.
Quantification of cell proliferation was assessed with
MTT, DNA, and BrdU assays. Fig. 3A shows the in vitro
MTT assay of normal human keratinocyte growth con-
trasted with that of HaCat cells (Fig. 3B) over a period
of 12–13 days in media with low or physiologic calcium.
For normal keratinocytes, after an initial lag phase of ap-
proximately 7–8 days the cell population of keratinocytes
in the low-calcium media begins to expand rapidly. Some
5 days later, the maximum cell number was reached (a total
of 12 elapsed days). However, the expansion of the cell
population in the physiologic-calcium medium was very
low even after 14 days of plating.
In contrast to the keratinocytes, HaCat cell populations
in physiologic-calcium media began to expand rapidly after
an initial lag phase of approximately 7–8 days, while the
expansion of the cell population in low-calcium medium
remained very low throughout the 12 days of study (as
illustrated in Fig. 3B).
Fig. 4A confirms these data assessing cell proliferation
by showing both total DNA and DNA synthesis levels in
cells over a range of calcium concentrations after 7–8 days
of culture. NHK proliferation was higher in low-calcium
media than in physiologic-calcium media, whereas HaCat
cell proliferation was lower in low-calcium media (Fig. 4B).
FIG. 2. Micrographs of (A) 40,6-diamidino-2-phenylindole
dihydrochloride (DAPI)–stained and (B) pancytokeratin-
immunostained normal human keratinocytes cultured in medium
with low calcium, (C) DAPI-stained and (D) pancytokeratin-
immunostained normal human keratinocytes cultured in me-
dium with physiologic calcium, (E) DAPI-stained and (F)
pancytokeratin-immunostained HaCat cells cultured in medium
with low calcium, (G) DAPI-stained and (H) pancytokeratin-
immunostained HaCat cells cultured in medium with physiologic
calcium. (Bar¼ 100 mm.) Color images available online at
www.liebertpub.com/ten.
Time in days
Tot
al li
ve c
ellc
dens
ity10
3 /m
m2
0.0
2.0
4.0
6.0
8.0
10.0
12.0
0 5 10 15
(D)Time in days
(A)
00
0.05
0.1
0.15
0.2
0.25
0.3
0.35
MT
T O
ptic
al D
ensi
ty
5 10 15
0.09mM Ca2+
2.0mM Ca2+
0 5 10 150
MT
T O
ptic
al D
ensi
ty
Time in days
0.05
0.1
0.15
0.2
0.25
0.3
0.35(B)
0.09mM Ca2+
2.0mM Ca2+
(C)
0 5 10 15
Time in days
0
0.5
1.0
1.5
2.0
2.5
3.0
Tot
al li
ve c
ell c
dens
ity10
3 /m
m2
0.51.01.251.52.02.5
0.050.51.01.251.52.02.5
0.05
FIG. 3. Total cellular viability of (A) normal human keratino-
cytes cultured in medium with low and physiologic levels of cal-
cium and (B) HaCat cells cultured in medium with low and
physiologic levels of calcium. (C) Virtual 3-(4,5-Dimethylthiazol-
2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (i.e., mod-
eled total live cell population) at different calcium concentrations.
Normal human keratinocytes (D) virtual MTT assay, HaCat cells.
Calcium concentration in mM(C) (D)
(A) (B)
0 0.5 1 1.5 2 2.5 32.1
2.2
2.3
2.4
2.5
2.6
2.7
0.91
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
Calcium concentration in mM
BrD
U o
ptic
al d
ensi
ty x
103
DN
A o
ptic
al d
ensi
ty x
104
DNABrDU
0 0.5 1 1.5 2 2.5 30
2
4
6
8
10
12
Tot
al c
ell d
ensi
ty 1
03/m
m2
Ca2+ concentration (mM)
0 0.5 1 1.5 2 2.5 30.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
DN
A o
ptic
al d
ensi
ty x
104
0
1.0
2.0
3.0
4.0
5.0
6.0
BrD
U o
ptic
al d
ensi
ty x
103
DNABrDU
0 0.5 1 1.5 2 2.5 3Ca2+ concentration (mM)
0
0.5
1.0
1.5
2.0
2.0
3.0
Tot
al c
ell d
ensi
ty 1
03/m
m2
FIG. 4. Total population DNA of normal human keratinocytes
(A) and HaCat cells (B) cultured in medium with varying levels of
calcium. Virtual DNA assay (total modeled cell density vs cal-
cium concentration in mM) for normal human keratinocytes (C)
and HaCat cells (D).
EXOGENOUS CALCIUM ON KERATINOCYTE AND HaCat CELL PROLIFERATION 5
Effect of calcium on differentiation of keratinocytes and
HaCats. Figure 5A–D illustrates that NHKs produced more
involucrin in physiologic-calcium media than in low-
calcium media while involucrin expression in HaCat cells
(Fig. 6A–D) remained low whether cultured in physiologic-
calcium media or low-calcium media.
Fig. 7 shows the involucrin expression relative to DNA
content of the cultures for both keratinocytes and HaCats
cultured in media with varying levels of calcium. These
results indicate that keratinocytes in a high-calcium media
clearly produce more involucrin per cell than when cultured
in a low-calcium media (Fig. 7A). In contrast, HaCat cells
showed no evidence of an increase in differentiation with
increasing levels of calcium (Fig. 7B).
Model simulations
Effect of calcium on cell proliferation. Fig. 5E and F
shows plots of cells after simulating a time equivalent to 12
days in 0.05mM and 2.0mM calcium in the normal human
keratinocyte simulations. In low calcium (0.05mM), the
model cells are confluent and evenly spaced, with virtually
no stratification. By contrast, cells in physiologic calcium
(2.0mM) are stratified, although they have not reached
confluence. Fig. 6E and F shows the equivalent plots for the
HaCat simulations. In low calcium (Figs. 5 and 6C) the
results appear very similar. However, in physiologic cal-
cium, close inspection reveal many more cells, forming
several fully confluent layers. No green cells are visible in
Fig. 6C and D, reflecting the imposed restriction that model
HaCat cells do not express involucrin and therefore cannot
terminally differentiate.
Fig. 3C shows the results of the virtual MTT assay (i.e.,
the number of live cells against simulation time) in 9 dif-
ferent calcium environments for modeled NHKs. After an
initial lag period of approximately 1 week, during which
the live cell population decreases slightly (because of the
apoptosis of unattached cells), the cell population in the
lower-calcium NHK simulations increases rapidly, reach-
ing a maximum level after approximately 11 days. In the
physiologic-calcium NHK simulations, the rate of popula-
tion expansion is much slower, with the final live cell
number totaling approximately 35–40% of that for the
lowest calcium. Intermediate-calcium models exhibit be-
havior between that observed in the low and high condi-
tions, with the monotonic trend that increasing calcium
results in a lower growth rate.
The final total cell population (including living and dead
cells) in the different NHK simulations is shown in Fig. 4C.
There is a curvilinear relationship between cell number and
calcium, with the highest total population occurring at the
lowest calcium.
The virtual MTT and DNA results for HaCat cells are
shown in Figs. 3D and 4D. It is apparent that in contrast to
the NHK results, increased calcium is associated with an
increase in population growth. This was observed to be due
to the tendency of cells in the higher-calcium simulations to
stratify, thus effectively increasing the space available for
cell growth. At the lowest calcium level, the model growth
results for NHKs and HaCats are identical.
Effect of calcium on cell differentiation. Inspection of
Fig. 5E and F shows a significantly greater degree of dif-
ferentiation in the higher-calcium NHK model, where
80–90% of the cells are committed (shown in green), in
contrast to the lower calcium case, where only approxi-
mately 10% of cells have permanently left the cycle. In the
latter case, most of this differentiation arises from the fact
that the initial seeded cell population has an involucrin
level randomly distributed between 0 and 1, and although
many of the cells with higher involucrin levels will fail to
attach and hence undergo apoptosis (see rules described in
Table 1), some will survive and differentiate. HaCat cells
have been modeled to not express involucrin and therefore
cannot terminally differentiate; thus, no green cells are
visible in Fig. 6E and F.
Fig. 7C shows the results of the virtual involucrin assay
for NHKs (i.e., the mean involucrin expression per modeled
cell after 7 days growth in 0.05mM calcium and a further 5
days at an adjusted calcium level, similar to the in vitro
assay). It is apparent that much more involucrin is ex-
pressed at calcium levels greater than 1.0mM. Again,
modeled HaCat cells do not express involucrin, so Fig. 7D
is simply a straight line.
DISCUSSION
The specific objective of this study was to find a mini-
mum set of behavioral differences that could explain the
different growth characteristics of human skin epithelial
cells versus human bladder epithelial cells (results outlined
in a previous publication1) in various calcium environ-
ments, and, further, to examine the hypothesis that the
behavior of a keratinocyte transformed cell line could
be replicated simply by ‘‘turning off’’ the differentiation
pathway.
The 2 major adaptations made to the model in order to
model normal keratinocyte-like, as opposed to urothelial-
like, behavior were (1) inclusion of calcium-dependent
differentiation and (2) inclusion of calcium-dependent
stratification. The first rule allowed model cells to increase
the concentration of a hypothetical substance analogous to
the cytoplasmic protein involucrin, known to be associated
with the differentiation process, by a factor linearly related
to the environmental calcium concentration [Ca2þ]e, and to
irreversibly leave the cell cycle (become committed) when
an arbitrary threshold was reached (i.e., for the ith cell):
cell{i}_involucrin¼ cell{i}_involucrinþ [Ca2þ]e/500;
If (cell{i} is a proliferating cell) AND (cell{i}_involucrin>0.5)
cell{i} becomes a committed cell
6 WALKER ET AL.
The second rule allows cells subject to a small com-
pressive force exerted by 3 or more neighbors to move
upwards, with the probability of the cell’s moving being
sigmoidally related to the calcium concentration. A sig-
moidal relationship was selected because intercellular
bonding strength has previously been shown to be related to
exogenous calcium in this way,13 and these bonds are
crucial in the stratification process as they allow a cell to
‘‘climb’’ over its neighbors.
In this case, the rule is slightly more complex because it
also takes into account the involucrin expressed by the cell
(it has been hypothesized that cells lose adhesion to the
substrate as they differentiate15), and the cell radius (a cell
subjected to compressive forces from its neighbors will first
become ‘‘squashed,’’ then stratify as it loses contact with
the substrate):
If (force on cell{i}> 0.1) AND (cell{i} is in contact with at
least 3 others)
radius_probability_index¼minimum radius/cell{i}_current
radius;
differentiation_probability_index¼ cell{i}_involucrin;
FIG. 5. Micrographs of (A) 40,6-diamidino-2-phenylindole di-
hydrochloride (DAPI)-stained and (B) involucrin-immunostained
normal human keratinocytes cultured in medium with low cal-
cium, (C) DAPI-stained and (D) involucrin-immunostained nor-
mal human keratinocytes cultured in medium with high calcium.
(E and F) Plots of modeled normal human keratinocytes after
12 days of simulated time in 0.05mM calcium (E) and 2.0mM
calcium (F). Red: stem cell; blue: transit-amplifying cell; pink:
mitotic cell; green: terminally differentiated cell.
FIG. 6. Micrographs of (A) 40,6-diamidino-2-phenylindole di-
hydrochloride (DAPI)-stained and (B) involucrin-immunostained
HaCat cells cultured in medium with low calcium, (C) DAPI-
stained and (D) involucrin-immunostained HaCat cells cultured in
medium with high calcium. (E and F) Plots of modeled HaCat
cells after 12 days of simulated time in 0.05mM calcium (E) and
2.0mM calcium (F). Red: stem cell; blue: transit-amplifying cell;
pink: mitotic cell.
(A)
(C)
0 0.5 1 1.5 2 2.5 3Calcium concentration in mM
Invo
lucr
in/t
otal
cel
l no.
(D)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5 30
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Calcium concentration in mM
Invo
lucr
in/t
otal
cel
l no.
30 0.5 1 1.5 2 2.5 3
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Calcium concentration in mM
Invo
lucr
in/D
NA
(B)
0 0.5 1 1.5 2 2.50
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Calcium concentration in mM
Invo
lucr
in/D
NA
FIG. 7. The effect of calcium concentration on the differentia-
tion of (A) normal human keratinocytes and (B) HaCat cells
in vitro. (C and D) Virtual involucrin assay–mean involucrin per
cell after 5 days in modified calcium normal human keratinocytes
(C) and HaCat cells (D).
EXOGENOUS CALCIUM ON KERATINOCYTE AND HaCat CELL PROLIFERATION 7
calcium_probability_index¼ tanh(pi*(env_ca� 1.5))þ 1)/
20;
If (product of probability indices> random number)
then cell{i}_layer¼ cell{i}_layerþ 1;
Inclusion of these 2 new rules has a conflicting effect on
cell population growth. The tendency of cells to differ-
entiate more quickly in higher calcium means that more
cells permanently leave the cell cycle, which tends to de-
crease the growth rate, whereas stratification increases the
effective space available for growth and thus tends to in-
crease the growth rate. In the case of the rule parameters
chosen, the differentiation effect dominates and cells grow
significantly more slowly in higher calcium levels (Fig.
3C), which agrees well with the in vitro observations.
The tendency of cells to grow in compact colonies, and
to stratify in physiologic calcium in vitro means that it is
extremely difficult to obtain a precise value for total cell
number, a parameter that it is trivial to extract from a
computational model simulation. Including results from
MTT, DNA, and BrdU assays (all indirect assessments of
cell number) gives the best estimate of the cell population
that is feasible for these cells in vitro. One indicator of the
accuracy of a computational model is its ability to replicate
the qualitative pattern of behavior seen in the in vitro ex-
periments. This was achieved for the calcium-dependent
growth, in that after approximately 2 weeks the model cell
population in the lowest calcium environment is roughly
2.5 times that in the highest calcium (a good agreement
with the MTT data). Similarly, an exact quantitative agree-
ment between modeled involucrin levels and those mea-
sured using immunofluorescence techniques in the in vitro
experiments cannot be expected, but the general pattern of
change shown in Fig. 7A (in vitro data) and C (modeled
data) is qualitatively similar. Hence we have demonstrated
that the rule changes summarized in Table 1 are sufficient
to represent keratinocyte behavior, in contrast to urothelial
cell behavior. The most important of these rules relate to
the ability of keratinocytes, but not urothelial cells, to
spontaneously stratify and differentiate in culture. How-
ever, these simple rules do not provide any information
about the mechanisms underlying these differences, which
will be related to the different genes and proteins expressed
in the 2 cell types, as well as a result of different culture
conditions routinely used in vitro. An experimental in-
vestigation with the aim of pin-pointing inherent versus
culture artifact induced differences between urothelial and
keratinocyte behavior is currently in the planning stages.
The tradeoff between the effects of differentiation and
stratification on proliferation leads to the question of
whether it is possible to have a situation in which the dif-
ferentiation pathway is somehow impaired, allowing cells
in higher calcium environments to grow more quickly than
in lower calcium. This was not observed for any of the
NHK models; for HaCat models, however, where the dif-
ferentiation rule was removed, the cell populations in 1mM
calcium and above expanded quickly, with those in the
highest calcium environments reaching up to 8 layers thick,
with no signs of leveling off at the end of the simulation
period. By contrast, at calcium levels below 1mM, little
stratification was evident, and population growth was ab-
rogated by contact inhibition.
In assembling the model rules for HaCat cells, we have
deliberately explored the assumption that this cell line has
lost the ability to differentiate and does not express in-
volucrin. However, Fig. 7B suggests that the involucrin
level in HaCat cell populations cultured in low calcium is
actually higher than levels in intermediate to high calcium.
At low calcium concentrations, however, we know that the
measurements are based on low cell numbers and thus are
unlikely to be as accurate as those based on much higher
cell numbers at higher calcium levels.
Inspection of MTT and DNA data (Figs. 3 and 4A and B)
shows that the growth of HaCat cells is compromised
compared to NHKs in vitro, particularly at low calcium le-
vels. By contrast, the model predicts that in low calcium the
growth of HaCat cells will be similar to that of NHKs,
whereas HaCat growth at higher calcium levels will actually
be enhanced by approximately a factor of 4. Hence, the
assumption we have tested in the model—that differentia-
tion is disabled in HaCat cells—is not sufficient to explain
the observed population growth characteristics. We there-
fore suggest that the current model is too simplistic to ac-
count for HaCat cells behavior. Some authors have found
deficiencies in differentiation in HaCats14 whereas others
have not.5 Interestingly, evidence exists that HaCat cells and
NHKs respond very differently to extracellular calcium. One
study investigated the effect of [Ca2þ]e on the induction of
p21 (a suppressor of cyclin-dependent kinases and hence an
inhibitor of cell proliferation) on NHKs and HaCat cells.16 It
was demonstrated that exposure of NHKs to physiologic
levels of calcium (1–1.5mM) induced an increase in in-
tracellular calcium levels, which activated S100C/A11 (a
process that initiates a cascade ultimately resulting in the
induction of p21), which slowed cell proliferation. In con-
trast HaCat cells required 5–10mM extracellular calcium
before a comparable increase in intracellular calcium with
induction of p21 occurred. This suggests that HaCat cells in
the process of spontaneous transformation have somehow
changed their response to extracellular calcium (mechanism
unknown) and hence become relatively insensitive to cal-
cium inhibition of proliferation. Our current model does not
include any differences in the cell’s ability to respond to
calcium. We also assumed that [Ca2þ]e-dependent inter-
cellular adhesion and hence contact inhibition of prolifera-
tion was similar for NHKs and HaCats, which may also
prove to be an oversimplification. Adaptation of our model
to include different parameters describing calcium-depen-
dent behavior for HaCat cells would thus be expected to
improve its accuracy.
Finally, it is hypothesized that epidermal tissue struc-
ture in vivo is regulated by the arrangement of NHKs into
8 WALKER ET AL.
epidermal proliferative units, where stem cells are located
in the center of proliferative cell clusters. Proceeding from
this, it is proposed that cells located on the cluster periphery
differentiate via the notch signaling pathway.17 It is also
possible that the heat shock transformation process dis-
rupted this cluster size control mechanism in HaCat cells,
resulting in the dysregulation of growth in physiologic
calcium levels. Our model does not currently include these
concepts, but our group is developing models of inter-
cellular signaling mechanisms. Linking the latter to the
current model, and eventually to models of key intracellular
signaling pathways, will allow us to explore these questions
in greater depth, and ultimately produce a powerful pre-
dictive tool for cell biologists and tissue engineers.
In summary, the current model, which describes the ef-
fect of extracellular calcium on NHK proliferation and dif-
ferentiation, indicates that a simple failure of HaCat cells to
differentiate at some stage is not sufficient to explain the
very different response of these cells to changing extra-
cellular calcium.
ACKNOWLEDGMENT
We gratefully acknowledge support from the Engineer-
ing Physical Sciences Research Council for Doctors Walker
and Sun for this project.
REFERENCES
1. Walker, D.C., Southgate, J.S., Hill, G., Holcombe, M., Hose,
D. R., Wood, S.M., Macneil, S., and Smallwood, R.H. The
epitheliome: modelling the social behaviour of cells. Bio-
systems 76, 89, 2004.
2. Walker, D.C., Hill, G., Wood, S.M., Smallwood, R.H., and
Southgate, J. Agent-based computational modeling of woun-
ded epithelial cell monolayers. IEEE. Trans. Nanobiosci. 3,
153, 2004.
3. Hennings, H., Michael, D., Cheng, C., Steinert, P., Holbrook,
K., and Yuspa, S.H. Calcium regulation of growth and dif-
ferentiation of mouse epidermal cells in culture. Cell 19, 245,
1980.
4. Bikle, D.D., Ng, D., Tu, C.L., Oda, Y., and Xie, Z. Calcium-
and vitamin D-regulated keratinocyte differentiation. Mol.
Cell. Endocrinol. 177, 161, 2001.
5. Boukamp, P., Petrussevska, R.T., Breitkreutz, D., Hornung,
J., Markham, A., and Fusenig, N.E. Normal keratinization in
a spontaneously immortalized aneuploid human keratinocyte
cell line. J. Cell Biol. 106, 761, 1988.
6. Sun, T., Mai, S., Norton, D., Haycock, J., Ryan, A., and
MacNeil, S. Self-organization of skin cells in three-
dimensional electrospun polystyrene scaffolds. Tissue Eng. 11,
1023, 2005.
7. Goberdhan, N.J., Dawson, R.A., Freedlander, E., and Mac-
Neil, S. A calmodulin-like protein as an extracellular mitogen
for the keratinocyte. Br. J. Dermatol. 129, 678, 1993.
8. Ralston, D.R., Layton, C., Dalley, A.J., Boyce, S.G., Freed-
lander, E., and MacNeil, S. The requirement for basement
membrane antigens in the production of human epidermal/
dermal composites in vitro. Br. J. Dermatol. 140, 605, 1999.
9. Marcus, J.S., Tulin, B., Cliff, S., Howell, J.A., Horrocks, M.,
and Chaudhuri, J.B. Human fibroblast culture on a crosslinked
dermal porcine collagen matrix. Biochem. Eng. J. 20, 217,
2004.
10. Higham, M.C., Dawson, R., Szabo, M., Short, R., Haddow,
D.B., and MacNeil, S. Development of a stable chemically
defined surface for the culture of human keratinocytes under
serum-free conditions for clinical use. Tissue Eng. 9, 919, 2003.
11. Little, M.C., Gawkrodger, D.J., and MacNeil, S. Differentia-
tion of human keratinocytes is associated with a progressive
loss of interferon gamma-induced intercellular adhesion
molecule-1 expression. Br. J. Dermatol. 135, 24, 1996.
12. Sun, T., Higham, M., Layton, C., Haycock, J., Short, R., and
MacNeil, S. Developments in xenobiotic-free culture of hu-
man keratinocytes for clinical use. Wound Repair Regen. 12,
626, 2004.
13. Baumgartner, W., Hinterdorfer, P., Ness, W., Raab, A.,
Vestweber, D., Schindler, H., and Drenckhahn, D. Cadherin
interaction probed by atomic force microscopy. Proc. Natl.
Acad. Sci. U. S. A. 97, 4005, 2000.
14. Houghton, P.J., Hylands, P.J., Mensah, A.Y., Hensel, A., and
Deters, A.M. In vitro tests and ethnopharmacological in-
vestigations: wound healing as an example. J. Ethnopharma-
col. 100, 100, 2005.
15. Watt, F.M. Influence of cell shape and adhesiveness on stra-
tification and terminal differentiation of human keratinocytes
in culture. J. Cell Sci. Suppl. 8, 313, 1987.
16. Sakaguchi, M., Miyazaki, M., Takaishi, M., Sakaguchi, Y.,
Makino, E., Kataoka, N., Yamada, H., Namba, M., and Huh,
N.H. S100C/A11 is a key mediator of Ca(2þ)-induced growth
inhibition of human epidermal keratinocytes. J. Cell Biol.
163, 825, 2003.
17. Savill, N.J., and Sherratt, J.A. Control of epidermal stem cell
clusters by Notch-mediated lateral induction. Develop. Biol.
258, 141, 2003.
Address reprint requests to:
Dawn Walker, Ph.D.
Department of Computer Science
Kroto Institute
University of Sheffield
South Yorkshire, United Kingdom
E-mail: [email protected]
EXOGENOUS CALCIUM ON KERATINOCYTE AND HaCat CELL PROLIFERATION 9