Corneal Fibroblasts Respond Rapidly to Changes in
Local Mechanical Stress
W. Matthew Petroll, Mridula Vishwanath, and Lisha Ma
PURPOSE. To investigate the response of corneal fibroblasts to
local changes in extracellular matrix (ECM) tension.
METHODS. Rabbit and human corneal fibroblasts were plated
inside fibrillar collagen matrices. After 18 to 72 hours, a glass
microneedle was inserted into the ECM and either pushed
toward a cell to reduce local tension, or pulled away to in-
crease tension. Time-lapse differential interference contrast
(DIC) imaging was performed both before and after needle
micromanipulation. ECM displacements were quantified, and
strain maps were generated by finite element modeling. In
some experiments, cells were treated with the Rho-kinase
inhibitor Y-27632 either 30 minutes before, or 1 hour after they
were pushed with the microneedle. Changes in focal adhesion
organization were also evaluated in a subset of cells expressing
green fluorescent protein (GFP)-zyxin, by simultaneous fluo-
rescent and DIC imaging.
RESULTS. Pulling on the ECM resulted in initial cell elongation,
followed by disengagement and retraction of pseudopodia. In
contrast, pushing the ECM toward a cell induced rapid short-
ening (contraction), presumably since existing cellular forces
were no longer counterbalanced by ECM tension. Pseudopo-
dial extension (spreading) was then observed at both ends of
the cell. The ECM was pulled inward during this secondary
spreading, and rapid turnover of focal adhesions was observed
along extending pseudopodia. Preincubation with Y-27632 or
cytochalasin D blocked both the initial contractile and second-
ary spreading responses.
CONCLUSIONS. Overall, the data suggest that corneal fibroblasts
actively respond to increases or decreases in local matrix stress
in an attempt to maintain tensional homeostasis (constant
tension), and that this response may be mediated by Rho
and/or Rac. (Invest Ophthalmol Vis Sci. 2004;45:3466–3474)
Mechanical loading has been shown to alter features such as
cell morphology6,7and sensitivity to growth factors.8–10In
most cells, the development of focal contacts and stress fibers
are tension-dependent processes,11–13and both in vitro and in
vivo studies have demonstrated that these structures tend to
align along the tensile axis under anisotropic conditions.14–17
Cell migration and spreading are also influenced by the me-
t is well established that mechanical stimuli play a key role in
regulating growth and function in a variety of cell types.1–5
chanical stiffness of the substrate. In general, cells on flexible
two-dimensional (2-D) substrates are more migratory and have
smaller focal adhesions than those on more rigid substrates.18
Cells also preferentially spread on more rigid substrates.19
Studies of wound healing in vivo and related experimental
models in vitro have also shown that cellular force generation
is regulated, in part, by the mechanical properties of the
extracellular matrix (ECM).6,9,14,20–23
The response of cells to dynamic changes in their mechan-
ical environment has also been investigated. For example,
Eastwood et al.20used a culture force monitor to measure how
dermal fibroblasts within three-dimensional (3-D) collagen ma-
trices respond to changes in tensional loading.24Cells were
first allowed to develop baseline tension in the 3-D matrix.
Subsequent stretching of the entire gel resulted in an initial
increase in the measured force. However, this was immediately
followed by a gradual cell-dependent reduction in force toward
the baseline level. Similarly, compressing the gel caused an
initial loss of tension that was followed by a cell-dependent
increase in force back to the baseline level. Based on these data
the authors conclude that cells within 3-D matrices respond to
changes in mechanical loading in a way that maintains “ten-
sional homeostasis” (constant tension) in their surrounding
matrix. This model is also supported by studies demonstrating
that cellular forces in 3-D matrices reach a constant value that
is independent of matrix stiffness.25Tensional homeostasis
may be fundamental to the regulation of tissue tension under
normal conditions, during development and also in response to
injury. In the cornea, large shifts in the distribution of ECM
tension can be induced by lacerating injury, penetrating kera-
toplasty, or refractive surgery. The response of corneal fibro-
blasts to changes in ECM stress may therefore play an impor-
tant role in both the acute and long-term clinical outcomes
after such insults.
Although the elegant studies by Eastwood et al.20have
provided important insights into the overall response of an
aggregate of dermal fibroblasts to tensional loading, the dy-
namic response of isolated cells to changes in local stress have
yet to be studied in a 3-D system. Thus, the specific changes in
mechanical activity (e.g., spreading, contraction, and migra-
tion) that underlie the changes in overall ECM stress are not
known. Furthermore, the response of corneal fibroblasts to
mechanical signals has received little attention. We recently
developed a new experimental model to assess directly the
cell–matrix mechanical interactions at the cellular level by
plating corneal fibroblasts at very low density inside 3-D fibril-
lar collagen matrices, and performing high-magnification time-
lapse differential interference contrast (DIC) and fluorescence
imaging.26–28With this approach, pseudopodial extensions
and retractions can be directly correlated with local collagen
matrix deformation in a 3-D environment. We have used this
model to study the behavior of both rabbit and human corneal
fibroblasts after serum removal, disruption of the F-actin cy-
toskeleton, and inhibition of Rho-kinase.26–28However, the
response of cells to mechanical stimuli has not yet been
evaluated. In this study, we investigated the dynamic response
of corneal fibroblasts to tensional loading, by performing time-
lapse imaging while locally deforming the collagen matrix
adjacent to isolated cells. We demonstrate for the first time that
From the Department of Ophthalmology, University of Texas
Southwestern Medical Center, Dallas, Texas.
Supported by Grant EY13322 from the National Institutes of
Health and funds from Research to Prevent Blindness, Inc.
Submitted for publication April 1, 2004; revised May 4, 2004;
accepted May 13, 2004.
Disclosure: W.M. Petroll, None; M. Vishwanath, None; L. Ma,
The publication costs of this article were defrayed in part by page
charge payment. This article must therefore be marked “advertise-
ment” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Corresponding author: W. Matthew Petroll, Department of Oph-
thalmology, University of Texas Southwestern Medical Center, 5323
Harry Hines Boulevard, Dallas, TX 75390-9057;
Investigative Ophthalmology & Visual Science, October 2004, Vol. 45, No. 10
Copyright © Association for Research in Vision and Ophthalmology
cells inside fibrillar collagen matrices undergo rapid changes in
cell mechanical activity that are consistent with the tensional
homeostasis model. The data also suggest that these responses
may be mediated, in part, by the small GTPases Rho and/or
Studies were performed using both primary rabbit corneal fibroblasts
(NRK) and a previously characterized telomerase-infected, extended
lifespan human corneal fibroblast cell line, HTK.29NRK cells were
harvested from New Zealand White Albino rabbit eyes (Pel-Freez,
Rogers, AR) as previously described.26Both cell types were cultured in
25-cm2tissue culture flasks (Costar, Cambridge, MA) using complete
medium consisting of modified Eagle’s minimum essential media
(MEM; Sigma-Aldrich, St. Louis, MO) supplemented with 1% Penicillin,
1% Streptomycin and 1% amphotericin B (Fungizone; BioWhittaker,
Inc., Walkersville, MD) and 10% fetal bovine serum (FBS, Sigma-
For some experiments, NRK cells were transfected to express
green fluorescent protein (GFP)-zyxin, using human zyxin in a
pEGFP-N1 vector (BD-Clontech Laboratories, Inc., Palo Alto, CA). This
probe has been used previously as a marker for focal contacts in
goldfish fin fibroblasts,30–32mouse melanoma cells,33and corneal
fibroblasts.26–28Transfection was then performed (Lipofectamine
PLUS; Invitrogen, Carlsbad, CA) as previously described.26
Hydrated collagen matrices were prepared by mixing neutralized bo-
vine dermal collagen (Vitrogen 100; Collagen Corp., Palo Alto, CA)
with 10? MEM to achieve a final collagen concentration of 2.48
mg/mL.26For plating cells inside the matrix, a 50-?L suspension of
NRK or HTK cells was mixed with 500 ?L of collagen solution. The
cell/collagen mixture was preincubated at 37°C for 5 minutes, and
30-?L aliquots (containing approximately 500 cells) were then poured
onto culture dishes (Delta T; Bioptechs, Inc., Butler, PA). Each aliquot
was spread over a central 12-mm diameter circular region on the dish
and was approximately 100 ?m thick. The dish was then placed in a
humidified incubator (37°C, 5% CO2) for 60 minutes for polymeriza-
tion and overlaid with 2 mL of complete medium.
Time-Lapse Digital Imaging
Microscopy was performed as previously described.26,27Briefly, we
used an inverted microscope with fluorescence and DIC imaging mod-
ules (TE300; Nikon, Tokyo, Japan) two high-speed filter wheels for
rapid selection of excitation and emission filters and shuttering of
epifluorescent illumination, and a high-resolution cooled CCD camera
(CoolSnap HQ; Roper Scientific, Tuscan, AZ). The hardware was con-
trolled by computer running image-analysis software (MetaVue; Uni-
versal Imaging Corp., Downingtown, PA). To maintain cell viability
during imaging, a microincubation system and objective heater was
used (Bioptechs, Inc.). A microperfusion pump (Bioptechs, Inc.) was
used to perfuse the cells continuously while on the microscope stage
with complete medium containing HEPES buffer at the rate of 6 mL/h.
Dishes were moved to the microscope stage 18 to 72 hours after
seeding on the gel, allowing them to develop a more consistent bipolar
spindle-shaped morphology, as is observed during in vivo wound
healing.16,34–36In each experiment, cells were allowed to acclimate to
the microincubation system for 30 to 60 minutes before time-lapse
imaging. The cell density was sparse enough to focus on the mechan-
ical activity of a single cell, minimizing the potential interference
caused by neighboring cells. The activity of a single cell was imaged for
up to 5 hours using either a 40? or 60? oil-immersion objective.
Nomarski DIC images and/or enhanced GFP (EGFP) images were
automatically acquired at 1- to 3-minute intervals, using the imaging
software (MetaVue; Universal Imaging Corp.). In most experiments,
3-D data sets were obtained at each time point by repeating the
acquisition at four to five sequential focal planes in z-steps of 2 to 3
?m. To minimize phototoxicity, neutral-density filters and 2 ? 2
on-chip camera binning was used.
Micromanipulation of ECM
In all experiments, glass microneedles with flame-polished tips at-
tached to a micromanipulator (Narishige Scientific Laboratory, Tokyo,
Japan) were used for local deformation of the ECM. Microneedles with
?35-?m diameter tips were made using a pipette puller. These needles
were strong enough to withstand the forces needed to stretch or
compress the collagen matrix. The needle was positioned at a 45°
angle to the microscope stage. After 30 minutes of time-lapse imaging
of a cell of interest, the needle was inserted axially into the collagen
lattice 50 to 100 ?m from the leading edge of the cell. After the needle
was inserted, time-lapse imaging was performed for an additional 30 to
60 minutes. The needle was then either pushed toward the cell (25–50
?m) to compress the collagen ECM, thereby decreasing the effective
matrix stiffness, or pulled away from the cell (25–50 ?m) to stretch the
ECM, thereby increasing the matrix stiffness (Table 1). All manipula-
tions were visualized using DIC imaging. We generally selected cells
that were less than 25 ?m from the top of the collagen matrix, to
minimize free body motion of the gel. The micropipette was inserted
only far enough to bring the tip into the same focal plane as the cell.
After the ECM was pushed or pulled, time-lapse imaging was continued
for an additional 2 to 4 hours.
In additional experiments, cells were treated with the Rho-kinase
inhibitor Y-27632 either 30 minutes before (10 cells) or 1 hour after
the needle push (5 cells). Four microliters of a 5-mM stock solution of
Y-27632 was added to the culture dish to achieve a final concentration
of 10 ?M, and the perfusion medium was simultaneously switched to
complete medium containing 10 ?M Y-27632. In other experiments,
cytochalasin-D (Sigma-Aldrich) was added to the culture dish (final
concentration, 25 ?M) before ECM micromanipulation, to assess the
effect of F-actin on the cellular response. Control experiments were
also performed on collagen lattices without cells to determine the
effect of needle pushing or pulling alone (i.e., without cellular force
generation) on the pattern of ECM deformation.
TABLE 1. Summary of Experiments
Needle Maneuver and Reagents
Experiments at Each
Time Point (n)
1 Day 2 Days3 DaysNRKHTK
Y-27632 3 needle push
Cytochalasin-D 3 needle push
NRK, rabbit corneal Keratocytes; HTK, human corneal Fibroblasts.
IOVS, October 2004, Vol. 45, No. 10
Fibroblasts Respond to Changes in ECM Tension3467
Image Processing and Analysis
Image processing was performed using MetaMorph (Universal Imaging
Corp.). ECM deformation was quantified by measuring the x, y coor-
dinates of landmarks in DIC images using the “measure pixel” feature
in the software. To display the ECM displacements, a custom-written
program (Visual Basic; Microsoft, Redmond, WA) program was used.
The program generated cross-marks (?) and tracks corresponding
with the start points and displacements, respectively, of ECM land-
marks from the measured x, y coordinates. Matrix deformation before,
during, and after micromanipulation was quantified. A similar ap-
proach was used to track adhesion movement in cells expressing
Finite Element Modeling
Finite element modeling (FEM) was used to visualize and quantify the
pattern of matrix deformation due to needle pushing, both with and
without cells. Finite element models were created using engineering
analysis software (Ansys, ver. 7.0; Ansys Inc., Canonsburg, PA), as
previously described by us.28,37Briefly, nodes were defined at coordi-
nates coinciding with ECM landmarks from the DIC images before ECM
micromanipulation. Boundary nodes were placed at the periphery of a
600-?m diameter circular field around this central set of nodes. A 2-D
plane stress model was created from the nodes with linear elastic
triangular elements. For simplicity, the matrix was assumed to be
isotropic, with a Young’s modulus of 3.89 ? 10?10N/?m2, an effective
thickness of 15 ?m, and a Poisson’s ratio of 0.3.28,38To generate maps
of ECM deformation, the displacements measured from time-lapse
recordings were applied to the corresponding nodes in the model. The
resultant strains induced on the matrix were calculated and displayed.
ECM Stress Induced by Needle Push
To investigate the ECM stress induced by needle micromanip-
ulation, we first tracked the ECM deformation induced by
pushing with microneedles using matrices without cells (Fig.
1A, red tracks, cross marks starting position). The FEM strain
maps parallel to the direction of movement (x-axis) confirm
that there was significant compression in front of needles (Figs.
1B, 1C, blue areas). Principal strain vectors reveal that although
some tension is produced perpendicular to this axis (Fig. 1D,
small white vectors), compression is clearly the dominant ef-
fect (large blue vectors) in front of the needle. Note that
maximum compression of the ECM was observed just in front
of the needle (see Movie 1 at www.iovs.org/cgi/content/full/
45/10/3466/DC1). We also found that tension was generated
by pulling with microneedles (not shown).
Fibroblast Response to ECM Compression
By 1 day after plating inside 3-D collagen matrices, both NRK
and HTK cells generally had a bipolar morphology with thin
pseudopodial processes, consistent with previous observa-
tions.26,27Cells were always aligned nearly parallel to the dish
on which the collagen matrix was plated. DIC imaging allowed
detailed visualization of the cells and the individual collagen
fibrils surrounding them. Fibroblasts repeatedly extended and
retracted pseudopodia at all time points after plating within
3-D matrices.26One to 2 days after the cells were plated,
pseudopodial extension generally occurred at the front of the
cells, whereas the rear was much less active and underwent
intermittent retractions, resulting in cell migration. The initial
axial insertion of the microneedle into the ECM did not alter
the normal pattern of cell behavior (Movie 2 at www.iovs.org/
cgi/content/full/45/10/3466/DC1). By 3 days after plating in-
side collagen matrices, fibroblasts continued to extend and
retract pseudopodia, but generally did not undergo significant
migration, consistent with previous observations.28
DIC image of a collagen matrix without cells, being compressed by
pushing with a glass microneedle from right to left (ECM displace-
ments are indicated by red tracks; ?, starting position before pushing
the needle). (B) Map of strain along the x-axis generated from the ECM
displacements using FEM (element solution). Pushing on the ECM with
a needle caused matrix compression (blue and green) in front of and
tension (red and orange) behind the region of contact of the needle.
Scale bars are in dimensionless units ?L/L (change in length/initial
length). Negative values: compression; positive values: tension. Black
bar: front of needle. (C) Contour map of strain along the x-axis
generated using the nodal solution. (D) Principal strain vectors: blue,
compression; white: tension.
The effects of needle micromanipulation on the ECM. (A)
3468Petroll et al.
IOVS, October 2004, Vol. 45, No. 10
Pushing the ECM toward a cell with a microneedle induced
rapid cellular shortening (contraction) with corresponding
ECM compression along the cell body (Fig. 2B, 2F, arrows;
compare with Figs. 2A, 2E). After this initial contraction, rapid
cell spreading was observed, and tractional force was gener-
ated as indicated by pulling in of the ECM (Figs. 2C, 2G, black
tracks). This secondary spreading response was generally ob-
served at both ends of cells (Fig. 2C) and is best appreciated in
time-lapse movies (Movies 3 and 4 at www.iovs.org/cgi/
content/full/45/10/3466/DC1). Note that during the secondary
spreading response, large amounts of ECM displacement were
produced for relatively small amounts of pseudopodial exten-
sion. “Bundling up” of collagen at the base of the pseudopodia
was also often observed as new pseudopodial extensions
pulled the collagen fibrils inward. The addition of Y-27632 1
hour after the needle push resulted in dramatic cell elongation
and relaxation of tension on the matrix (Fig. 2D; Movie 5 at
an active response to needle pushing that is, in part, Rho-kinase
We also followed the dynamic changes in focal adhesion
organization in response to ECM micromanipulation, by trans-
fecting cells to express GFP-zyxin. GFP-zyxin was organized
into focal adhesions that were most easily visualized along
pseudopodial processes (Figs. 3A, 3B, arrows), consistent with
previous observations.26,27During cellular contraction after a
35-?m needle push, ECM deformation (Fig. 3B, blue tracks)
correlated with the inward movement of existing focal adhe-
sions toward the cell body (Fig. 3B, white tracks). In contrast,
the secondary spreading response was associated with the
rapid turnover and formation of focal adhesions along extend-
ing pseudopodia (Fig. 3C, arrows), which resulted in additional
pulling in of the ECM (Fig. 3C, blue tracks; Movie 6 at www.
Although there were differences in the magnitude of the
initial cell contraction and secondary spreading, the same over-
all pattern of cell mechanical activity was observed in response
to needle pushing in 20 of 21 cells. There was only one
exception in which a cell moved away from the needle after
the needle push. To quantify the amount of initial cellular
contraction in response to ECM micromanipulation, we mea-
sured the distance between ECM landmarks at the ends of the
cell and calculated the distance between them. Analysis was
performed only on experiments in which both ends of the cells
and the surrounding matrix were clearly visible (total 14 cells).
All cells analyzed showed some amount of contraction imme-
diately after needle push, with an average shortening of 26.9%
? 6.8% (range, 14%–37%; Table 2). Note that some amount of
ECM compression (representative of
20 of 21 experiments). (A–D) Rabbit
corneal fibroblast 2 days after plating
inside collagen matrix. (A) Before
needle push. (B) Pushing the ECM 45
?m toward the cell induced rapid
cellular contraction (36% shortening)
and ECM compression along the cell
body (arrows). (C) This initial con-
traction was followed by the exten-
sion of pseudopodia (arrows) and
rapid pulling in of the ECM (traction)
at both ends of the cell (black tracks,
?: start position beginning 1 minute
after needle push). (D) Subsequent
addition of Y-27632 induced cell
elongation and dramatic relaxation of
cell-induced matrix tension (black
tracks, ?: start position after adding
Y-27632). This response to Y-27632
was observed in all five cells evalu-
ated. (E–G) Rabbit corneal fibroblast
1 day after plating inside collagen
matrix. (E) Before needle push. (F)
Cellular contraction (arrows) was
observed after a 25-?m needle push.
(G) This initial contraction was fol-
lowed by cell spreading (arrows)
and traction (black tracks, ?: start
position beginning 1 minute after
IOVS, October 2004, Vol. 45, No. 10
Fibroblasts Respond to Changes in ECM Tension3469
passive cellular shortening would be expected due to the
effect of the needle push alone. We estimated this shortening
to be less than 5%, using data obtained from needle pushing
without cells (Fig. 1).
Effects of Y-27632 and Cytochalasin D on the
Cellular Response to ECM Compression
Addition of the Rho-kinase inhibitor Y-27632 to serum-contain-
ing (S?) medium prior to pushing with the needle induced
dramatic cell elongation and relaxation of tension on the ma-
trix within 5 minutes (Fig. 4B, white arrows and black tracks;
compare with 4A).28The active cellular contraction normally
observed after needle push was essentially blocked by
Y-27632, as indicated by minimal ECM displacements (Fig. 4C,
black tracks). For the cell shown in Fig 4C, shortening of only
5.9% was observed; this is much less than the minimum short-
ening measured with complete medium (14%; Table 2). The
secondary spreading response was also blocked by pretreat-
ment with Y-27632. As shown is Figure 4D, there was no cell
spreading or matrix deformation observed, even 1 hour after
pushing with the needle (black tracks; see also Movie 7 at
Similarly, in cells pretreated with cytochalasin D, the nor-
mally observed responses to ECM compression were absent.
Incubation with cytochalasin D resulted in cell elongation and
relaxation of cell-induced ECM stress (Fig. 4F, white arrows
and black tracks; compare with 4E). Cellular shortening of only
8.5% was observed after pushing with the needle (Fig. 4G). In
addition, no secondary cell activity or ECM displacements were
observed after the initial needle push (Fig. 4H). Note that in
cells pretreated with either Y-27632 or cytochalasin D, reper-
fusion with complete medium after the needle push resulted in
cellular contraction and tractional force generation (not
shown); thus, the cells were still viable.
We also mapped the pattern of ECM deformation after the
needle push using FEM (Fig. 5), for both a large needle push
(40 ?m; Fig. 5A, 5B), and a smaller needle push (30 ?m; Figs.
5C, 5D). The strain maps verify that there was compression
along the cell body immediately after pushing the needle in S?
medium in both cases (Figs. 5A, 5C, blue regions). Cellular
shortening of 36% and 24% was measured for the cells in
Figures 5A and 5C, respectively. Pulling in of the matrix during
the secondary spreading response was also demonstrated in
maps showing the strain produced beginning after the initial
contraction (Figs. 5B, 5D). This effect was indicated by stretch-
ing of the matrix at the ends of cells and additional compres-
sion along the cell body. Note that during secondary spreading,
maximum compression was usually generated near the base of
extending pseudopodia (Figs. 5B, 5D, dark blue areas), consis-
tent with the pattern of ECM deformation observed in actively
spreading cells both on planar elastic substrates and inside
FEM modeling also confirmed that there was inhibition of
both the initial and secondary responses after preincubation
with Y-27632 (Figs. 5E, 5F; 30-?m needle push). The compres-
sion produced by the needle push is located primarily near the
needle tip (Fig. 5E, blue region), not along the cell body as
normally observed (compare with Figs. 5A, 5C). Note that a
similar pattern of compression was observed in matrices with-
out cells (compare with Fig. 1B). Furthermore, little change in
ECM stress was observed after the initial push (Fig. 5F, yellow
indicates strains close to zero).
Fibroblast Response to ECM Stretch
Inserting a microneedle in front of a cell and pulling it away
stretched the ECM (Fig. 6B, red tracks) which resulted in cell
elongation (compare small arrows in Figs. 6A, 6B). Labeling
with GFP-zyxin demonstrated that focal adhesions (Figs. 6E, 6F,
arrows) initially remained attached to the ECM, because ECM
displacement and focal adhesion displacements correlated
highly (Fig. 6F, compare blue and white tracks). Later, pseu-
dopodia often disengaged and retracted (Figs. 6C, 6G). Adhe-
sions were occasionally left behind (Figs. 6F, 6G, arrowheads),
suggesting that they were “torn off” as the cell retracted. After
retraction, repeated extension and retraction of pseudopodia
were observed as cells partially respread (Movies 8 and 9 at
www.iovs.org/cgi/content/full/45/10/3466/DC1). New focal
adhesions were formed during respreading (Fig. 6H, arrows),
but very little ECM displacement was observed (Figs. 6D, 6H,
red and blue tracks). These results were observed in 10 of the
11 cells studied. In the other experiment, when the ECM was
onstrating the response to ECM compression. (A) Rabbit corneal fibro-
blast 2 days after plating inside collagen matrix. GFP-zyxin was orga-
nized into focal adhesions that were most easily visualized along
pseudopodial processes (arrows). (B) During cellular contraction after
a 35-?m needle push, ECM deformation (blue tracks) correlated with
the inward movement of existing focal adhesions (arrows) toward the
cell body (white tracks). (C) The secondary spreading response was
associated with the formation of new focal adhesions at pseudopodial
tips (arrows), and rearward movement of existing adhesions, which
resulted in additional pulling in of the ECM (blue tracks).
Overlays of GFP-zyxin (green) and DIC (red) images dem-
TABLE 2. Cellular Contraction after Needle Push
1425.3 ? 6.4 13.5–34.826.9 ? 6.814–37
Shortening is expressed as the mean ? SD.
3470 Petroll et al.
IOVS, October 2004, Vol. 45, No. 10
stretched by pulling near the trailing end of a migrating cell,
the cell migrated away from the needle.
It should be noted that for all experimental manipulations,
similar responses were observed for NRK and HTK cells. Fur-
thermore, there were no apparent differences in the response
to micromanipulation between the three time points evalu-
In this study, we investigated for the first time the response of
isolated cells to local mechanical stimulation using a 3-D col-
lagen matrix model. Changing local ECM stress using mi-
croneedles induced rapid and reproducible response patterns
in both rabbit and human corneal fibroblasts. Pushing the ECM
reduced local tension in the matrix and resulted in rapid
cellular contraction with corresponding ECM compression.
This initial contraction is probably due to the release of pre-
existing cellular forces, similar to the shortening produced by
releasing one end of a rubber band that is under tension. After
this initial contraction, respreading was observed at both ends
of cells, and tractional force was generated as indicated by
pulling in of the ECM. During this secondary spreading re-
sponse, large amounts of ECM displacement were produced
for relatively small amounts of pseudopodial extension, which
would be expected for a low-compliance substrate (similar to
a tire spinning in the mud). Pulling on the ECM resulted in
initial cell elongation, followed by disengagement and retrac-
tion of pseudopodia. Raucher and Sheetz40have demonstrated
that the addition of amphiphilic compounds or fluorescent
lipids that expand the plasma membrane and decrease mem-
brane tension (as measured using laser tweezers) stimulates
spreading of NIH 3T3 cells on rigid substrates. Conversely,
increasing membrane tension by osmotically swelling cells
reduces the lamellipodial extension rate. Our results are con-
sistent with these findings, in that ECM compression (which
reduces membrane tension) induced cell spreading and ECM
stretch (which increases membrane tension) caused pseudopo-
Many groups have described the existence of a steady level
of cell-mediated tension in large aggregates of cells.6,14,20,23,25
Brown et al.24investigated the responses of dermal fibroblasts
to changes in mechanical loading, using a culture force moni-
tor to measure the cell-mediated changes in mechanical ten-
D blocked the cellular response to
ECM micromanipulation (representa-
tive of 10 Y-27632 experiments and 4
cytochalasin D experiments). (A–D)
Rabbit corneal fibroblast 1 day after
plating inside collagen matrix. (B)
Cell elongation (arrows) and ECM
relaxation was observed after adding
Y-27632 (black tracks, ?: position
just before addition of Y-27632). (C)
Subsequent pushing with a needle
(note shadow of needle on right) in-
duced little cell contraction (5.9%
shortening, black tracks). (D) There
was also no secondary spreading or
traction after needle push, as indi-
cated by minimal ECM displacement
(black tracks, ?: starting position 1
minute after push). (E–H) Human
corneal fibroblast 2 days after plating
inside collagen matrix. (F) After cy-
tochalasin D was added, elongation
of the cell (arrows) and relaxation of
cell-induced matrix stress were ob-
served (black tracks). (G) Pushing on
the ECM in front of the cell caused
ECM compression in front of the nee-
dle, but little cell contraction (4.7%
shortening, black tracks). (H) There
was also no secondary spreading or
traction after needle push, as indi-
cated by minimal ECM displacements
(black tracks, ?: starting position 1
minute after push).
Y-27632 and cytochalasin
IOVS, October 2004, Vol. 45, No. 10
Fibroblasts Respond to Changes in ECM Tension 3471
sion across entire 3-D matrices. It was demonstrated that fibro-
blasts respond to changes in mechanical loading in a way that
maintains “tensional homeostasis” in the surrounding matrix.
The response of isolated fibroblasts we observed in this study
is also consistent with the tensional homeostatic model. After
the release of tension due to needle push, cells actively pull in
the ECM during respreading to partially decompress the ma-
trix. Similarly, after the initial increase in tension due to pulling
with the microneedle, the stretched cells attempt to release
tension on the matrix by disengaging their pseudopodia. Thus,
in both cases, corneal fibroblasts respond in a way that tends to
counteract the initial change in stress. In previous studies,
spontaneous rupture of adhesions and retraction of cells have
been shown to reduce ECM tension.26,27However, we were
unable to detect such ECM relaxation in this study. This is most
likely because the large load being borne by the ECM after
needle pulling masked the smaller forces generated by the
cells. Future studies using smaller needle displacements during
pulling, or more flexible (thinner) needles could be used to
overcome this limitation.
For FEM analysis of fibroblast-induced matrix distortion,
linear elastic, isotropic material properties were used. Our
previous matrix calibration experiments suggest that this is a
reasonable assumption when studying the normal cellular pat-
tern of force generation,41but it should be noted that individ-
ual collagen fibrils with which a cell interacts can undergo
much larger displacements than neighboring fibrils, particu-
larly at the leading edge.26Furthermore, large pushes or pulls
with the microneedle may induce stresses that are outside the
linear response range for the collagen matrix. For this reason,
FEM was used to generate strain maps (to visualize the pattern
of ECM deformation), but was not used to estimate forces.
The cellular response to transient mechanical stimulation
with microneedles has been investigated by Lo et al.19and
Wang et al.,42who used planar elastic substrates. In their
model, pulling near the trailing end of a migrating 3T3 cell
caused the cell to reverse direction and move toward the
microneedle. In contrast, pushing the substrate toward the
leading edge of a cell caused the cell to retract its leading edge
and migrate away from the needle. We did not observe this
type of migratory behavior in response to mechanical stimula-
tion in this study. However, significant differences exist be-
tween these two experimental models. First, most of the cells
we studied were not undergoing rapid migration before needle
manipulation. Second, the mechanical stiffness of collagen
matrices is generally much less than that of polyacrylamide
substrates. Finally, cell mechanical interactions on a planar
substrate coated with nonfibrillar collagen are likely different
from that which occurs within a 3-D fibrillar ECM.
Previous studies have established that the Rho-family of
small guanosine triphosphatases (GTPases) such as Rho, Rac,
and Cdc42 play a central role in regulating the cytoskeletal
changes associated with cell mechanical activity. These GTP-
binding proteins function as molecular switches, alternating
between the active GTP-bound state and the inactive GDP-
bound state. Activated Rho stimulates the formation of stress
fibers, the development of large focal adhesions (focal con-
tacts), and cellular contraction (shortening).43–46In contrast,
activated Rac induces the creation of smaller focal complexes,
actin polymerization, and cell spreading.43–48Recent studies
the initial (A, C, E) and secondary (B,
D, F) responses to needle push in S?
medium (A–D) and in medium con-
taining Y-27632 (E, F). Scale bars are
in dimensionless units ?L/L (change
in length/initial length). Negative val-
ues: compression; positive values:
tension. Black bars: the front of nee-
dle. (A, B) Same cell as in Figures
2A–C. (A) After a 40-?m needle
push, compression was observed
along the cell body (blue regions).
(B) Additional pulling in of the ma-
trix (traction) was observed over the
next 30 minutes. This is indicated by
decompression of the matrix at the
ends of cells (red, yellow, and green
regions) and additional compression
along the cell body (blue regions).
Only the strain produced after the
initial response is shown (starting 1
minute after the needle push). (C, D)
Human corneal fibroblast 2 days after
plating inside collagen matrix. (C) Af-
ter a 30-?m needle push, compres-
sion was observed along the cell
body (blue and green regions). (D)
Additional pulling in of the matrix
was observed over the next 50 min-
utes. This is indicated by stretching
of the matrix at the ends of cells (red
and orange regions) and additional
compression along the cell body
(blue and green regions). Only the
strain produced after the initial re-
sponse is shown (starting 4 minutes
after the needle push). (E, F) Same
FEM strain maps of both
cell as in Figures 4A–D. Both the initial and secondary responses to a 30-?m needle push were inhibited by preincubation with Y-27632. The
compression produced by the needle push is located primarily near the needle tip (E; blue region); a similar pattern of compression was observed
in matrices without cells (compare with Fig. 1B). Cellular shortening of only 5.9% was detected. Very little ECM stress was generated after the initial
push (F; yellow indicates strains close to zero).
3472Petroll et al.
IOVS, October 2004, Vol. 45, No. 10
suggest that Rho and Rac may be involved in the cellular
response to a variety of mechanical signals. For example, cyclic
deformation strain of cultured airway smooth muscle cells
induces increased stress fiber and focal adhesion formation and
contractility coincident with a fourfold increase in Rho acti-
vity.49In addition, reorientation of vascular endothelial cells in
response to shear stress requires Rho-induced depolarization
(cell rounding), followed by Rho/Rac mediated migration in
the direction of flow.50In vascular smooth muscle cells, non-
cyclic uniaxial mechanical stretching was shown to downregu-
late Rac and suppress cell spreading, whereas decreasing me-
chanical tension (by inhibiting Rho-kinase or myosin light
chain kinase) increased cell spreading through upregulation of
Rac. A similar spreading response was observed by inhibiting
Rho-kinase in the current study. Increased spreading was also
observed after reducing ECM tension by pushing with mi-
croneedles. Rac activation can be stimulated by platelet-de-
rived growth factor (PDGF),43,51and PDGF induces both cell
spreading and tractional force generation by corneal fibroblasts
(manuscript in preparation), similar to the secondary response
observed after needle pushing in the present study.
In addition to inducing cell elongation and relaxation, inhi-
bition of Rho-kinase also blocked both the initial and secondary
responses to ECM compression with a microneedle. This sup-
ports our contention that the initial contraction results from
preexisting cellular forces. Without these preexisting forces,
the change in cellular tension induced by a needle push was
apparently too small to induce the secondary spreading re-
sponse. Taken together, the data suggest that Rho and Rac
activation may play a central role in the fibroblast response to
local mechanical stimulation. Additional studies more specifi-
cally targeting Rho and Rac signaling pathways are needed to
clarify the molecular mechanisms underlying these important
The authors thank Ju ¨rgen Wehland and coworkers (BGF, Braun-
schweig, Germany) for providing the EGFP-zyxin expression vector,
James Jester for providing the HTK cells, and Charles Chuong (Univer-
sity of Texas at Arlington) for advice regarding finite element modeling.
1. Shyy JY-J, Chien S. Role of integrins in endothelial mechanosensing
of shear stress. Circ Res. 2002;91:769–775.
2. Sadoshima J, Izumo S. The cellular and molecular response of
cardiac myocytes to mechanical stress. Annu Rev Physiol. 1997;
3. Liu M, Tanswell AK, Post M. Mechanical force-induced signal
transduction in lung cells. Am J Physiol. 1999;277:L667–L683.
4. Brown TD. Techniques for mechanical stimulation of cells in vitro:
a review. J Biomech. 2000;33:3–14.
5. Tummina SJ, Mitton KP, Arora J, Zelenka P, Epstein DL, Russell P.
Mechanical stretch alters the actin cytoskeletal network and signal
transduction in human trabecular meshwork cells. Invest Ophthal-
mol Vis Sci. 1998;39:1361–1371.
6. Eastwood M, Porter RA, Kahn U, McGrouther DA, Brown RA.
Quantitative analysis of collagen gel contractile forces generated
by dermal fibroblasts and the relationship to cell morphology.
J Cell Physiol. 1996;166:33–42.
7. Tomasek JJ, Hay ED. Analysis of the role of microfilaments and
microtubules in acquisition of bipolarity and elongation of fibro-
blasts in hydrated collagen gels. J Cell Biol. 1984;99:536–549.
8. Ingber DE, Folkman J. Mechanochemical switching between
growth and differentiation during fibroblast growth factor stimu-
ECM stretching produced by pulling
the ECM away from cells with a glass
microneedle (representative of 10/11
experiments). (A–D) Rabbit corneal
fibroblast, 2 days after plating inside
collagen matrix. Needle pull is 50
?m to the right. (A) Before needle
pull. (B) On pulling the ECM in front
of the cell with a microneedle, the
cells and ECM were stretched, as vi-
sualized by matrix displacements
(red tracks). (C, D) Later, pseudopo-
dia disengaged and retracted (com-
pare arrows in C and D) and then
extended and retracted processes
without inducing significant ECM de-
formation (red tracks, ?: position 1
minute after needle pull). (E–H) GFP-
zyxin (green) and DIC (red) overlay
of rabbit corneal fibroblast, 1 day af-
ter plating inside collagen matrix.
Needle pull is 30 ?m to the right. (E)
Before needle pull. (F) On pulling
the ECM in front of the cell with a
microneedle, the cells and ECM were
stretched, as visualized by matrix dis-
placements (blue tracks). Labeling
with GFP-zyxin demonstrated that fo-
cal adhesions (E, F, arrows) initially
remained attached to the ECM, since
ECM displacement and focal adhe-
sion displacements correlated highly
(compare blue and white tracks). (G,
H) After the initial stretch, the pseu-
dopodia disengaged and retracted.
After retraction, repeated extensions
and retractions of pseudopodia were observed as cells partially respread. New focal adhesions were formed as the cells respread (H, arrows), but
no significant ECM displacement was observed (G, H, blue tracks, ?: position 4 minutes after needle pull). Adhesions were occasionally left behind
(F, G, arrowheads), suggesting that they were “torn off” as the cell retracted.
IOVS, October 2004, Vol. 45, No. 10
Fibroblasts Respond to Changes in ECM Tension 3473
lated angiogenesis in vitro: role of extracellular matrix. J Cell Biol.
9. He Y, Grinnell F. Stress relaxation of fibroblasts activates a cyclic
AMP signaling pathway. J Cell Biol. 1994;126:457–464.
10. van Bockxmeer FM, Martin CE, Constable IJ. Effect of cyclic AMP
on the cellular activity and DNA synthesis of chorioretinal fibro-
blasts maintained in collagen matrices. Exp Cell Res. 1984;155:
11. Riveline D, Zamir E, Balaban NQ, et al. Focal contacts as
mechanosensors: externally applied local mechanical force in-
duces growth of focal contacts by an mDia1-dependent and ROCK-
independent mechanism. J Cell Biol. 2001;153:1175–1185.
12. Tamariz E, Grinnel F. Modulation of fibroblast morphology and
adhesion during collagen matrix remodeling. Mol Biol Cell. 2002;
13. Burridge K, Chrzanowska-Wodnicka C. Focal adhesions, contrac-
tility, and signaling. Ann Rev Cell Dev Biol. 1996;12:463–519.
14. Kolodney MS, Wysolmerski RB. Isometric contraction by fibro-
blasts and endothelial cells in tissue culture: a quantitative study.
J Cell Biol. 1992;117:73–82.
15. Takakuda K, Miyairi H. Tensile behavior of fibroblasts cultured in
collagen gel. Biomaterials. 1996;17:1393–1397.
16. Petroll WM, Cavanagh HD, Barry-Lane P, Andrews P, Jester JV.
Quantitative analysis of stress fiber orientation during corneal
wound contraction. J Cell Sci. 1993;104:353–363.
17. Wakatsuki T, Elson EL. Reciprocal interactions between cells and
extracellular matrix during remodeling of tissue constructs. Bio-
phys Chem. 2003;100:593–605.
18. Pelham RJ Jr, Wang Y. Cell locomotion and focal adhesions are
regulated by substrate flexibility. Proc Natl Acad Sci USA. 1997;
19. Lo CM, Wang HB, Dembo M, Wang YL. Cell movement is guided
by the rigidity of the substrate. Biophys J. 2000;79:144–152.
20. Eastwood M, McGrouther DA, Brown RA. A culture force monitor
for measurement of contraction forces generated in human dermal
fibroblast cultures: Evidence for cell matrix mechanical signalling.
Biochim Biophys Acta. 1994;1201:186–192.
21. Elsdale T, Bard J. Collagen substrata for studies on cell behavior.
J Cell Biol. 1972;54:626–637.
22. Bell E, Ivarsson B, Merril C. Production of a tissue-like structure by
contraction of collagen lattices by human fibroblasts of different
proliferative potential in vivo. Proc Natl Acad Sci USA. 1979;76:
23. Delvoye P, Wiliquest P, Leveque JL, Nusgens BV, Lapiere CM.
Measurement of mechanical forces generated by skin fibroblasts
embedded in a three-dimensional collagen gel. J Invest Dermatol.
24. Brown RA, Prajapati R, McGrouther DA, Yannas IV, Eastwood M.
Tensional homeostasis in dermal fibroblasts: mechanical responses
to mechanical loading in three-dimensional substrates. J Cell
25. Freyman TM, Yannas IV, Yokoo R, Gibson LJ. Fibroblast contractile
force is independent of the stiffness which resists the contraction.
Exp Cell Res. 2002;272:153–162.
26. Petroll WM, Ma L. Direct, dynamic assessment of cell-matrix inter-
actions inside fibrillar collagen lattices. Cell Motil Cytoskeleton.
27. Petroll WM, Ma L, Jester JV. Direct correlation of collagen matrix
deformation with focal adhesion dynamics in living corneal fibro-
blasts. J Cell Sci. 2003;116:1481–1491.
28. Vishwanath M, Ma L, Jester JV, Otey CA, Petroll WM. Modulation
of corneal fibroblast contractility within fibrillar collagen matrices.
Invest Ophthalmol Vis Sci. 2003;44:4724–4735.
29. Jester JV, Huang J, Fisher S, et al. Myofibroblast differentiation of
normal human keratocytes and hTERT, extended-life, human cor-
neal fibroblasts. Invest Ophthalmol Vis Sci. 2003;44:1850–1858.
30. Kaverina I, Krylyshkina O, Small JV. Microtubule targeting of
substrate contacts promotes their relaxation and dissociation.
J Cell Biol. 1999;146:1033–1044.
31. Kaverina I, Krylyshkina O, Gimona M, Beningo K, Wang YL, Small
JV. Enforced polarisation and locomotion of fibroblasts lacking
microtubules. Curr Biol. 2000;10:739–742.
32. Beningo KA, Dembo M, Kaverina I, Small JV, Wang YL. Nascent
focal adhesions are responsible for the generation of strong pro-
pulsive forces in migrating fibroblasts. J Cell Biol. 2001;153:881–
33. Rottner K, Krause M, Gimona M, Small JV, Wehland J. Zyxin is not
colocalized with vasodilator-stimulated phosphoprotein (VASP) at
lamellipodial tips and exhibits different dynamics to vinculin, pax-
illin, and VASP in focal adhesions. Mol Biol Cell. 2001;12:3103–
34. Cukierman E, Pankov R, Stevens DR, Yamada KM. Taking cell-
matrix adhesions to the third dimension. Science. 2001;294:1708–
35. Møller-Pedersen T, Cavanagh HD, Petroll WM, Jester JV. Corneal
haze development after PRK is regulated by volume of stromal
tissue removal. Cornea. 1998;17:627–639.
36. Møller-Pedersen T, Li HF, Petroll WM, Cavanagh HD, Jester JV.
Confocal microscopic characterization of wound repair after pho-
torefractive keratectomy. Invest Ophthalmol Vis Sci. 1998;39:
37. Roy P, Petroll WM, Chuong CJ, Jester JV. Effect of cell migration on
the maintenance of tension on collagen matrix. Ann Biomed Eng.
38. Roeder BA, Kokini K, Sturgis JE, Robinson JP, Voytik-Harbin SL.
Tensile mechanical properties of three-dimensional type I collagen
extracellular matrices with varied microstructure. Trans ASME.
39. Dembo M, Wang YL. Stresses at the cell-to-substrate interface
during locomotion of fibroblasts. Biophys J. 1999;76:2307–2316.
40. Raucher D, Sheetz MP. Cell spreading and lamellipodial extension
rate is regulated by membrane tension. J Cell Biol. 2000;148:127–
41. Roy P, Petroll WM, Cavanagh HD, Chuong CJ, Jester JV. An in vitro
force measurement assay to study the early mechanical interaction
between corneal fibroblasts and collagen matrix. Exp Cell Res.
42. Wang H, Dembo M, Hanks SK, Wang Y. Focal adhesion kinase is
involved in mechanosensing during fibroblast migration. Proc Nat
Acad Sci. 2001;98:11295–11300.
43. Sander EE, ten Klooster JP, van Delft S, van der Kammen RA,
Collard JG. Rac downregulates Rho activity: reciprocal balance
between both GTPases determines cellular morphology and mi-
gratory behavior. J Cell Biol. 1999;147:1009–1021.
44. Parizi M, Howard EW, Tomasek JJ. Regulation of LPA-promoted
myofibroblast contraction: role of rho, myosin light chain kinase,
and myosin light chain phosphatase. Exp Cell Res. 2000;254:210–
45. Totsukawa G, Yamakita Y, Yamashiro S, Hartshorne DJ, Sasaki Y.
Distinct roles of ROCK (Rho-kinase) and MLCK in spatial regula-
tion of MLC phosphorylation for assembly of stress fibers and focal
adhesions in 3T3 fibroblasts. J Cell Biol. 2000;150:797–806.
46. Rottner K, Hall A, Small JV. Interplay between Rac and Rho in the
control of substrate contact dynamics. Curr Biol. 1999;9:640–
47. Svitkina TM, Borisy GG. Arp2/3 complex and actin depolymerizing
factor/cofilin in dendritic organization and treadmilling of actin
filament array in lamellipodia. J Cell Biol. 1999;145:1009–1026.
48. Demali KA, Burridge K. Coupling membrane protrusion and cell
adhesion. J Cell Sci. 2003;116:2389–2397.
49. Smith PG, Roy C, Zhang YN, Chauduri S. Mechanical stress in-
creases RhoA activation in airway smooth muscle cells. Am J
Respir Cell Mol Biol. 2003;28:436–442.
50. Wojciak-Strothard B, Ridley AJ. Shear stress-induced endothelial
cell polarization is mediated by Rho and Rac but not Cdc42 or PI
3-kinases. J Cell Biol. 2003;161:429–439.
51. Grinnell F. Fibroblast-collagen matrix contraction: growth-factor
signalling and mechanical loading. Trends Cell Biol. 2000;10:362–
3474 Petroll et al.
IOVS, October 2004, Vol. 45, No. 10