![FIG. 2. G 0 (a), G00 (b) of HASM cells vs frequency under control conditions ( , n 256), and after 10 min treatment with histamine [1024 M] ( , n 195), DBcAMP [1023 M] ( , n 239) and cytochalasin D [2 3 1026 M] ( , n 171). The solid lines are the fit of Eq. (1) to the data and have been obtained by minimizing mean square residuals for all G0 and G00 data over the measurement bandwidth (G0 53.6 kPa, F0 2.5 3 108 rad s, m 1.41 Pa s). G was computed as ag̃, with a being derived from a finite element analysis of a representative bead-cell geometry. (c) Extrapolation of Eq. (1) to higher frequencies yields crossover at coordinate G0,F0 . (d) G0 and G00 vs f under control conditions.](/figures/fig-2-g-0-a-g00-b-of-hasm-cells-vs-frequency-under-control-2i5058l7.webp)
VOLUME 87, N
UMBER 14 PHYSICAL REVIEW LETTERS 1O
CTOBER 2001
Scaling the Microrheology of Living Cells
Ben Fabry,
1,
* Geoffrey N. Maksym,
2
James P. Butler,
1
Michael Glogauer,
3
Daniel Navajas,
4
and Jeffrey J. Fredberg
1
1
Physiology Program, Harvard School of Public Health, 665 Huntington Avenue, Boston, Massachusetts 02115
2
School of Biomedical Engineering, Dalhousie University, 5981 University Avenue, Halifax B3H 3J5, Canada
3
Division of Experimental Medicine, Harvard Medical School, 221 Longwood Avenue, Boston, Massachusetts 02115
4
Unitat Biofísica i Bioenginyeria, Universitat de Barcelona-IDIBAPS, Casanova 143, 08036 Barcelona, Spain
(Received 27 November 2000; published 13 September 2001)
We report a scaling law that governs both the elastic and frictional properties of a wide variety of
living cell types, over a wide range of time scales and under a variety of biological interventions. This
scaling identifies these cells as soft glassy materials existing close to a glass transition, and implies that
cytoskeletal proteins may regulate cell mechanical properties mainly by modulating the effective noise
temperature of the matrix. The practical implications are that the effective noise temperature is an easily
quantified measure of the ability of the cytoskeleton to deform, flow, and reorganize.
DOI: 10.1103/PhysRevLett.87.148102 PACS numbers: 87.16.Ka, 64.70.Pf, 83.85.Vb, 87.19.Rr
Mechanical stresses and resulting deformations play
central roles in cell contraction, spreading, crawling,
invasion, wound healing, and division, and have been
implicated in regulation of protein and DNA synthesis and
programed cell death [1]. If the cytoskeleton were simply
an elastic body, it would maintain its structural integrity by
developing internal elastic stresses to counterbalance what-
ever force fields it might be subject to. However, those
same elastic stresses would tend to oppose —or even pre-
clude altogether —other essential mechanical functions
such as cell spreading, crawling, extravasation, invasion,
division, and contraction, all of which require the cell to
“flow” similar to a liquid. A liquidlike cell, however, would
be unable to maintain its structural integrity. The classical
resolution of this paradox has been the idea that cytoskele-
tal polymers go through a sol-gel transition, allowing
the cytoskeleton to be fluidlike in some circumstances
(the sol phase) and solidlike in others (the gel phase)
[2–4]. The data presented here suggest that, rather than
being thought of as a gel, the cytoskeleton may be thought
of more properly as a glassy material existing close to a
glass transition, and that disorder and metastability may
be essential features underlying its mechanical functions.
We coated ferrimagnetic microbeads (
4.5 mm diameter)
with a synthetic RGD (Arg-Gly-Asp)-containing peptide
(Integra Life Sciences) and bound them specifically to in-
tegrin receptors on the surface of human airway smooth
muscle (HASM) cells (Figs. 1a and 1b). The beads were
magnetized horizontally and then twisted vertically by an
external homogeneous magnetic field that was varying si-
nusoidally in time (Fig. 1c).
Lateral bead displacement in response to the resulting
oscillatory torque was detected by a charge-coupled de-
vice camera (JAI CV-M10) mounted on an inverted micro-
scope. Image acquisition (exposure time: 0.1 ms) was
phase locked to the twisting field so that 16 images were
acquired during each twisting cycle. Heterodyning was
used at twisting frequencies .1 Hz. The images were
analyzed using an intensity-weighted center-of-mass algo-
rithm in which subpixel arithmetic allowed the determina-
tion of bead position with an accuracy of 5 nm (rms).
The specific torque,
T, is the mechanical torque per bead
volume, and has dimensions of stress (Pa). The ratio of
the complex specific torque
˜
T to the resulting complex
bead displacement
˜
d defines a complex elastic modulus of
the cell ˜g
苷
˜
T共 f兲兾
˜
d共 f兲, and has dimensions of Pa兾nm.
For each bead, we computed the elastic modulus g
0
(the
real part of ˜g), the loss modulus g
00
(the imaginary part
of ˜g), and the loss tangent h (the ratio g
00
兾g
0
). These
measurements could be transformed into traditional elas-
tic and loss moduli by a geometric factor a that depends
on the shape and thickness of the cell, and the degree of
bead embedding, where
˜
G 苷 a ˜g. Finite element analysis
of cell deformation for a representative bead-cell geome-
try (assuming homogeneous and isotropic elastic proper-
ties with 10% of the bead diameter embedded in a cell
5 mm high) sets a to 6.8 mm. This geometric factor need
serve only as a rough approximation because it cancels out
in the scaling procedure described below, which is model
independent.
FIG. 1. (a) Ferrimagnetic beads (arrow) bind avidly to the actin
cytoskeleton (stained with phalloidin) of HASM cells via cell ad-
hesion molecules (integrins). (b) Scanning electron microscopy
of a bead bound to the cell surface. (c) A magnetic field intro-
duces a torque which causes the bead to rotate and to displace.
M denotes the direction of the bead’s magnetic moment.
148102-1 0031-9007兾01兾87(14)兾148102(4)$15.00 © 2001 The American Physical Society 148102-1
VOLUME 87, N
UMBER 14 PHYSICAL REVIEW LETTERS 1O
CTOBER 2001
Bead displacement amplitude followed a log-normal dis-
tribution that has been analyzed in detail elsewhere [5].
The loss tangent and the frequency dependence of G
0
and
G
00
varied little between beads, however. Importantly, G
0
and G
00
were found to be constant over a wide range of
specific torque amplitudes (1.8–130 Pa) and lateral bead
displacement amplitudes (5–500 nm), implying linear me-
chanical behavior in this range.
G
0
increased weakly with frequency and followed a
power law, f
0.17
(Fig. 2a). G
00
also followed a weak power
law at low frequencies, but beyond 10 Hz progressively
stronger frequency dependence emerged, approaching
but never quite attaining a power-law exponent of 1,
which would be characteristic of a Newtonian viscosity
(Fig. 2b). G
0
was larger than G
00
for twisting frequencies
below 300 Hz (Fig. 2d), and the loss tangent remained
almost constant between 0.3 and 0.4 over the lower four
frequency decades. We found similar behavior when
human bronchial epithelial cells were probed between
0.1 and 10 Hz using atomic force microscopy (data not
shown). Comparable findings over more limited ranges of
10
-2
10
0
10
2
10
4
10
3
10
2
10
5
f [Hz]
G'
G"
[Pa]
10
5
10
4
10
3
10
2
10
1
10
-2
10
0
10
2
10
4
10
6
10
8
f [Hz]
G' [Pa]
(G
0
,Φ
0
/2π)
10
4
10
3
10
2
10
1
G' [Pa]
10
4
10
5
10
3
10
2
10
1
10
-2
10
-1
10
0
10
1
10
2
10
3
f [Hz]
G" [Pa]
(a)
(b)
(c)
(d)
FIG. 2. G
0
(a), G
00
(b) of HASM cells vs frequency under
control conditions (䊏, n 苷 256), and after 10 min treatment
with histamine [10
24
M] (䊐, n 苷 195), DBcAMP [10
23
M]
(䉱, n 苷 239) and cytochalasin D [2 3 10
26
M] (䉭, n 苷 171).
The solid lines are the fit of Eq. (1) to the data and have
been obtained by minimizing mean square residuals for all G
0
and G
00
data over the measurement bandwidth (G
0
苷 53.6 kPa,
F
0
苷 2.5 3 10
8
rad兾s, m 苷 1.41 Pa s). G was computed as
a ˜g, with a being derived from a finite element analysis of a
representative bead-cell geometry. (c) Extrapolation of Eq. (1)
to higher frequencies yields crossover at coordinate 共G
0
, F
0
兲.
(d) G
0
and G
00
vs f under control conditions.
frequency have been obtained with different techniques
in several other cell types [6–10]. Experimental models
of cell rheology using semidilute F-actin solutions [11],
however, display rheological characteristics that can differ
rather substantially from the weak power law behavior
observed in living cells (Fig. 2).
Addition of the contractile agonist histamine to HASM
cells caused G
0
and G
00
to increase (Figs. 2a and 2b).
The loss tangent and the exponent of the power law fell
slightly. By contrast, ablating baseline contractile tone
with N
6
,2
0
-O-dibutyryladenosine 3
0
,5
0
-cyclic monophos-
phate (DBcAMP) caused G
0
and G
00
to fall, and the loss
tangent and exponent to increase. When actin filaments
were disrupted with cytochalasin D, G
0
and G
00
fell even
more, and the loss tangent and exponent increased further.
These data, taken together, cannot be explained by
receptor-ligand dynamics as would be the case if there were
repetitive peeling of the bead away from the cell surface
and subsequent reattachment with each bead oscillation.
Neither can these data be readily explained by changes in
cell membrane mechanics. Rather, these data are more
consistent with the notion that the bead binds avidly to the
cytoskeleton via focal adhesions, and that the cytoskeleton
is deformed as the bead rotates (Fig. 1). Accordingly, G
0
and G
00
predominantly reflect the mechanical properties
of the cytoskeleton. This view is further supported by
experiments using beads coated with acetylated low den-
sity lipoprotein (acLDL); those beads bind to scavenger
receptors which do not form focal adhesion complexes
[12]. We found that baseline values of G
0
and G
00
mea-
sured with acLDL-coated beads were smaller by fivefold
compared with beads bound to integrins, and demon-
strated an attenuated response to contractile agonists.
Structural damping.—The data in Fig. 2 conform
closely to an empirical law that is known in the engineer-
ing literature as structural damping or hysteretic damping
[13–18]. Accordingly,
˜
G 苷 G
0
µ
v
F
0
∂
x21
共1 1 jh兲G共2 2 x兲 cos
p
2
共x 2 1兲
1 jvm , (1)
where
h 苷 tan共x 2 1兲p兾2, and v is the radian frequency
2pf. G
0
and F
0
are scale factors for stiffness and fre-
quency, respectively, G denotes the gamma function, m
is a Newtonian viscous term, and j is the unit imaginary
number
p
21. G
0
and m depend on bead-cell geometry.
h has been called the hysteresivity or the structural damp-
ing coefficient. The elastic modulus corresponds to the
real part of Eq. (1), which increases for all v according to
the power-law exponent, x 2 1. The loss modulus corre-
sponds to the imaginary part of Eq. (1) and includes a com-
ponent that is a frequency-independent fraction (
h) of the
elastic modulus; such a direct coupling of the loss modu-
lus to the elastic modulus is the characteristic feature of
structural damping behavior [18]. The loss modulus also
includes a Newtonian viscous term, jvm, which comes
into play only at higher frequencies. At low frequencies,
148102-2 148102-2
VOLUME 87, N
UMBER 14 PHYSICAL REVIEW LETTERS 1O
CTOBER 2001
the loss tangent h approximates h. In the limit that x ap-
proaches unity, the power-law slope approaches zero, G
0
approaches G
0
, and h approaches zero. Thus, Eq. (1) de-
scribes a relationship between changes of the exponent of
the power law and the transition from solidlike (x
苷 1)to
fluidlike (x 苷 2) behavior.
Unexpectedly, we could fit Eq. (1) to the data in Fig. 2
using only a single value for G
0
and for F
0
. This implies a
common intersection of the G
0
vs f curves when extrapo-
lated outside the measurement range (Fig. 2c). Further, m
was found to vary little with cellular manipulations. This
leaves x as the sole free parameter in Eq. (1).
Master curves.—We estimated x directly from the mea-
sured power-law dependence of G
0
on f. We then defined
a normalized stiffness
G as being G
0
measured at 0.75 Hz
(an arbitrary choice) divided by G
0
, which causes the geo-
metric factor a to cancel out. G vs x for all manipu-
lations followed the relationship shown in Fig. 3a (black
symbols). Similarly, h (measured as the ratio of G
00
兾G
0
at
0.75 Hz) vs x followed the relationship shown in Fig. 3b
(black symbols). When we studied a variety of cell types,
corresponding data were found in every case to conform
to Eq. (1); G
0
and m differed between cell types, but F
0
was invariant (2.5 3 10
8
rad兾s).
Despite important differences in the elastic and fric-
tional material properties, all data collapsed onto the very
same relationships as did the HASM cells (Fig. 3). These
relationships thus represent universal master curves in that
a single free parameter, x,defined the constitutive elastic
and frictional behaviors for a variety of cytoskeletal ma-
nipulations, for five frequency decades, and for diverse cell
types. The normalization that leads to these master curves
is model independent. Nonetheless, these master curves
fell close to the predictions from Eq. (1) (solid black lines).
Soft glassy materials.— An attractive physical interpre-
tation for the parameter x comes from the physics of soft
glassy materials (SGM). The empirical criteria that de-
fine this class of material are that G
0
and G
00
increase with
FIG. 3 (color). Master curves (G and h at 0.75 Hz vs x)of
HASM cells (black, n 苷 256), human bronchial epithelial cells
(blue, n 苷 142), mouse embryonic carcinoma cells (F9) cells
(pink, n 苷 50), mouse macrophages (J774A.1) (red, n 苷 46)
and human neutrophils (green, n 苷 42) under control condi-
tions (䊏), treatment with histamine (䊐), FMLP (䉬), DBcAMP
(䉭), and cytoD (䉱). Equation (1) predicts the black solid
curves: lnG 苷 共x 2 1兲 ln共v兾F
0
兲 with F
0
苷 2.5 3 10
8
rad兾s,
and
h 苷 tan关共x 2 1兲p兾2兴. Error bars indicate 61 standard
error.
weak power-law dependencies upon frequency, and that
the loss tangent is frequency insensitive and of the order
0.1 [19,20]. Accordingly, the data in Figs. 2 and 3 estab-
lish that the cells studied here behaved as soft glassy ma-
terials. Surprisingly, other materials in the class include
foams, emulsions, colloid suspensions, pastes, and slurries
[20,21]. Existing models of viscoelasticity have thus far
failed to provide a satisfactory explanation of the rheology
of SGMs [17,20]. Moreover, it is unclear how such diverse
systems express mechanical behavior that is so much alike.
It has been suggested by Sollich [20] that the common
rheological features of this class may be not so much a
reflection of the particular molecules or molecular mecha-
nisms as they are a reflection of a generic system property
at some higher level of structural organization [20,22]. The
generic properties that these SGMs share are that each is
composed of elements that are discrete, numerous, aggre-
gated with one another via weak attractive interactions, and
arrayed in a geometry that is structurally disordered and
metastable. Glassy behavior in such systems is a conse-
quence of structural relaxation in which the elements have
to cross energy barriers that are large compared with ther-
mal energies. A disordered metastable configuration arises
if the elements are quenched before they can attain a mini-
mum energy (equilibrium) state.
Sollich [20] extended the earlier work of Bouchaud [23]
to develop a unified theory of soft glassy rheology (SGR).
SGR theory considers that the individual elements of the
matrix exist within an energy landscape containing many
wells, or traps, formed by neighboring elements. These
traps are of differing energy depth, and each is regarded as
being so deep that the element is unable to escape the well
by thermal fluctuations. Instead, elements are imagined
to be agitated and rearranged by their mutual interactions
within the matrix [20]. A clear notion of the source of
the agitation remains to be identified, but this agitation can
be represented nonetheless by an effective temperature, or
noise level,
x. The elements are unable to escape their
wells when x 苷 1, in which case the system is perfectly
elastic. When x . 1, however, the elements can hop ran-
domly between wells and, as a result, the system can flow
and become disordered. Flow and disorder are the essential
features of a glassy material. Thus, as x approaches unity
from above, the system approaches a glass transition. In
the limit of linear responses at low frequencies, SGR theory
[20] leads to a constitutive equation that we now identify
as being identical to the structural damping law [Eq. (1)].
Accordingly, SGR theory provides the following inter-
pretation of the free parameter x and the master curves
reported here (Fig. 3). Agents that activate the contractile
apparatus and/or polymerize cytoskeletal proteins move
the cell towards a glass transition. A decreasing noise
temperature is consistent with the formation of structure,
which in turn may be associated with the notion of
de Gennes [24] that chemical bonding is equivalent to
attractive interactions; the system becomes more ordered
and approaches a frozen state. Conversely, relaxing
148102-3 148102-3
VOLUME 87, N
UMBER 14 PHYSICAL REVIEW LETTERS 1O
CTOBER 2001
agonists and agents that disrupt the cytoskeleton cause the
system to become more disordered and move towards a
fluid state. SGR theory also identifies the parameter
G
0
in Eq. (1) as being the stiffness at the glass transition,
and the parameter F
0
as being the frequency at which
elements attempt to escape their energy wells. If so,
then G
0
represents an upper limit of stiffness and might
correspond to the stiffness defined by the purely elastic
cytoskeletal model of Satcher and Dewey [25].
Time-scale invariance.—The attempt rate F
0
is not to
be confused with the frequency at which elements escape
their energy wells, and does not correspond to a relax-
ation time scale [19]. Rather, the inverse Fourier trans-
form of Eq. (1) predicts a wide range of relaxation time
scales [13,15], relaxing according to the power law t
1
2x
.
Power-law behavior implies that the relaxation processes
are not tied to any particular internal time scale.
Regulation of cell mechanics.—Although glasslike be-
havior has been reported for protein macromolecules [26],
we do not necessarily expect that SGR theory will apply
to mechanical properties at the level of discrete molecu-
lar interactions. Moreover, the data presented here were
restricted to small amplitude deformations coupled to the
cytoskeleton via integrins, and to a limited subset of cells
and substrates. Nonetheless, these data lead to the prospect
of a major conceptual simplification. The observation
of universal master relationships describing cell rheology
supports the hypothesis that details of the molecular inter-
actions are manifested largely through their effects on x.
This behavior is reminiscent of statistical mechanics in sys-
tems that have only a weak dependence of aggregate be-
havior on the details of the molecular interactions. We
speculate that some key functions of diverse cytoskeletal
proteins, such as their influence on the ability of the cy-
toskeleton to deform, flow, and reorganize, may be un-
derstood mainly through their ability to modulate x.We
have no evidence, however, that x is a regulated vari-
able. SGR theory cannot predict how much a given drug
might change x, but it does predict the effects of any drug-
induced change in x on cytoskeletal rheology. As a prac-
tical matter, it provides a conceptual framework for assays
of drug potency.
The rheological behavior reported here stands in contrast
with prevailing theories of cell mechanics, which hold that
cell rheology arises from an interaction of distinct elas-
tic and viscous components expressing a limited range of
characteristic relaxation times. Except at frequencies in ex-
cess of several tens of Hz, the data reported here support
the hypothesis that the locus of both the frictional and the
elastic stresses is within formed structures of the cytoskele-
ton [9,17,20], and that the dominant frictional stress does
not correspond to a viscous stress.
To the extent that the concept of the sol-gel transition
can be applied usefully to the mechanics of these formed
cytoskeletal structures, the data presented here suggest that
the gel phase may be thought of more properly as a glassy
material existing close to a glass transition; if so, then
disorder and metastability may be essential features un-
derlying mechanical functions of the cytoskeleton. A key
thermodynamic distinction between the sol-gel transition
(in a strong gel) and the glass transition is that the former
is an equilibrium process, with structural elements found at
energy minima, whereas the latter is a nonequilibrium pro-
cess with elements trapped in metastable states. Finally,
we wish to point out that it is unclear at this time how
the level of metastability and intracellular agitation as ex-
pressed by the effective noise temperature might depend
upon how far the cell departs from thermodynamic equi-
librium due to its energy metabolism, but this would seem
to be an important question.
We thank S. Shore, P. Moore, W. Goldmann, C. F.
Dewey, D. Tschumperlin, S. Mijailovich, W. Möller,
M. Harp, R. Panettieri, R. Rogers, and D. Weitz. This
study was supported by NIH HL33009, HL65960,
HL59682, and DGESIC-PM980027.
*Corresponding author.
Email address: bfabry@hsph.harvard.edu
[1] P. A. Janmey, Physiol. Rev. 78
, 763 (1998).
[2] P. A. Janmey et al., Nature (London) 345
, 89 (1990).
[3] T. P. Stossel, Science 260
,
1086 (1993).
[4] M. Tempel, G. Isenberg, and E. Sackmann, Phys. Rev. E
54
, 1802 (1996).
[5] B. Fabry et al., J. Appl. Physiol. 91
, 986 (2001).
[6] S. G. Shroff, D. R. Saner, and R. Lal, Am. J. Physiol. 269
,
C286 (1995).
[7] R. E. Mahaffy et al., Phys. Rev. Lett. 85
, 880 (2000).
[8] W. H. Goldmann and R. M. Ezzell, Exp. Cell Res. 226
, 234
(1996).
[9] G. N. Maksym et al., J. Appl. Physiol. 89
, 1619 (2000).
[10] S. Yamada, D. Wirtz, and S. C. Kuo, Biophys. J. 78
, 1736
(2000).
[11] T. Gisler and D. A. Weitz, Phys. Rev. Lett. 82
, 1606 (1999).
[12] N. Wang, J. P. Butler, and D. E. Ingber, Science 260
, 1124
(1993).
[13] W. Weber, Ann. Phys. Chem. 54
, 1 (1841).
[14] F. Kohlrausch, Ann. Phys. Chem. 128
, 1 (1866).
[15] J. Hildebrandt, Bull. Math. Biophys. 31
, 651 (1969).
[16] Y. C. Fung, Am. J. Physiol. 213
, 1532 (1967).
[17] J. J. Fredberg and D. Stamenovic, J. Appl. Physiol. 67
,
2408 (1989).
[18] S. H. Crandall, J. Sound Vib. 11
, 3 (1970).
[19] P. Sollich et al., Phys. Rev. Lett. 78
, 2020 (1997).
[20] P. Sollich, Phys. Rev. E 58
, 738 (1998).
[21] V. Trappe and D. Weitz, Phys. Rev. Lett. 85
, 449 (2000).
[22] J. Phillips, Rep. Prog. Phys. 59
, 1133 (1996).
[23] J. Bouchaud, J. Phys. I (France) 2
, 1705 (1992).
[24] P. G. de Gennes, Scaling Concepts in Polymer Physics
(Cornell University Press, Ithaca, 1979), pp. 134 –152.
[25] R. L. Satcher, Jr. and C. F. Dewey, Jr., Biophys. J. 71
, 109
(1996).
[26] A. L. Lee and A. J. Wand, Nature (London) 411
, 501
(2001).
148102-4 148102-4