20. E. Smirnova, D. Shurland, S. Ryazantsev, A. Van Der
Bliek, J. Cell Biol. 143, 351 (1998).
21. A. M. Labrousse, M. D. Zappaterra, D. A. Rube, A. M.
van der Bliek, Mol. Cell 4, 815 (1999).
22. H. Sesaki, R. E. Jensen, J. Cell Biol. 147, 699 (1999).
23. S. Arimura, N. Tsutsumi, Proc. Natl. Acad. Sci. U.S.A.
99, 5727 (2002).
24. A. M. van der Bliek, Trends Cell Biol. 9, 96 (1999).
25. J. E. Hinshaw, Annu. Rev. Cell Dev. Biol. 16, 483 (2000).
26. J. E. Hinshaw, S. L. Schmid, Nature 374, 190 (1995).
27. J. E. Hinshaw, Curr. Opin. Struct. Biol. 9, 260 (1999).
28. M. H. Stowell, B. Marks, P. Wigge, H. T. McMahon,
Nature Cell Biol. 1, 27 (1999).
29. B. Marks et al., Nature 410, 231 (2001).
30. N. H. Fukushima, E. Brisch, B. R. Keegan, W. Bleazard,
J. M. Shaw, Mol. Biol. Cell 12, 2756 (2001).
31. S. Sever, Curr. Opin. Cell Biol. 14, 463 (2002).
32. S. L. Newmyer, A. Christensen, S. Sever, Dev. Cell 4,
929 (2003).
33. H. Gao, D. Kadirjan-Kalbach, J. E. Froehlich, K. W. Oster-
young, Proc. Natl. Acad. Sci. U.S.A. 100, 4328 (2003).
34. S. Miyagishima, K. Nishida, T. Kuroiwa, Trends Plant
Sci. 8, 432 (2003).
35. S. Miyagishima et al., Plant Cell 13, 2257 (2001).
36. S. Y. Miyagishima et al., Plant Cell 15, 655 (2003).
37. K. Nishida et al., Proc. Natl. Acad. Sci. U.S.A. 100,
2146 (2003).
38. A. Legesse-Miller, R. H. Massol, T. Kirchhausen, Mol.
Biol. Cell 14, 1953 (2003).
39. M. Fujiwara, S. Yoshida, Biochem. Biophys. Res. Com-
mun. 287, 462 (2001).
40. R. S. McAndrew, J. E. Froehlich, S. Vitha, K. D. Stokes,
K. W. Osteryoung, Plant Physiol. 127, 1656 (2001).
41. S. Sever, H. Damke, S. L. Schmid, J. Cell Biol. 150,
1137 (2000).
42. J. Nunnari, unpublished data.
43. A. G. McArthur et al., FEMS Microbiol. Lett. 189, 271
(2000).
44. M. Marti et al., J. Biol. Chem. 278, 24837 (2003).
45. C. A. Vater, C. K. Raymond, K. Ekena, I. Howald-Stevenson,
T. H. Stevens, J. Cell Biol. 119, 773 (1992).
46. D. Hoepfner, M. van den Berg, P. Philippsen, H. F.
Tabak, E. H. Hettema, J. Cell Biol. 155, 979 (2001).
47. J. Nunnari, H. Gao, K. W. Osteryoung, unpublished data.
48. F. van den Ent, L. Amos, J. Lowe, Curr. Opin. Micro-
biol. 4, 634 (2001).
49. K. W. Osteryoung, K. D. Stokes, S. M. Rutherford, A. L.
Percival, W. Y. Lee, Plant Cell 10, 1991 (1998).
50. R. Strepp, S. Scholz, S. Kruse, V. Speth, R. Reski, Proc.
Natl. Acad. Sci. U.S.A. 95, 4368 (1998).
51. K. W. Osteryoung, R. S. McAndrew, Annu. Rev. Plant
Physiol. Plant Mol. Biol. 52, 315 (2001).
52. K. D. Stokes, K. W. Osteryoung, Gene 320, 97 (2003).
53. S. Vitha, R. S. McAndrew, K. W. Osteryoung, J. Cell
Biol. 153, 111 (2001).
54. S. Vitha et al., Plant Cell 15, 1918 (2003).
55. R. McAndrew, C. Chi-Ham, J. E. Froehlich, K. W.
Osteryoung, unpublished data.
56. A. Gaikwad, V. Babbarwal, V. Pant, S. K. Mukherjee,
Mol. Gen. Genet. 263, 213 (2000).
57. M. El-Shami, S. El-Kafafi, D. Falconet, S. Lerbs-Mache,
Mol. Genet. Genomics 267, 254 (2002).
58. X. Ma, W. Margolin, J. Bacteriol. 181, 7531 (1999).
59. J. Errington, R. A. Daniel, D. J. Scheffers, Microbiol.
Mol. Biol. Rev. 67, 52 (2003).
60. J. Kiessling et al., J. Cell Biol. 151, 945 (2000).
61. K. D. Stokes, R. S. McAndrew, R. Figueroa, S. Vitha,
K. W. Osteryoung, Plant Physiol. 124, 1668 (2000).
62. J. L. Marrison et al., Plant J. 18, 651 (1999).
63. S. Walter, J. Buchner, Angew. Chem. Int. Ed. Engl. 41,
1098 (2002).
64. O. A. Koksharova, C. P. Wolk, J. Bacteriol. 184, 5524
(2002).
65. H. Fulgosi, L. Gerdes, S. Westphal, C. Glockmann, J.
Soll, Proc. Natl. Acad. Sci. U.S.A. 99, 11501 (2002).
66. J. Maple, N. H. Chua, S. G. Moller, Plant J. 31, 269 (2002).
67. K. S. Colletti et al., Curr. Biol. 10, 507 (2000).
68. R. Itoh, M. Fujiwara, N. Nagata, S. Yoshida, Plant
Physiol. 127, 1644 (2001).
69. L. Griparic, A. M. van der Bliek, Traffic 2, 235 (2001).
70. E. Smirnova, L. Griparic, D. L. Shurland, A. M. van der
Bliek, Mol. Biol. Cell 12, 2245 (2001).
71. K. R. Pitts, Y. Yoon, E. W. Krueger, M. A. McNiven,
Mol. Biol. Cell 10, 4403 (1999).
72. S. Jakobs et al., J. Cell Sci. 116, 2005 (2003).
73. K. L. Cerveny, J. M. McCaffery, R. E. Jensen, Mol. Biol.
Cell 12, 309 (2001).
74. Q. Tieu, J. Nunnari, J. Cell Biol. 151, 353 (2000).
75. P. Fekkes, K. A. Shepard, M. P. Yaffe, J. Cell Biol. 151,
333 (2000).
76. A. D. Mozdy, J. M. McCaffery, J. M. Shaw, J. Cell Biol.
151, 367 (2000).
77. Q. Tieu, V. Okreglak, K. Naylor, J. Nunnari, J. Cell Biol.
158, 445 (2002).
78. K. L. Cerveny, R. E. Jensen, Mol. Biol. Cell 14, 4126 (2003).
79. Y. Yoon, E. W. Krueger, B. J. Oswald, M. A. McNiven,
Mol. Cell. Biol. 23, 5409 (2003).
80. D. I. James, P. A. Parone, Y. Mattenberger, J.-C. Mar-
tinou, J. Biol. Chem. 278, 36373 (2003).
81. H. Berman et al., Nucleic Acids Res. 28, 235 (2000).
82. PDBID: 1IYG, W. Ohashi et al., unpublished data.
83. D. Rube, S. Gandre, A. van der Bliek, 14th Interna-
tional C. elegans Conference, 29 June to 3 July 2003,
Los Angeles, Abstract 845B.
84. W. B. Huttner, A. Schmidt, Curr. Opin. Neurobiol. 10,
543 (2000).
85. K. Farsad et al., J. Cell Biol. 155, 193 (2001).
86. A. Schmidt et al., Nature 401, 133 (1999).
87. H. Hashimoto, Int. Rev. Cytol. 222, 63 (2003).
88. N. Garrido et al., Mol. Biol. Cell 14, 1583 (2003).
89. W. J. Jeong et al., Plant Physiol. 129, 112 (2002).
90. J. Nunnari et al., Mol. Biol. Cell 8, 1233 (1997).
91. We regret that space constraints have prevented us from
citing many relevant papers. We thank S. Merchant and S.
Miyagishima for critical reading of the manuscript; P. Beech
and A. van der Bliek for sharing relevant findings before
publication; S. Vitha, D. Yoder, S. Miyagishima, K. Naylor, A.
Stone, C. Song, and H. Gao for providing images for figures;
and all the members of our laboratories for invaluable
contributions. Supported by NSF grants 0092448 (K.W.O.),
0313520 (K.W.O.), and 0110899 (J.N.); NIH grant
R01GM62942A ( J.N.); and the Michigan State University
Center for Plant Products and Technologies (K.W.O.).
Cell Migration: Integrating Signals from
Front to Back
Anne J. Ridley,
1
Martin A. Schwartz,
2
Keith Burridge,
5
Richard A. Firtel,
6
Mark H. Ginsberg,
7
Gary Borisy,
8
J. Thomas Parsons,
3
Alan Rick Horwitz
4
Cell migration is a highly integrated multistep process that orchestrates embryonic mor-
phogenesis; contributes to tissue repair and regeneration; and drives disease progression in
cancer, mental retardation, atherosclerosis, and arthritis. The migrating cell is highly polarized
with complex regulatory pathways that spatially and temporally integrate its component
processes. This review describes the mechanisms underlying the major steps of migration and
the signaling pathways that regulate them, and outlines recent advances investigating the
nature of polarity in migrating cells and the pathways that establish it.
O
ur liaison with cell migration, as hu-
mans, begins shortly after concep-
tion, accompanies us throughout life,
and often contributes to our death. Although
migratory phenomena are apparent as early
as implantation, cell migration orchestrates
morphogenesis throughout embryonic devel-
opment (1). During gastrulation, for example,
large groups of cells migrate collectively as
sheets to form the resulting three-layer em-
bryo. Subsequently, cells migrate from vari-
ous epithelial layers to target locations, where
they then differentiate to form the specialized
cells that make up different tissues and or-
gans. Analogous migrations occur in the
adult. In the renewal of skin and intestine,
fresh epithelial cells migrate up from the
basal layer and the crypts, respectively. Mi-
gration is also a prominent component of
tissue repair and immune surveillance, in
which leukocytes from the circulation mi-
grate into the surrounding tissue to destroy
invading microorganisms and infected cells
and to clear debris. The importance of cell
migration however, goes far beyond humans
and extends to plants and even to single-
celled organisms (2).
Migration contributes to several important
pathological processes, including vascular dis-
ease, osteoporosis, chronic inflammatory dis-
eases such as rheumatoid arthritis and multiple
1
Ludwig Institute for Cancer Research, Royal Free and
University College School of Medicine, London W1W
7BS, UK.
2
Departments of Microbiology and Biomed-
ical Engineering, Cardiovascular Research Center and
Mellon Prostate Cancer Research Institute;
3
Depart-
ment of Microbiology;
4
Department of Cell Biology;
University of Virginia School of Medicine, Charlottes-
ville, VA 22908, USA.
5
Department of Cell and Devel-
opmental Biology and Lineberger Comprehensive
Cancer Center, University of North Carolina, Chapel
Hill, NC 27599, USA.
6
Section of Cell and Develop-
mental Biology, Division of Biological Sciences and
Center for Molecular Genetics, University of Califor-
nia, San Diego, 9500 Gilman Drive, La Jolla, CA
92093–0634, USA.
7
Department of Cell Biology, The
Scripps Research Institute, 10550 North Torrey Pines
Road, La Jolla, CA 92037, USA.
8
Department of Cel-
lular and Molecular Biology, Northwestern University,
School of Medicine, Chicago, IL 60611, USA.
R EVIEWS
5 DECEMBER 2003 VOL 302 SCIENCE www.sciencemag.org1704
on May 2, 2018 http://science.sciencemag.org/Downloaded from
sclerosis, cancer, and mental retardation. Thus,
understanding the fundamental mechanisms un-
derlying cell migration holds the promise of
effective therapeutic approaches for treating
disease, cellular transplantation, and the prepa-
ration of artificial tissues.
Over the past few years, immense progress
has been made in understanding cell migration,
including the establishment of polar structures,
the regulation of the dynamic processes of actin
and microtubule polymerization, and the regu-
lation of spatial and temporal signal transduc-
tion. This review summarizes and highlights
some of these advances in the context of the
need to integrate and coordinate the many cel-
lular events that compose migration.
The Migration Cycle
Our present understanding of cell migration is a
composite derived from studies of different cell
types and environments. In general, cell migra-
tion can be usefully conceptualized as a cyclic
process (3). The initial response of a cell to a
migration-promoting agent is to polarize and
extend protrusions in the direction of migration.
These protrusions can be large, broad lamelli-
podia or spike-like filopodia, are usually driven
by actin polymerization, and are stabilized by
adhering to the extracellular matrix (ECM) or
adjacent cells via transmembrane receptors
linked to the actin cytoskeleton. These adhe-
sions serve as traction sites for migration as the
cell moves forward over them, and they are
disassembled at the cell rear, allowing it to
detach. Interestingly, the movement of cell
sheets shows some features of single-cell mi-
gration; however, the polarization extends
across the sheet.
Although many aspects of this picture are
shared among different cell types, the details
can differ greatly. For example, these steps are
observed most distinctly in slow-moving cells
such as fibroblasts, but are not as obvious in
fast-moving cells such as neutrophils, which
seem to glide over the substratum. In addition,
a cell’s migratory behavior depends on its en-
vironment. Somitic cells migrating in vivo, for
example, show large single protrusions and
highly directed migration, in contrast to the
multiple small protrusions they display on pla-
nar substrates; and cancer cells can modify their
morphology and nature of migration in re-
sponse to environmental changes (4, 5).
The Protrusive Machinery
Actin filaments are intrinsically polarized with
fast-growing “barbed” ends and slow-growing
“pointed” ends, and this inherent polarity is
used to drive membrane protrusion. However,
the organization of filaments depends on the
type of protrusion: In lamellipodia, actin fila-
ments form a branching “dendritic” network,
whereas in filopodia they are organized into
long parallel bundles (6). Actin polymerization
in lamellipodia is mediated by the Arp2/3 com-
plex, which binds to the sides or tip of a pre-
existing actin filament and induces the forma-
tion of a new daughter filament that branches
off the mother filament (6, 7). Activation of the
Arp2/3 complex is localized by WASP/WAVE
family members, which are themselves activat-
ed at the cell membrane (6) (see below). Push-
ing of the membrane, the actual protrusive
event, is believed to occur not by elonga-
tion of the actin filament per se but by an
“elastic Brownian ratchet” mechanism, in
which thermal energy bends the nascent
short filaments, storing elastic energy. Un-
bending of an elongated filament against
the leading edge would then provide the
driving force for protrusion (7).
Several actin-binding proteins regulate the
rate and organization of actin polymerization in
protrusions by affecting the pool of available
monomers and free ends (7, 8). For example,
profilin prevents self-nucleation by binding to
actin monomers and also serves to selectively
target monomers to barbed ends. Filament elon-
gation is terminated by capping proteins, there-
by restricting polymerization to new filaments
close to the plasma membrane. In addition, dis-
assembly of older filaments, which is needed to
generate actin monomers for polymerization at
the front end, is assisted by proteins of the
ADF/cofilin family, which sever filaments and
promote actin dissociation from the pointed end.
Other proteins play supporting roles in the den-
dritic network: Cortactin stabilizes branches,
whereas filamin A and ␣-actinin stabilize the
entire network by cross-linking filaments (6).
Filopodial protrusion is thought to occur
by a filament treadmilling mechanism, in
which actin filaments within a bundle elon-
gate at their barbed ends and release actin
monomers from their pointed ends (6). The
long and unbranched filament organization is
consistent with assembly occurring by elon-
gation rather than by branched nucleation.
Many proteins are enriched at filopodial tips,
including Ena/VASP proteins, which bind
barbed ends of actin filaments and antagonize
both capping and branching, thereby allow-
ing continuous elongation of filaments and
fascin, which bundles actin filaments and
might thereby generate the stiffness needed to
allow efficient pushing of the plasma mem-
brane in filopodia (6).
Fig. 1. The polarized cell. (A) PIP
3
: Leading-edge localization of a green fluorescent protein (GFP)
fusion of the PH domain of Akt/PKB in chemotaxing Dictyostelium cells. (Micrograph by R. Meili and
R. Firtel.) (B) Phosphorylated ␣4 integrin: Localization at the leading edge of phosphorylated ␣4
integrin expressed in migrating CHO cells. The localization was assayed by immunostaining with an
antibody directed against the phosphorylated ␣4 integrin cytoplasmic domain. [Micrograph by L. E.
Goldfinger and M. H. Ginsberg, reproduced from The Journal of Cell Biology, 2003, Vol. 162, p. 732,
by copyright permission of The Rockefeller University Press] (C) PTEN: Localization of a GFP fusion
of PTEN in chemotaxing Dictyostelium cells. PTEN is absent from the leading edge but present along
the lateral sides and posterior of the cell. (Micrograph by R. Meili and R. Firtel.) (D) Activated Rac:
Activated Rac localizes preferentially with an effector in the leading edge of migrating 3T3 cells.
The interaction of GFP-V12Rac with an effector domain was assayed by fluorescence resonance
energy transfer (FRET). The enhanced interaction in the leading edge is due to locally regulated
membrane targeting of the V12Rac. Red and blue represent high and low intensities of FRET (that
is, of interaction), respectively. (Micrograph by M. Del Pozo, W. B. Kiosses, and M. A. Schwartz.)
R EVIEWS
www.sciencemag.org SCIENCE VOL 302 5 DECEMBER 2003 1705
on May 2, 2018 http://science.sciencemag.org/Downloaded from
The supramolecular design of lamellipo-
dia and filopodia endows them with the ca-
pacity to perform distinct functions. Biophys-
ical considerations suggest that the dendritic
organization of lamellipodia provides a tight
brush-like structure that is able to push along
a broad length of plasma membrane (7).
Through localized activation of the Arp2/3
complex, the lamellipodium could be induced
to grow in a particular direction, providing
the basis for directional migration. In con-
trast, filopodia, with their parallel bundle or-
ganization, are particularly well designed to
serve as sensors and to explore the local
environment, although they are not essential
for chemotaxis.
Rho family proteins: Central regulators
of protrusion. Rho family small guano-
sine triphosphate (GTP)– binding proteins
(GTPases) are pivotal regulators of actin and
adhesion organization and control the forma-
tion of lamellipodia and filopodia. They are
conformationally regulated by the binding of
GTP and GDP: When bound to GTP, they are
active and interact with their downstream
target proteins, which include protein ki-
nases, lipid-modifying enzymes, and activa-
tors of the Arp2/3 complex (9). Rho GTPases
are activated by guanine nucleotide exchange
factors (GEFs) and inactivated by GTPase
activating proteins (GAPs). Of the Rho
GTPases, Rac, Cdc42, and RhoG are required
for protrusion of lamellipodia and filopodia.
The major targets for Rac and Cdc42 that
mediate actin polymerization in protrusions are
the WASP/WAVE family of Arp2/3 complex
activators. Rac stimulates lamellipodial exten-
sion by activating WAVE proteins (10). Cdc42
binds to WASP proteins, and in vitro this stim-
ulates the Arp2/3 complex to induce dendritic
actin polymerization (6). However, this inter-
action may not account for Cdc42’s ability to
induce filopodia, because cells lacking WASPs
are still able to form filopodia (11); and, as
described above, filopodia contain parallel actin
filaments and not a dendritic network. RhoG
does not interact directly with WASPs but ap-
pears to act upstream of Rac by binding to and
activating a Rac-GEF complex (12).
WAVE/WASP proteins may themselves
regulate the activity of Rac and Cdc42 by
binding to GAPs and GEFs (13–15), and
could thereby generate positive or negative
feedback loops to regulate the extent of
Cdc42/Rac-induced actin polymerization.
WAVE/WASP proteins can also be regulated
by other stimuli apart from Cdc42 and Rac,
including Src family kinases, the adaptor pro-
teins Nck and WIP, and phosphoinositides (7,
14, 16–19).
Polarizing the Cell: A Keystone of
Migration
For a cell to migrate, it must be polarized, which
means that the molecular processes at the front
and the back of a moving cell are different.
Establishing and maintaining cell polarity in re-
sponse to extracellular stimuli appear to be me-
diated by a set of interlinked positive feedback
loops involving Rho family GTPases, phospho-
inositide 3-kinases (PI3Ks), integrins, microtu-
bules, and vesicular transport (Figs. 1 and 2).
Although the following discussion synthesizes
information from multiple cell systems, the rel-
ative contributions of the various signals depend
on the cell type and the specific stimulus.
Cdc42: A master regulator of cell polar-
ity. Cdc42 is a master regulator of cell polar-
ity in eukaryotic organisms ranging from
yeast to humans. Cdc42 is active toward the
front of migrating cells (20), and both inhi-
bition and global activation of Cdc42 can
disrupt the directionality of migration (9).
One way in which Cdc42 influences polarity
is by restricting where lamellipodia form (21)
(see below). Cdc42 can also affect polarity by
localizing the microtubule-organizing center
(MTOC) and Golgi apparatus in front of the
nucleus, oriented toward the leading edge.
Cdc42-induced MTOC orientation may con-
tribute to polarized migration by facilitating
microtubule growth into the lamella and
microtubule-mediated delivery of Golgi-
derived vesicles to the leading edge, provid-
ing membrane and associated proteins needed
for forward protrusion (9, 22). Reorganiza-
tion of the MTOC appears to be more
important for the migration of slow-moving
cells, because in fast-moving cells such as
neutrophils and T cells, it is usually located
behind the nucleus (23).
The effects of Cdc42 on MTOC position
appear to be exerted mainly through a path-
way involving the Cdc42 effector PAR6,
which exists in a complex with PAR3 and an
atypical protein kinase C (aPKC) (24). The
molecular mechanism by which the PAR6/
PAR3/aPKC complex orients the MTOC is
incompletely understood, but recent evidence
suggests that it could occur as a result of local
capture of microtubules at the leading edge
via APC, a protein that binds tubulin and
localizes to the ends of microtubules (9), via
CLIP170 and IQGAP (22) and/or via the
microtubule-based dynein/dynactin motor
protein complex (24).
A downstream target of Cdc42, the kinase
PAK1, can itself mediate Cdc42 activation
downstream of heterotrimeric GTP-binding pro-
tein (G protein)–coupled receptors, which are
activated by many chemoattractants. These in-
teractions define a positive feedback loop be-
tween Cdc42 and PAK1, resulting in high Cdc42
activity at the leading edge (25). Other feedback
loops involving integrins may also contribute to
maintaining local Cdc42 activation (9, 26).
PI3Ks and PTEN: The gradient amplifiers. A
surprising aspect of chemotaxis is the ability of
cells to respond directionally to very shallow
chemoattractant gradients (less than a 10% dif-
ference in the concentration of chemoattractant
between the front and rear of a cell) (27). Such a
small difference in signaling between the front
and rear needs to be amplified into steeper
intracellular signaling gradients in order to gen-
erate a cellular response. The phosphoinosi-
tides PtdIns(3,4,5)P
3
(PIP
3
) and PtdIns(3,4)P
2
[PI(3,4)P
2
] are key signaling molecules that be-
come rapidly and highly polarized in cells that
are exposed to a gradient of chemoattractant
(Fig. 2). This amplification process involves
both localized accumulation and activation of
PI3Ks, which generate PIP
3
/PI(3,4)P
2
, and the
phosphatase PTEN, which removes them. In
Dictyostelium, for example, PI3Ks rapidly accu-
mulate at the leading edge of cells in response to
a chemoattractant, whereas PTEN becomes re-
stricted to the sides and the rear (27, 28) (Fig. 1).
Cells with altered PI3K or PTEN activity can
usually migrate but exhibit a significantly re-
duced ability to move directionally up a che-
moattractant gradient. Although it is not yet clear
what regulates the localization of PI3Ks, Cdc42
activation is implicated in PTEN exclusion from
protrusions in leukocytes, and PIP
3
appears to be
required for localizing Cdc42 activity (25).
These results imply that there is a network of
positive feedback loops between Cdc42, PI3K
products, and PTEN that work together to initi-
ate and maintain the polarity of migrating cells,
although a Cdc42 paralog has not yet been iden-
tified in Dictyostelium.
Localized Rac activation: Initiating and
maintaining protrusion. How do Cdc42 and
PI3Ks lead to activation of the actin polymeriza-
tion machinery required for active protrusion?
The key event appears to be defining where Rac
is active (Fig. 1). This is probably achieved by
activating or delivering a Rac exchange factor
locally, and indeed several Rac GEFs are acti-
vated by PI3K products (29). Once Rac is active,
several feedback loops have been identified that
help maintain directional protrusion. First, Rac
can itself stimulate the recruitment and/or acti-
vation of PI3Ks at the plasma membrane, which
then act upstream of Rac by PIP
3
-sensitive Rac
GEFs (21, 29). Second, microtubules and Rac
may form a positive feedback loop in which
microtubule polymerization activates Rac, and
Rac in turn stabilizes microtubules (22). Third,
integrin engagement leads to Rac activation and
membrane targeting (30), and so new adhesions
formed at the leading edge will stimulate Rac,
which in turn induces recruitment and clustering
of activated integrins to the edge of lamellipodia
(31, 32). PIP
3
also contributes to integrin activa-
tion (33) and may thereby further enhance the
positive feedback to Rac.
Defining the tail. Is restriction of PIP
3
, ac-
tive Cdc42, and Rac to the front of the cell
sufficient to make the back of the cell follow
the front? In several cell types, inhibition of
Rho leads to the formation of an extended tail,
possibly because actomyosin-based contractili-
ty in the body of the cell is decreased. Rho may
R EVIEWS
5 DECEMBER 2003 VOL 302 SCIENCE www.sciencemag.org1706
on May 2, 2018 http://science.sciencemag.org/Downloaded from
also act in the tail by stabilizing microtubules,
which would then promote focal adhesion turn-
over (see below) (22, 34). One model for how
migrating cells maintain polarity is based on the
fact that Rho and Rac are mutually antagonistic,
each suppressing the other’s activity (35). Ac-
tive Rac at the leading edge of cells would
suppress Rho activity, whereas Rho would be
more active at the sides and rear of the cell and
suppress Rac activity, thereby preventing Rac-
mediated protrusion at sites other than the lead-
ing edge (36, 37). However, active Rac has
been implicated in detachment at the rear of
migrating cells (38), and also Rho can lead to
Rac activation (39).
Integrins and Adhesion in Migration
For migration to occur, a protrusion must
form and then stabilize by attaching to the
surroundings. Although many different re-
ceptors are involved in the migration of
different cell types, the integrins are a ma-
jor family of migration-promoting recep-
tors. These receptors act as the “feet” of a
migrating cell by supporting adhesion to
the ECM or other cells and by linking via
adapters with actin filaments on the inside
of the cell. As described above, integrins
activate migration-related signaling mole-
cules. They are also recipients of “inside-
Fig. 2. Steps in cell migration. Polarity is intrinsic to a migrating cell (A).
Cdc42, along with Par proteins and aPKC, are involved in the generation
of polarity. Several additional proteins are implicated in polarity, which
results in directed vesicle trafficking toward the leading edge, organiza-
tion of microtubules (in some cells), and the localization of the MTOC (in
some cells) and Golgi apparatus in front of the nucleus. In the presence
of a chemotactic agent, PIP
3
is produced at the leading edge through the
localized action of PI3K, which resides at the leading edge, and PTEN, a
PIP
3
phosphatase that resides at the cell margins and rear. PTEN and
myosin II are implicated in restricting protrusions to the cell front. The
migration cycle begins with the formation of a protrusion (B). WASP/
WAVE proteins are targets of Rac and Cdc42 and other signaling path-
ways and regulate the formation of actin branches on existing actin
filaments by their action on the Arp2/3 complex. Actin polymerization, in
turn, is regulated by proteins that control the availability of activated
actin monomers (profilin) and debranching and depolymerizing proteins
(ADF/cofilin), as well as capping and severing proteins. Protrusions are
stabilized by the formation of adhesions. This process requires integrin
activation, clustering, and the recruitment of structural and signaling
components to nascent adhesions. Integrins are activated by talin bind-
ing and through PKC-, Rap1-, and PI3K-mediated pathways. Integrin
clustering results from binding to multivalent ligands and is regulated by
Rac. At the cell rear, adhesions disassemble as the rear retracts (C). This
process is mediated by several possibly related signaling pathways that
include Src/FAK/ERK, Rho, myosin II, calcium, calcineurin, calpain, and the
delivery of components by microtubules. Many of these molecules may
also regulate the disassembly of adhesions at the cell front, behind the
leading edge.
R EVIEWS
www.sciencemag.org SCIENCE VOL 302 5 DECEMBER 2003 1707
on May 2, 2018 http://science.sciencemag.org/Downloaded from
out signaling”; that is, activation to a high-
affinity state by cytoplasmic signals (40).
The integrins are heterodimeric recep-
tors consisting of ␣ and  chains with large
ligand-binding extracellular domains and
short cytoplasmic domains. The binding of
ligands to the extracellular portion of inte-
grins leads to conformational changes in
the receptors by changing interactions be-
tween the ␣- and -chain cytoplasmic do-
mains (41) and to integrin clustering. This
combination of occupancy and clustering
initiates intracellular signals such as pro-
tein tyrosine phosphorylation, activation of
small GTPases, and changes in phospholip-
id biosynthesis that regulate the formation
and strengthening of adhesion sites, the
organization and dynamics of the cytoskel-
eton, and cell polarity during migration
(40). Although integrins themselves do not
have any catalytic activity, signals are trans-
mitted through direct and indirect interactions
with many partners of integrins.
Activated integrins preferentially local-
ize to the leading edge, where new adhe-
sions form (31). Integrin affinity is regulat-
ed in large part by alterations in the con-
formation of the integrin extracellular do-
mains that result from interactions at the
integrin cytoplasmic tail (42). Activation of
key intermediates such as the GTPase Rap1
or PKC increase integrin affinity. Con-
versely, activation of Raf-1 kinase often
suppresses integrin activation (43). The cy-
toskeletal linker protein talin promotes in-
tegrin activation by binding to a subset of
integrin -subunit tails and disrupting inte-
grin ␣-–subunit tail interactions (42, 44).
The signaling potential of integrins can
also be modified by posttranslational mod-
ifications of the cytoplasmic domains. For
example, integrin ␣4 phosphorylation on
serine blocks the binding of paxillin, a
signaling adapter protein. In migrating
cells, ␣4 phosphorylation at the leading
edge (Fig. 1) and the consequent release of
bound paxillin are required to maintain sta-
ble lamellipodia of cells migrating on li-
gands for integrin ␣41(45).
Formation of adhesions. The mecha-
nism by which adhesions assemble in mi-
grating cells is a major challenge that is
only beginning to be addressed. Presum-
ably it begins with small-scale clustering
due to the multivalent nature of the ECM to
which the cell is adhering. Some cells,
particularly rapidly migrating ones such as
leukocytes, have few visible integrin clus-
ters, and thus very small submicroscopic
adhesions are probably important for their
migration. In other cells, small adhesions
known as focal complexes can be observed
at the leading edge. Formation of these
adhesions depends on Rac and Cdc42, and
these adhesions stabilize the lamellipodium
by mediating attachment to the ECM,
thereby contributing to efficient migration.
However, cells with large integrin clusters
(“focal adhesions”) are tightly adherent and
are typically either nonmigratory or move
very slowly. The assembly of focal adhe-
sions involves Rho as well as myosin-in-
duced contractility.
During their formation, some protein
components enter adhesions with similar
kinetics, which suggests that they exist in
preformed cytoplasmic complexes (46).
However, other components enter adhe-
sions with very distinct kinetics, which is
consistent with a model in which a regula-
tory event initiates the serial addition of
different proteins. Paxillin, for example, is
present in nascent adhesions, whereas ␣-ac-
tinin appears more prominently in “older”
adhesions (46).
Tractional forces. By connecting the
ECM to the intracellular cytoskeleton, in-
tegrins serve as both traction sites over
which the cell moves and as mechanosen-
sors, transmitting information about the
physical state of the ECM into the cell and
altering cytoskeletal dynamics (3, 47, 48).
Because migrating cells must be able to
detach, yet exert traction on the substratum,
migration speed is a biphasic function of
the strength of cell attachment. The latter is
determined by the density of adhesive li-
gands on the substrate, the density of adhe-
sion receptors on the cells, and the affinity
of the receptors for the adhesive ligands
(3). Thus, shifts in any of these parameters
can have a dramatic effect on migration.
The force transmitted to sites of adhe-
sion derives from the interaction of myosin
II with actin filaments that attach to these
sites. Myosin II activity is regulated by
myosin light-chain (MLC) phosphoryl-
ation, which is either directly positively
regulated by MLC kinase (MLCK) or Rho
kinase (ROCK) or negatively regulated by
MLC phosphatase, which is itself phospho-
rylated and inhibited by ROCK. Whereas
MLCK is regulated by intracellular calcium
concentration as well as by phosphoryl-
ation by a number of kinases, ROCK is
regulated by binding Rho-GTP (49). MLC
phosphorylation activates myosin, resulting
in increased contractility and transmission
of tension to sites of adhesion.
In migrating cells, the strongest forces
have been reported to be transmitted to the
focal complexes at the leading edge and the
retracting regions at the rear (47). In con-
trast, in more adhesive cells, force trans-
mitted through a focal adhesion to the sub-
stratum is proportional to the adhesion’s
cross-sectional area (50). It is striking that
the tractional forces measured in many
studies far exceed what should be needed
for cell translocation. One explanation is
that cells in tissue culture may be respond-
ing to a “wound” environment, which acti-
vates Rho and thus stimulates contractility.
Because traction forces are unevenly dis-
tributed over migrating cells, integrin sig-
naling is a means of reporting these force
differences to the cell.
Adhesion disassembly at the front. Ad-
hesion disassembly is observed both at the
leading edge, where it accompanies the for-
mation of new protrusions, and at the cell
rear, where it promotes tail retraction. At
the front of migrating cells, adhesions at
the base of a protrusion disassemble as new
adhesions form at the leading edge (46).
However, some adhesions persist and ma-
ture into larger, more stable structures. Lit-
tle is known about adhesion disassembly
versus maturation; however, targeting of
microtubules has been implicated as one
factor that promotes adhesion disassembly
(34). Both protein kinases and phospha-
tases also appear to be central to the regu-
lation of adhesion turnover and stability
(51). For example, cells lacking the ty-
rosine kinases FAK or Src have more and
larger adhesions and migrate poorly (46,
52). The interaction of FAK with Src and
the adapter proteins Cas and Crk, which in
turn activate Rac-specific GEFs, appears to
regulate adhesion turnover. Adhesion turn-
over in migrating cells is also regulated by
a complex of Rac-associated proteins (53)
and by the mitogen-activated protein kinase
ERK (54). The emerging evidence favors a
model for adhesion turnover in which acti-
vation of the protein tyrosine kinases FAK
and Src accompanies the formation of an
adhesion signaling complex that in turn
mediates the localized activation of Rac
and ERK. These signals then contribute to
the turnover of adhesions at the leading edge.
Adhesion disassembly and retraction at
the rear. At the rear of migrating cells,
adhesions must also disassemble. In fibro-
blasts, the rearmost adhesions often tether
the cell strongly to the substratum, result-
ing in a long tail to the site of anchorage.
The tension can be sufficient to physically
break the linkage between integrin and the
actin cytoskeleton, with the result that in-
tegrin is left behind while the rest of the
cell moves on; a similar behavior has been
observed in vivo (3). High tension exerted
on the rear adhesions contributes to detach-
ment (3). Several lines of evidence point to
a role for myosin II in this event as well as
in the maintenance of polarity. Dictyosteli-
um cells deficient in myosin II or its regu-
lator PAKa show impaired retraction and
the formation of multiple pseudopodia
along the sides of the cell (55). A similar
phenotype is seen in monocytes or neutrophils
in which myosin II assembly is blocked through
inhibition of Rho or Rho kinase (36, 37). Al-
R EVIEWS
5 DECEMBER 2003 VOL 302 SCIENCE www.sciencemag.org1708
on May 2, 2018 http://science.sciencemag.org/Downloaded from