Cell 123, December 2, 2005 ©2005 Elsevier Inc. 769
Teaching 120 undergraduates the essentials of cellular and
developmental biology in one semester is a daunting pros-
pect. In one lecture, we cover all aspects of cell-cell and
cell-matrix adhesion, so it’s quite a roller-coaster ride. How-
ever, survey courses like this do reveal the essential facts
of the discipline. One of these is that cadherin-based cell-
cell adherens junctions mediate adhesion and are respon-
sible for anchoring the actin cytoskeleton. If you pick up
any undergraduate cell biology textbook, you can see “how
things work” (Figure 1): cadherins mediate homophilic adhe-
sion and the cytoplasmic tails of cadherin cis-dimers bind to
intracellular proteins; β-catenin and the related protein p120
bind to the tail directly at different sites, and α-catenin then
binds to β-catenin; and finally, actin filaments bind directly
to α-catenin. In this issue of Cell, work from the Weis and
Nelson groups shakes up this traditional image of the adhe-
rens junction (Drees et al., 2005; Yamada et al., 2005). They
show that α-catenin, rather than being a stable link to actin,
may instead act as a key regulator of actin dynamics.
How did this static view become accepted as “the way
junctions work”? Adherens junctions (AJs) were first iden-
tified as electron-dense structures near the apical end of
the lateral cell interface in epithelial cells. Beneath them, the
actin cytoskeleton is organized into a belt of bundled actin
filaments that runs around the apical end of the cell (Figure
1A). The molecular machinery that comprises AJs has been
identified over the past 20 years. Cadherins were estab-
lished as homophilic adhesion molecules in both cultured
cells and early mouse embryos. Subsequently, the catenins
were identified as proteins that coimmunoprecipitate with
cadherins. The catenins are a set of both related and unre-
lated proteins. Of these, β-catenin is a member of the Arma-
dillo-repeat superfamily. It binds to a conserved sequence in
the distal part of the cadherin tail. p120 is distantly related to
β-catenin and binds to the juxtamembrane region of cadher-
ins. α-catenin is a distant relative of vinculin, an actin binding
protein. Biochemical data suggest that E-cadherin and β-
and α-catenin form a complex that is roughly stoichiometric
and can be isolated from cells even under relatively harsh
conditions (Ozawa and Kemler, 1992), although, under other
conditions, α-catenin can be preferentially dislodged from
Genetic data support central roles for fruit-fly DE-cad-
herin, mouse E-cadherin, and fly β-catenin (Armadillo) in
cell adhesion and the architecture of epithelial tissues. For
example, mice lacking either E-cadherin or α-E-catenin (the
family members expressed in epithelia) have very similar
phenotypes (reviewed in Jamora and Fuchs, 2002). Both
mutations disrupt the trophectodermal epithelium of the
early embryo prior to implantation.
Can 1000 Reviews Be Wrong? Actin,
a-Catenin, and Adherens Junctions
Julie Gates1 and Mark Peifer1,2,*
1Lineberger Comprehensive Cancer Center
2Department of Biology
University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA
Coupling between cell adhesion and the actin cytoskeleton is thought to require a stable
link between the cadherin-catenin complex and actin that is mediated by a-catenin. In this
issue of Cell, the Weis and Nelson groups call this static model into question, showing that
a-catenin can directly regulate actin dynamics (Drees et al., 2005; Yamada et al., 2005).
Figure 1. A Textbook Model for How the
Adherens Junction Complex Connects
(A) Epithelial cells are joined by adherens junctions,
positioned near the apical end of the lateral cell
interface. Belts of actin filaments underlie adher-
(B) Cadherins mediate homophilic adhesion. The
cytoplasmic tails of cadherin cis-dimers bind to
intracellular proteins. β-catenin binds to the tail
directly, and α-catenin then binds to β-catenin.
Actin filaments then bind to α-catenin. In this
model, cadherins are linked directly to actin via the
catenins. The catenins could also mediate interac-
tions to actin via binding to proteins such as ZO-1,
spectrin, vinculin, afadin, and α-actinin.
770 Cell 123, December 2, 2005 ©2005 Elsevier Inc.
AJs and the underlying actin belt are interdependent. Dis-
ruption of fruit-fly β-catenin leads to widespread defects in
the polarity of the actin cytoskeleton (Cox et al., 1996), sug-
gesting that AJs play a key role in maintaining this actin belt.
Likewise, disruption of the cortical-actin cytoskeleton dis-
rupts AJs (Quinlan and Hyatt, 1999). The current view that
this interdependence relies on a direct interaction emerged
from a series of biochemical experiments. The three regions
of vinculin homology in α-catenin are referred to as the VH1,
VH2, and VH3 domains. Biochemical data have shown
that the VH1 domain binds to β-catenin, whereas the VH3
domain binds to actin (Rimm et al., 1995). α-catenin could
also interact with actin via a dazzling array of binding part-
ners, including α-actinin, vinculin, spectrin, zonula occlu-
dins-1 (ZO-1), and afadin (Figure 1B). Although the cadherin-
catenin complex bound to actin has never been isolated,
these biochemical and genetic data seem to provide ample
support for the textbook view that the catenins provide a
stable tether between cadherin and the actin cytoskeleton.
The Weis and Nelson groups (Drees et al., 2005; Yamada
et al., 2005) bring complementary approaches to the sub-
ject. The Nelson lab was one of the first to establish the role
of cadherins in the polarity of epithelial cells and has contrib-
uted key information about the assembly and disassembly
of AJs. The Weis lab solved the structures of domains of
many proteins involved in cell adhesion, including those in
β- and α-catenin. Their work has led to a new view of the
relationship between cadherins, catenins, and actin.
The first tenet of the current model (Figure 1B) tested by
Weis and Nelson is the idea that the actin cytoskeleton is
physically and stably linked to the cadherin-catenin complex
(Yamada et al., 2005). As they point out, the current model is
based on the “commutative property” applied to biochem-
istry—if E-cadherin and β-catenin bind to α-catenin and α-
catenin binds to actin, then E-cadherin must bind to actin. In
a series of direct tests, they show that this is not accurate.
They offer compelling data suggesting that, although both
binary interactions can be observed, α-catenin assembled
into the cadherin-catenin complex does not bind to actin.
This is demonstrated both in biochemical assays and in a
new assay where cadherin-catenin complexes are assem-
bled onto isolated patches of plasma membrane. Further-
more, neither vinculin nor α-actinin can mediate this inter-
action. Finally, using FRAP (fluorescence recovery after
photobleaching), they show that cortical actin is much more
dynamic than the cadherin-catenin complex, a finding con-
trary to the existence of a highly stable connection (Figure
If α-catenin does not provide a stable connection
between AJs and actin, then what does? One possibil-
ity is that there is no direct or indirect connection between
actin and AJs, either stable or transient. However, several
observations make this unlikely. Although the circumfer-
ential belt of actin underlying AJs could be maintained by
proteins whose concentration is locally raised by transient
association with the AJ complex, this would not explain why
altering the actin cytoskeleton destabilizes AJs. Likewise, if
the AJ is not tethered to actin, how does contraction of the
actin belt during apical constriction trigger cell constriction
during embryonic development (for example, during neu-
ral-tube formation and mesoderm invagination)? In the fruit
fly Drosophila, Wieschaus and colleagues recently showed
that AJs are repositioned in response to actin-based apical
constriction during mesoderm invagination, and they dem-
onstrate that reducing AJ function by depleting β-catenin
allows the actin-myosin ring to contract without changing
a cell’s shape, strongly supporting the existence of some
kind of physical link between actin and AJs (Dawes-Hoang
et al., 2005).
A second possibility is that other molecules mediate a
direct connection. One attractive candidate is the nectin-
afadin system (Figure 2B; reviewed in Takai and Nakanishi,
2003). Nectins are immunoglobulin-superfamily adhesion
molecules that also are localized to AJs and bind to afadin.
Afadin is a cytoplasmic plaque protein that binds to actin as
well as to several actin-associated proteins and thus could
serve as an alternate link. It is also possible that actin bind-
ing proteins other than those tested by the Weis and Nelson
groups may be recruited to the cadherin-catenin complex.
However, the FRAP data would suggest that none of these
interactions is stable.
A third possibility is that the link is mediated by many
weak and transient interactions (Figure 2C), which cannot
be detected biochemically. Interestingly, a similar model
has been proposed for the function of the cadherin extra-
cellular domain. Cadherins interact homophilically with the
cadherins of neighboring cells, mediating cell-cell adhesion.
The current model suggests that the individual cadherin-
cadherin bond is relatively weak. However, given that many
cadherins are clustered into AJs, the sum of these weak,
transient interactions creates an adhesive interface that can
be strong yet easily remodeled.
Without a stable linkage between AJs and actin, one is
still left with the fact that all three core components of AJs
are essential for adhesion and for the polarization of the actin
cytoskeleton in epithelial cells. Moreover, site-specific muta-
tions in the α-catenin binding site of fly β-catenin disrupt
adhesion and polarity (Orsulic and Peifer, 1996), suggest-
ing that recruitment of α-catenin into the cadherin-catenin
complex is also essential. If α-catenin does not provide a
stable link to AJs, what role does it fulfill? Nelson, Weis, and
their colleagues suggest that α-catenin acts as a molecular
switch that regulates actin dynamics at AJs.
How might α-catenin act as a molecular switch? α-catenin
can exist as a monomer, a homodimer, and in heterodimers
with β-catenin. Both homodimerization and interaction
with β-catenin occur via its N-terminal VH1 domain. Thus,
homodimerization and heterodimerization compete with
one another. In contrast, the interaction with actin involves
the C-terminal VH3 domain. In a simple world, this would
mean that actin binding and AJ association would be inde-
pendent. However, the Nelson group (Yamada et al., 2005)
now demonstrates that α-catenin cannot simultaneously
bind to β-catenin and actin. How can this be? In their study,
the Weis lab (Drees et al., 2005) provides evidence that α-
catenin behaves in an allosteric fashion, with its affinity for
Cell 123, December 2, 2005 ©2005 Elsevier Inc. 771
one ligand affected by binding to another. The α-catenin-β-
catenin heterodimer has a high affinity for cadherin but has
reduced affinity for actin, whereas dimeric α-catenin cannot
bind to the cadherin-catenin complex but has a high affinity
for actin (Figure 2A). This suggests that interactions at the
N terminus somehow modulate the actin affinity of the C
terminus of α-catenin, thereby allowing α-catenin to switch
between different states.
Drees et al. (2005) next consider what roles α-catenin
might have when it is not in the cadherin-catenin complex.
Because α-catenin homodimers bind to actin, the authors
speculate that α-catenin may influence actin dynamics
at AJs. To test this possibility, they examined whether α-
catenin can influence actin polymerization mediated by the
Arp2/3 complex (containing actin-related proteins 2 and 3).
The Arp2/3 complex nucleates actin polymerization from the
sides of existing filaments, which creates a branched, den-
dritic actin array. Using a pyrene-actin fluorescence assay,
they found that the actin polymerization that is normally
observed in the presence of the Arp2/3 complex and the
activation domain of the Wiskott-Aldrich syndrome protein
(WASP) was suppressed by the addition of the α-catenin
homodimer. The degree of suppression is concentration
dependent and correlates with the binding of the α-catenin
homodimer to actin. Interestingly, the amount of Arp3 bound
to actin decreases when the concentration of the α-catenin
homodimer is increased, suggesting that the α-catenin
homodimer competes with the Arp2/3 complex for binding
to actin filaments. α-catenin might inhibit actin branching,
facilitating formation of the belt of unbranched actin fila-
ments (Figure 2A).
One caveat to this analysis is that the suppression of the
Arp2/3 complex requires very high concentrations of the
α-catenin homodimer, with mild effects observed at 1 µM
and complete suppression requiring 7.5 µM. In contrast,
concentrations only in the nM range of the vasodilator-stim-
ulated phosphoprotein (VASP) are needed to increase the
rate of actin polymerization from spectrin/F-actin seeds in
the presence of capping proteins (Barzik et al., 2005). The
requirement for such a high concentration of the α-catenin
homodimer raises questions about the physiological rele-
vance of this result. However, Drees et al. (2005) found that
α-catenin bound to E-cadherin-β-catenin complexes can
leave this complex and bind to actin in solution. Therefore,
they propose that E-cadherin and β-catenin recruit α-catenin
to AJs. However, because this interaction is transient, the
subsequent dissociation of α-catenin from this complex
leads to a local increase in the concentration of α-catenin
in the vicinity of apical actin. Whether this is sufficient to cre-
ate the concentration needed to allow α-catenin to influence
Arp2/3-mediated actin polymerization at AJs remains to be
determined. A recent study examining the local concentra-
tion of several actin regulators in fission yeast shows that
techniques to address this are now available (Wu and Pol-
However, α-catenin is not the only regulator of actin
dynamics found at the AJ. To date, several actin regulators
have been shown to localize to either established or nascent
AJs (Figure 2C), including Ena/VASP proteins (Vasioukhin
et al., 2001), formin-1 (Kobielak et al., 2004), the Arp2/3
complex (Kovacs et al., 2002), and its activator cortactin
(Helwani et al., 2004). The Arp2/3 complex and formin-1
can each nucleate actin filaments, although the geometry
of the resulting filaments differs. Formin-1 nucleates polym-
erization of linear actin filaments, whereas the Arp2/3 com-
plex nucleates polymerization of branched actin filaments.
Ena/VASP proteins promote the continued elongation of
existing filaments by binding to the quickly growing barbed
end and preventing the binding of capping proteins (Bar-
zik et al., 2005). The localization of both Ena/VASP proteins
and formin-1 to AJs is dependent on α-catenin (Kobielak
et al., 2004; Vasioukhin et al., 2000). Furthermore, formin-1
can directly interact with α-catenin (Kobielak et al., 2004),
although it is not known whether α-catenin can bind to both
β-catenin and formin-1 at the same time. Interestingly, both
cortactin (Helwani et al., 2004) and a component of the
Arp2/3 complex, p34 (Kovacs et al., 2002), coimmunopre-
cipitate with E-cadherin, suggesting that their AJ localiza-
tion may depend on a direct or indirect association with the
cadherin-catenin complex. These actin regulatory proteins
Figure 2. Revised Models for How the
Adherens Junction Complex Connects
(A) The model proposed by Weis and Nelson: an
α-catenin monomer binds to β-catenin; α-catenin
homodimers, released from cadherin-catenin
complexes, bind to actin and antagonize Arp2/3
(B and C) Alternative models explain the associa-
tion between actin and adherens junctions.
(B) A direct connection to actin is mediated by
other junction proteins, such as nectin and afadin.
(C) The connection is mediated by the cumula-
tive effect of several weak transient interactions
between actin binding proteins (such as ZO-1,
spectrin, vinculin, afadin, and α-actinin) and adhe-
rens junction components, facilitated by cadherin
772 Cell 123, December 2, 2005 ©2005 Elsevier Inc.
may provide a molecular toolkit that can dynamically alter
the state of actin at AJs. Dynamic changes in actin occur as
AJs are assembled, suggesting that this process is highly
regulated. In order to enable diverse responses, perhaps
cells can regulate the release of distinct subsets of the actin
regulators that are localized to AJs depending on the situa-
tion. For example, formins may be involved in the formation
of linear actin cables at nascent junctions. These additional
players might also help account for the very rapid polymer-
ization of actin induced by cadherin clustering, as observed
by several labs. Likewise, regulators of the microtubule cyto-
skeleton might localize at AJs, helping to explain the effects
of cell adhesion on microtubules.
The Weis and Nelson model suggests that α-catenin
continuously shuttles between an “inactive” cadherin bound
pool and an actin bound pool that is active in regulation of
the cytoskeleton (Figure 2A). One challenge to this model
comes from the analysis of fusion proteins that covalently link
E-cadherin and α-catenin, eliminating the need for β-catenin
as a bridge. Fusion proteins that join E-cadherin (minus its
β-catenin binding site) to either full-length α-catenin or the
α-catenin VH3 domain each rescue cell adhesion (Nagafu-
chi et al., 1994). Furthermore, they also confer resistance
to extraction by nonionic detergents, a common assay for
cytoskeletal association, and disrupting the actin cytoskel-
eton interferes with the strong adhesion mediated by these
fusion proteins. These data suggest that if α-catenin acts as
an allosteric regulator of the cytoskeleton, it may be able to
do so when recruited to AJs by covalent linkage to E-cad-
herin. This could relieve the allosteric repression conferred
by β-catenin binding. However, this work was done in the
presence of wild-type α-catenin. Wild-type α-catenin might
also be recruited to AJs by homodimerization with the α-
catenin-E-cadherin fusion protein and may be able to regu-
late actin. Although E-cadherin-α-catenin fusion proteins
rescued adhesion, dividing cells did not round up, and cell
motility within an epithelial sheet was impaired (Nagafuchi et
al., 1994), suggesting that AJs built around this fusion pro-
tein may not allow for rapid remodeling.
Recent work extends these findings in vivo. Pacquelet and
Rorth (2005) examined the ability of DE-cadherin-α-catenin
fusion proteins to carry out an array of functions during
Drosophila oogenesis. Fusions similar to those studied by
Nagafuchi et al. (1994) fully rescue the adhesive functions for
DE-cadherin in both the germline and somatic follicle cells.
However, these fusion proteins do not rescue DE-cadherin-
dependent migration of one group of somatic follicle cells,
the border cells (Niewiadomska et al., 1999). But this func-
tion is rescued if the juxtamembrane region of DE-cadherin
is included in the fusion protein. This region binds to p120,
a distant relative of β-catenin. Thus, in vivo, DE-cadherin-α-
catenin fusion proteins can restore virtually all of the known
functions of DE-cadherin. This may limit the generality of the
allosteric regulation model but is also subject to the caveat
that these experiments were conducted in the presence of
wild-type α-catenin. It will be of interest to test the function of
these fusion proteins in flies carrying mutations in α-catenin,
when such mutants become available.
In addition to regulating the actin cytoskeleton, α-catenin
may have other unexpected roles in the biology of epithe-
lial cells. One striking example is provided by work from
the Fuchs group, in which α-E-catenin was knocked out in
the skin cells of mice (Vasioukhin et al., 2001). This led to
expected defects in cellular junctions and severe defects in
the organization of the skin. However, there were some sur-
prising consequences: skin cells underwent hyperprolifera-
tion both in vivo and in vitro, and the MAP kinase pathway
was constitutively activated. Perhaps, like its binding partner
β-catenin, α-catenin has a dual signaling function, with per-
haps some role in the nucleus.
It is encouraging to those of us still engaged in studying
adhesion that 20 years of research on its molecular mecha-
nisms raises many exciting new questions, just as old ques-
tions have been answered. We look forward to seeing what
the textbooks will say about AJs in 20 years’ time.
Barzik, M., Kotova, T.I., Higgs, H.N., Hazelwood, L., Hanein, D., Gertler,
F.B., and Schafer, D.A. (2005). J. Biol. Chem. 280, 28653–28662.
Cox, R.T., Kirkpatrick, C., and Peifer, M. (1996). J. Cell Biol. 134, 133–
Dawes-Hoang, R.E., Parmar, K.M., Christiansen, A.E., Phelps, C.B.,
Brand, A.H., and Wieschaus, E.F. (2005). Development 132, 4165–
Drees, F., Pokutta, S., Yamada, S., Nelson, W.J., and Weis, W.I. (2005).
Cell, this issue.
Helwani, F.M., Kovacs, E.M., Paterson, A.D., Verma, S., Ali, R.G., Fan-
ning, A.S., Weed, S.A., and Yap, A.S. (2004). J. Cell Biol. 164, 899–910.
Jamora, C., and Fuchs, E. (2002). Nat. Cell Biol. 4, E101–E108.
Kobielak, A., Pasolli, H.A., and Fuchs, E. (2004). Nat. Cell Biol. 6, 21–30.
Kovacs, E.M., Goodwin, M., Ali, R.G., Paterson, A.D., and Yap, A.S.
(2002). Curr. Biol. 12, 379–382.
Nagafuchi, A., Ishihara, S., and Tsukita, S. (1994). J. Cell Biol. 127, 235–
Niewiadomska, P., Godt, D., and Tepass, U. (1999). J. Cell Biol. 144,
Orsulic, S., and Peifer, M. (1996). J. Cell Biol. 134, 1283–1301.
Ozawa, M., and Kemler, R. (1992). J. Cell Biol. 116, 989–996.
Pacquelet, A., and Rorth, P. (2005). J. Cell Biol. 170, 803–812.
Quinlan, M.P., and Hyatt, J.L. (1999). Cell Growth Differ. 10, 839–854.
Rimm, D.L., Koslov, E.R., Kebriaei, P., Cianci, C.D., and Morrow, J.S.
(1995). Proc. Natl. Acad. Sci. USA 92, 8813–8817.
Takai, Y., and Nakanishi, H. (2003). J. Cell Sci. 116, 17–27.
Vasioukhin, V., Bauer, C., Yin, M., and Fuchs, E. (2000). Cell 100, 209–
Vasioukhin, V., Bauer, C., Degenstein, L., Wise, B., and Fuchs, E. (2001).
Cell 104, 605–617.
Wu, J.Q., and Pollard, T.D. (2005). Science 310, 310–314.
Yamada, S., Pokutta, S., Drees, F., Weis, W.I., and Nelson, W.J. (2005).
Cell, this issue.