News & Views | Published:

Cell biology

A break in the chain?

How chains of proteins link transmembrane cell–cell adhesion molecules to the cell's inner scaffold was standard textbook material. But recent research challenges the accepted model, opening a new chapter in the field.

Many of the adhesion molecules that enable cells to stick together are anchored to part of the cell's internal scaffold — the actin cytoskeleton. This association is usually indirect and serves to transmit tension to sites of adhesion, to cluster adhesion molecules and to provide a framework for the assembly of complexes of signalling proteins. How the adhesion molecules connect to the cytoskeleton has often been difficult to define, but one of the most widely accepted chains of attachment is that of the cadherin adhesion proteins to actin filaments through the β-catenin and α-catenin proteins, with α-catenin providing the link to actin (Fig. 1). (Indeed, ‘catenin’ derives from the Latin for chain.) Two Cell papers from the groups of W. James Nelson and William Weis1,2 challenge the validity of this linkage, and particularly the role of α-catenin.

Figure 1: The conventional model for how a cadherin links to actin filaments.

The cytoplasmic domain binds to β-catenin (beige), which complexes with α-catenin (green). In turn, α-catenin provides the link to actin, either directly or indirectly through one of several other proteins (?). Work from Nelson, Weis and colleagues1,2 challenges this static model, instead proposing that the binding of α-catenin to β-catenin is incompatible with the simultaneous binding of α-catenin to actin (Fig. 2).

Cadherins mediate cell–cell adhesion in many cell types. Much of what is known about these proteins derives from work on E-cadherin, the type found in the epithelial cells that line ducts and cavities in the body. E-cadherin is concentrated within cell–cell contacts known as adherens junctions, and the extracellular domains of the protein hold adjacent cells together. Adherens junctions associate with a circumferential belt of actin filaments around the cell, and together these structures determine the shape of epithelial cells and maintain epithelial integrity. During tissue development, contraction of the circumferential belt can transmit tension to adherens junctions, causing cells to constrict in their apical regions, inducing the epithelium to curve or fold. This underlies the generation of structures such as the neural tube, which develops from a localized folding of the neural epithelium and eventually goes on to form the brain and spinal cord.

The first indications that cadherins might be linked to actin filaments came from the localization of adherens junctions with bundles of actin filaments. Furthermore, cadherins partitioned with much of the cytoskeleton in crude subcellular fractionation procedures. Then, nearly 20 years ago, cadherins were shown to interact with at least two cytoplasmic proteins, α- and β-catenin. Binding studies demonstrated that β-catenin forms a bridge between cadherin and α-catenin, which in turn binds to actin in vitro. From these binary interactions, a chain of attachment seemed logical (Fig. 1), and this model was accepted by the field (and textbooks) with little challenge, except for the qualification that many other actin-binding proteins also bind to α-catenin3 and might contribute to this linkage.

The Nelson and Weis groups1,2 provide several lines of evidence that are inconsistent with this textbook linear model. They report1 that α-catenin does not bind simultaneously to actin and to β-catenin in complex with the E-cadherin intracellular domain. They developed a system to test the binding of purified components to ‘membrane patches’ enriched for cadherins but stripped of the associated proteins. Although β-catenin bound to these preparations and conferred α-catenin binding, these complexes were unable subsequently to bind to either G- or F-actin.

Recognizing that additional α-catenin-binding components may be required, the authors tested two prime candidates, vinculin and α-actinin, but these revealed little (vinculin) or no (α-actinin) binding to the cadherin–catenin complexes, and they failed to mediate actin binding to the membranes. Cell extracts, which would be expected to contain the putative missing factor, also failed to confer actin binding in this system.

Finally, the authors used FRAP, a technique that analyses the dynamics of fluorescently tagged proteins in cells or subcellular compartments, to monitor the cadherin–catenin complex and actin in adherens junctions. All the components of the cadherin–catenin complex had similar dynamics, and were mostly immobile, presumably held in the complex. By contrast, actin and several actin-binding proteins were far more mobile, arguing against a stable linkage between the cadherin–catenin complex and actin filaments.

So what is α-catenin doing instead of making a static bridge between actin and the catenin–cadherin complex? In their second paper2, Nelson, Weis and their colleagues demonstrated that α-catenin exists in three forms in the cell: singly, paired with itself (a homodimer) or paired with β-catenin. The monomer preferentially binds to the β-catenin–E-cadherin complex, whereas the homodimer binds to filamentous actin (Fig. 2). Significantly, they found that the α-catenin homodimer competes with the Arp2/3 protein complex for binding to actin. When bound to actin filaments, the Arp2/3 complex initiates the assembly of new actin filaments, generating a branched actin network. By preventing binding of the Arp2/3 complex to actin, the α-catenin homodimer inhibits new filament assembly in vitro. This inhibition requires α-catenin concentrations higher than those estimated in cells, but the authors speculate that recruitment of α-catenin to adherens junctions may provide a reservoir for the protein, increasing its local concentration next to junctions. Once there, it may contribute to the conversion of a branched actin network to bundles of parallel filaments. This must occur when adherens junctions are being established in response to cell–cell contact, although other mechanisms may also contribute to this conversion4.

Figure 2: The states of α-catenin.

Nelson, Weis and colleagues1,2 find that in the cytoplasm, α-catenin can exist singly (a monomer), paired with itself (a homodimer), which binds to filamentous actin, or paired with β-catenin (a heterodimer). When complexed with β-catenin, it can associate with the cytoplasmic domain of cadherins but not with actin.

The authors conclude that α-catenin does not link cadherins directly to actin and that the cadherin–catenin complex is not stably associated with actin. Both conclusions are provocative and demand re-evaluation of older work. The first point, for example, is difficult to reconcile with experiments by Nagafuchi, Tsukita and colleagues5,6,7 that study the function of α-catenin by fusing it or its domains directly to E-cadherin. These chimaeric constructs, particularly one that included just the carboxyl terminus of α-catenin (containing the actin-binding domain), bestowed on cells lacking E-cadherin the adhesion properties of epithelial cells. Some of these experiments implied that both E-cadherin and a chimaera containing the actin-binding domain of α-catenin were tethered to the cytoskeleton and could resist significant forces when pulled with optical tweezers7. These experiments, although suggestive, are not definitive, as chimaeras can exhibit unique characteristics not shared by the parent proteins, and the carboxy terminus of α-catenin can bind to other actin-binding proteins3.

The second conclusion seems to fly in the face of evidence that considerable tension can be exerted on adherens junctions. This is seen not only during development, but also in other situations; for instance, when the wound-response factor thrombin increases endothelial permeability. This occurs in part by stimulating cell contraction. The increased tension is transmitted to the adherens junctions, changing the cell borders from a smooth to a jagged contour, with bundles of actin filaments terminating at these sites rather than running parallel to the junctions.

Can relatively dynamic linkages allow the transmission of force? With integrins (another class of adhesion molecule), tension does seem to be conveyed through low-affinity associations made between the cytoplasmic domain and intermediates such as talin. In part, this may be explained by the ‘Velcro’ principle, where many weak interactions generate an adhesion capable of transmitting large forces. Another possibility is that tension itself exposes ‘cryptic’ binding sites on α-catenin or other components, which then enhance the binding. The Nelson and Weis groups demonstrate that α-catenin does undergo conformational changes as a consequence of binding to another molecule. Thus, the interaction of β-catenin with the amino terminus of α-catenin inhibits the binding of the α-catenin carboxy terminus to actin. Could there be conditions, such as mechanical tension, that expose this actin-binding site on α-catenin while it is still bound to β-catenin?

The death knell of a research field occurs when its evolving ideas are captured by textbooks as ‘settled’ dogma, stifling further enquiry. The Nelson and Weis papers1,2 not only identify new functions of α-catenin but, by challenging long-accepted concepts, they should reopen a seemingly settled question and reinvigorate the field.


  1. 1

    Yamada, S., Pokutta, S., Drees, F., Weis, W. I. & Nelson, W. J. Cell 123, 889–901 (2005).

  2. 2

    Drees, F., Pokutta, S., Yamada, S., Nelson, W. J. & Weis, W. I. Cell 123, 903–915 (2005).

  3. 3

    Kobielak, A. & Fuchs, E. Nature Rev. Mol. Cell Biol. 5, 614–625 (2004).

  4. 4

    Kobielak, A., Pasolli, H. A. & Fuchs, E. Nature Cell Biol. 6, 21–30 (2004).

  5. 5

    Nagafuchi, A., Ishihara, S. & Tsukita, S. J. Cell Biol. 127, 235–245 (1994).

  6. 6

    Sako, Y., Nagafuchi, A., Tsukita, S., Takeichi, M. & Kusumi, A. J. Cell Biol. 140, 1227–1240 (1998).

  7. 7

    Imamura, Y., Itoh, M., Maeno, Y., Tsukita, S. & Nagafuchi, A. J. Cell Biol. 144, 1311–1322 (1999).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.