The cytoskeletal protein talin plays a key role in coupling the integrin family of cell adhesion molecules to the actin cytoskeleton. In this paper I present a brief review on talin and summarize our recent studies, in which we have taken both genetic and structural approaches to further elucidate the function of the protein.
- actin cytoskeleton
- vinculin-binding site
There is now extensive evidence that the cytoskeletal protein talin plays a pivotal role in coupling the integrin family of cell adhesion molecules to the actin cytoskeleton . Moreover, talin also activates integrins , probably by relieving the electrostatic interaction between the α- and β-subunit cytoplasmic tails . The structure and biochemical properties of talin make it perfectly suited for such a role. It is an elongated (approx. 60 nm) flexible molecule , reportedly an antiparallel dimer . The globular talin head contains a FERM (four-point-one, ezrin, radixin, moesin) domain (residues 86–400) with binding sites for β-integrin cytodomains , and two signalling proteins, FAK (focal adhesion kinase)  and the type 1γ661 isoform of PIPK (phosphatidyl-inositol-4-phosphate 5-kinase) [8,9], which regulate the dynamic properties of integrin-containing cell–extracellular matrix junctions or FAs (focal adhesions). The FERM domain also contains an actin-binding site , and there is an adjacent membrane insertion sequence  which may account for the ability of talin to create holes in liposomes [12,13]. The talin rod, which is responsible for dimer formation, contains a conserved C-terminal actin-binding site [14,15] homologous to that in Hip1R (Huntingtin-interacting protein 1-related) and yeast Sla2p , a second integrin-binding site [17,18], and several binding sites for the cytoskeletal protein vinculin [19,20], which itself has multiple binding partners, including F-actin . The talin head reportedly binds acidic phospholipids , and PtdIns-(4,5)P2 activates the integrin-binding sites in talin , although both require further investigation. Whether talin is regulated by an intramolecular head–tail interaction that is relieved by PtdIns(4,5)P2 as for other FERM proteins is unclear, although talin can adopt a globular conformation in low-salt buffers in a manner that depends on the talin head .
Integrin signalling via FAK and Src promotes binding of talin to PIPK , although this effect may be indirect. Src phosphorylates Tyr649 in the C-terminal region of the PIPK type 1γ661 isoform, and this suppresses phosphorylation of the adjacent Ser650 by ERK (extracellular-signal-regulated kinase); phosphorylation of Ser650 inhibits the interaction between PIPK and talin . Binding of talin to PIPK activates the kinase, and results in translocation of the complex to the plasma membrane . This suggests a model in which the localized production of PtdIns(4,5)P2 at the plasma membrane activates talin, which then binds to and activates integrins, and in so doing provides a link between integrins and the actin cytoskeleton. Recent data also suggest a role for phospholipase D2 in integrin adhesion, via a mechanism which involves a different isoform of PIPK type 1γ . Recruitment of additional components to the initial integrin–talin–F-actin complexes might then (i) stabilize the complex and facilitate maturation into larger FAs and (ii) regulate the dynamic properties of FAs and therefore cell motility.
There are two talin genes in mammals
The Tln1 and Tln2 genes each encode proteins composed of 2541 amino acids (approx. 74% identity), and it seems probable that they will bind to a similar repertoire of proteins. The genes each contain 57 exons, and intron/exon boundaries are totally conserved, but the Tln2 gene is approx. 190 kb, whereas Tln1 is approx. 30 kb in size . The difference is because of the much larger introns present in Tln2. Northern blots suggest that Tln2 is expressed in a more restricted set of tissues than the ubiquitously expressed Tln1, and Tln2 gene trap mice show high levels of expression in skeletal and cardiac muscle as well as brain (Y. El-Jai, S.J. Monkley, C.A. Pritchard and D.R. Critchley, unpublished work). Interestingly, there are two talin genes in Dictyostelium (TalA and TalB) that play distinct roles in adhesion  and morphogenesis  respectively, although Drosophila  and Caenorhabditis elegans  appear to have only a single talin gene.
Additional roles for talin
Interestingly, new unexpected roles for talin have recently emerged. For example, the talin–PIPK interaction is required for the final stages of clathrin-mediated endocytosis in the synpase . Talin has been reported to co-localize and to co-immunoprecipitate with β1-integrins in Golgi/ER (endoplasmic reticulum) fractions , and HeLa cells transfected with a Tln1 antisense RNA showed defects in integrin processing . Tln1−/− ES (embryonic stem) cells also showed a dramatic reduction in β1-integrin protein , and the possibility that a talin–PIPK plays a role in the formation of integrin-containing Golgi transport vesicles merits investigation. A recent report shows that talin is important in the negative regulation of DE-cadherin transcription in Drosophila ovarian follicle cells , although the mechanism has not been elucidated. It will be important to establish whether talin plays a wider role in regulating cadherin gene expression in other tissues and organisms. Talin has been localized to the mid-body in mammalian cells and is therefore implicated in cytokinesis . Indeed, knock-out of the TalA gene in Dictyostelium resulted in multinucleate cells , and TalA has been shown to be crucial for myosin II-independent and adhesion-dependent cytokinesis . (Interestingly, TalA has recently been shown to bind myosin VII .)
A conditional mouse Tln1 allele for studies of the role of talin1 in development
Our original Tln1 knock-out studies in mouse  were restricted by the fact that Tln1−/− mice die at approx. 8.5 days post-coitus (only slightly later than the β1-integrin knock-out), in part due to failure to complete gastrulation, thought to be one of the earliest integrin-mediated developmental events. Therefore, it was not possible to study the role of talin1 in tissue morphogenesis, or to isolate Tln1−/− MEFs (mouse embryo fibroblasts) for in vitro studies. To address these problems, we have recently generated mice that carry a Tln1 conditional allele (Tln1+/fl, where fl is floxed) in which two loxP sites flank coding exons 1 to 4. We have confirmed that Tln1−/− mice (derived by intercrossing with Cre-recombinase mice) have a phenotype very similar to the conventional Tln1 knock-outs described previously , and mice carrying the conditional Tln1 allele are now available for studies on the role of talin1 in specific tissues during development and in the adult.
We have also successfully derived Tln1−/fl and Tln1+/fl MEFs. Transient co-transfection of Cre-recombinase with an FA marker such as GFP (green fluorescent protein)–paxillin into the Tln1−/fl cells by nucleofection (approx. 70% transfection efficiency) showed a rapid reduction (within 24 h) of GFP–paxillin-positive FAs, reduced spread-cell area and a marked reduction in talin immunoreactivity as judged by Western blotting. In contrast, Tln1+/fl MEFs retained the wild-type phenotype after Cre transfection (S.J. Monkley, C.A. Pritchard and D.R. Critchley, unpublished work). This result is consistent with previous Tln1 knock-out studies in ES cells , and the Tln1−/fl MEFs should be a useful tool for further talin structure–function studies. The immediate questions under investigation are as follows:
(i) Is integrin binding to the talin1 FERM domain essential for FA assembly, since a second integrin-binding site has been mapped to the talin rod  (residues 1984–2113) . Crystallographic data show that the β3-integrin cytodomain docks into the talin FERM F3 subdomain, and that mutation of talin R358A abolishes integrin binding . Expression of a talin R358A mutant in the conditional Tln-1−/fl knock-out cells should establish whether it can rescue the FA defects induced by Cre. However, NMR  and crystallographic  data show that the F3 subdomain also binds PIPK via the same hydrophobic groove, and the R358A mutation again inhibits binding. Therefore, interpretation of the above results may be less clear cut than anticipated, although the talin F3 subdomain/PIPK structure may help to identify mutations that selectively inhibit either integrin or PIPK binding.
(ii) Is the C-terminal actin-binding site in talin1 essential for FA assembly? The C-terminal actin-binding site in talin is reportedly the dominant actin-binding site , so mutations in this site should block the ability of talin to couple integrins to actomyosin. Indeed, optical trap experiments showed that a truncated talin lacking this site was unable to support the formation of the weak linkage between fibronectin-coated beads and the actin cytoskeleton in Tln1−/− fibroblasts derived from ES cells . However, we have recently characterized an actin-binding site in the talin FERM domain , and mutations that selectively inactivate one or other of these sites will be required to establish their role in talin function.
(iii) Do the talin head and rod domains function independently? Calpain 2 (m-calpain) cleavage of talin1 is involved in FA turnover . However, the talin rod has been localized to FAs . Moreover, m-calpain is implicated in activation of the integrin-binding site(s) in talin . Since both the talin FERM and rod domains contain integrin- and actin-binding sites, they might function independently. In activated platelets, the talin head released by calpain cleavage (and not the rod) is selectively retained with integrin αIIbβ3 in Triton X-100 cytoskeletons , suggesting that the talin FERM domain may serve to link integrins to F-actin in platelets.
(iv) Can talin2, α-actinin-1, tensin or filamin rescue the Tln1−/− phenotype? Our ES-derived Tln1−/− cells up-regulate talin2 with passage, and this may explain why they regain their FAs, although filamin is also up-regulated. Therefore it will be important to establish whether the Tln2 cDNA can rescue the Tln1−/− MEF phenotype. In the same way, it will also be important to test whether α-actinin-1, filamin and tensin, all of which bind integrins and F-actin, can compensate for loss of talin1.
The talin–vinculin interaction
One of the most well-characterized talin binding proteins is vinculin, which co-localizes with talin in many different types of integrin-mediated cell junctions, although vinculin also co-localizes with cadherins in cell–cell junctions . Despite the extensive literature on vinculin, the precise function of the protein in FAs is still far from clear . Vinculin null cells are less well spread, are more motile, and have fewer and smaller FAs than wild-type cells . The spreading defect may be owing to the fact that the SH3 (Src homology 3) protein vinexin fails to localize to FAs in the absence of vinculin  (vinexin has previously been implicated in epidermal-growth-factor-mediated cell spreading ). Moreover, vinculin binds the Arp (actin-related protein)2/3 complex, an interaction implicated in lamellipodia formation . (Given this, it is perhaps surprising that vinculin null cells are more motile.) The finding that the recruitment of vinculin to FAs is stimulated by applied force , and that vinculin reduces FA turnover , suggests that vinculin may stabilize the initial integrin–talin–F-actin complexes, perhaps by cross-linking talin to F-actin. However, vinculin appears to play additional roles; e.g. vinculin null cells are more resistant to apoptosis, apparently owing to the fact that FAK/paxillin signalling is constitutive in the absence of vinculin .
The recent crystal structures of vinculin show how an interaction between the vinculin head and tail renders many of the ligand-binding sites cryptic [55,56]. This intramolecular interaction can be relieved by a talin peptide  containing one of the vinculin-binding sites contained within the talin rod, raising the possibility that talin may play a role in vinculin activation. PtdIns(4,5)P2 has also been shown to activate vinculin , although the physiological significance of this has been challenged by the observation that vinculin mutants deficient in PtdIns(4,5)P2 binding still localize to FAs and suppress cell motility . However, FA turnover in cells expressing such mutants is reduced, and vinculin has been suggested to regulate FA dynamics by acting as a PtdIns-(4,5)P2 sensor .
Characterization of the VBSs (vinculin-binding sites) in the talin rod
We previously identified several VBSs in the talin rod using a series of overlapping recombinant talin polypeptides and a vinculin SDS-gel blot assay , and three VBSs were further localized to a stretch of approx. 25 amino acids (each predicted to be a single amphipathic α-helix) using a yeast two-hybrid assay . More recently, we have synthesized peptides representing each of the 63 predicted α-helices in the talin rod and identified an additional 8 VBS sequences, giving a total of 11 VBSs. The new VBSs were probably missed in the original study owing to the fact VBSs tend to be clustered within the talin rod, and so deletion of one VBS would not necessarily abolish vinculin binding to a large talin polypeptide. Given that all of the approx. 63 talin helices are amphipathic in nature, we sought to explain why only approx. 11 helices bind vinculin. Using talin VBS mutant peptides, we have defined a consensus for vinculin binding (LXXAAXXVAXXVXXLIXXA, where X is any amino acid residue). Positions 1, 8, 12, 15 and 16 require an aliphatic residue and will not tolerate alanine, whereas positions 4, 5 and 9 are less restrictive . Thus the key determinants of vinculin binding are hydrophobic residues that are located on one face of an amphipathic α-helix.
The structure of VBSs in talin
To define further the molecular basis of the vinculin–talin interaction, we have determined the structure of several regions of the talin rod which contain VBSs. The crystal structure of the N-terminal part of the rod (residues 482–789) , and the NMR structure of an adjacent region (residues 755–889)  show that the rod is composed of a series of amphipathic helical bundles. Based on these two structures, we have developed a model of residues 482–889 that is made up of a 5-helix bundle that interacts with the adjacent 7-helix bundle via a hydrophobic patch on each of the bundles. VBS1 (helix 4) is contained within the first bundle, whereas the 7-helix bundle contains an additional four VBSs. We have used a variety of binding assays to monitor the interaction between recombinant talin polypeptides and vinculin, including competitive ELISA, gel filtration and NMR. These show that although VBS1 is cryptic, one or more of the four VBSs in the 7-helix bundle can bind vinculin [60,61].
The crystal structures of several talin VBS peptides bound to the N-terminal region of vinculin (Vd1; residues 1–258) [60,62] confirm that the key determinants of vinculin binding are hydrophobic residues on one face of the talin VBS helix. These are normally buried in the core of the talin amphipathic helical bundles, and indeed contribute to the stability of the fold. Interestingly, NMR shows that a talin 4-helix bundle comprising residues 755–889 undergoes a marked conformational change upon Vd1 binding; indeed, one helix unfolds completely . Together, this suggests a model in which vinculin Vd1 extracts a VBS helix from a helical bundle within the talin rod, and the VBS hydrophobic residues that determine binding specificity become buried within a hydrophobic groove within the N-terminal vinculin Vd1 α-helical bundle . To test this model we have introduced mutations into the talin 755–889 4-helix bundle, which stabilizes the fold as judged by CD at increasing temperatures. As predicted, these mutations markedly reduce vinculin binding, indicating that helical-bundle stability may be an important factor in determining whether a given VBS within the talin rod is available for vinculin binding (B. Patel, A.R. Gingras, A.A. Bobkov, L.M. Fujimoto, R.C. Liddington, D. Mazzeo, J. Emsley, G.C. Roberts, I.L. Barsukov and D.R. Critchley, unpublished work).
How many VBSs are active in whole talin?
To address this question, we used competitive ELISA; we assayed the ability of GST (glutathione S-transferase)–vinculin Vd1 to bind to a constitutively active talin VBS1 polypeptide on plastic, in the presence of increasing concentrations of smooth-muscle talin or recombinant talin polypeptides. The results showed that talin was not an effective competitor in this assay, although talin was able to bind GST–Vd1 in pull down assays. We conclude that one or more of the VBSs in talin are available for binding, but that the affinity of the interaction is low relative to that of talin polypeptides, which have been engineered to contain a constitutively active VBS.
Vinculin binding to talin polypeptides causes a major conformational change in the talin molecule, exposing previously cryptic protease-cleavage sites . Therefore, we explored the possibility that vinculin binding to intact talin might similarly induce local conformational changes in the talin rod, which would be revealed by enhanced sensitivity to protease cleavage. Interestingly, binding of GST–Vd1 to talin exposed a previously cryptic trypsin-cleavage site between residues 898 and 899. Moreover, the N-terminal fragment liberated from the rod that should contain residues 482–898 failed to accumulate as a Coomassie Blue-positive band, although it could be detected using a monoclonal antibody 8d4 (epitope residues 482–655). There are five VBSs within residues 482–889, whereas the adjacent C-terminal region is devoid of VBSs. We conclude that there are one or more active VBSs within the N-terminal region of the talin rod, which upon Vd1 binding undergo a conformational change that renders this region protease-sensitive (B. Patel, A.R. Gingras, A.A. Bobkov, L.M. Fujimoto, R.C. Liddington, D. Mazzeo, J. Emsley, G.C. Roberts, I.L. Barsukov and D.R. Critchley, unpublished work).
What determines the activity of the other VBSs in the talin rod remains to be elucidated. One attractive possibility is that the force exerted on talin by actomyosin contraction may help to destabilize the helical bundles within the talin rod, thereby exposing the VBSs. Alternatively, phosphorylation may play a role. There are a total of 28 tyrosine residues in talin, only 13 of which are found in the over 2000 amino acids that make up the rod. Interestingly, the tyrosines appear to be located at the top or bottom of the helical bundles, and it will be interesting to see whether phosphorylation of these tyrosines affects the activity of the VBSs in the talin rod. However, currently there are no published data to indicate that the talin rod is tyrosine phosphorylated within the cell.
Cell Architecture: from Structure to Function: A Focus Topic at BioScience2005, held at SECC Glasgow, U.K., 17–21 July 2005. Edited by S. Cockroft (University College London, U.K.), Y. Goda (University College London, U.K.), R. Insall (Birmingham, U.K.) and M. Wakelam (Birmingham, U.K.).
Abbreviations: ER, endoplasmic reticulum; ES, embryonic stem; FA, focal adhesion; FAK, FA kinase; FERM, four-point-one, ezrin, radixin, moesin; GFP, green fluorescent protein; MEF, mouse embryo fibroblast; GST, glutathione S-transferase; PIPK, phosphatidylinositol-4-phosphate 5-kinase; VBS, vinculin-binding site
- © 2005 The Biochemical Society