The regions encoding murine talin1 residues 1788–1973 (6h), 1815–1973 (5h) and 1843–1973 (4h) were synthesized by PCR using a mouse talin1 cDNA as template, and cloned into the expression vector pet-151TOPO (Invitrogen). Talin polypeptides were expressed in Escherichia coli BL21 STAR (DE3) cultured in M9 minimal media containing ¹⁵NH₄Cl and/or ¹³C-glucose. His-tagged talin polypeptides were purified by nickel-affinity chromatography. The His-tag was removed by cleavage with AcTEV protease (Invitrogen), and the proteins were further purified by anion-exchange.

NMR spectra of all the proteins were obtained at 298 K using Bruker AVANCE DRX 600 or AVANCE DRX 800 spectrometers both equipped with CryoProbes, with 1 mM protein in 20 mM sodium phosphate pH 6.5, 50 mM NaCl, 2 mM DTT, 10% (v/v) ²H₂O. Spectra were processed with TopSpin (Bruker) and analysed using ANALYSIS [13]. Resonance assignment was carried out as described [14]; assignments of 1815–1973 have been deposited in the BioMagResBank (http://www.bmrb.wisc.edu) with the accession number 15625.

These were carried out as described previously [8]; structural statistics are presented in Table 1. The set of 20 lowest energy structures has been submitted to the Protein Data Bank (www.rcsb.org), accession number 2kvp. Molecular models were generated using PYMOL [15].

Far-UV CD spectra were recorded, using a JASCO J-715 spectropolarimeter, over the wavelength range 200–250 nm (scan rate 50 nm min⁻¹) in a quartz cell of 0.1 cm path length with ∼25 μM protein in 20 mM sodium phosphate, pH 6.5, 50 mM NaCl.

Analytical gel filtration chromatography using Superdex-75 (10/300) GL (Amersham Biosciences) was used to measure binding of talin polypeptides to the vinculin Vd1 domain. Polypeptides were incubated at various temperatures for 30 min prior to loading onto the column, which was pre-equilibrated and eluted with 20 mM Tris pH 8.0, 150 mM NaCl, 2 mM DTT at a flow rate of 0.8 mL/min at room temperature.

Our earlier studies of talin residues 1843–1973, which contains a vinculin-binding site referred to as VBS3, showed that its [¹H, ¹⁵N]-heteronuclear single quantum coherence (HSQC) spectrum had good dispersion with peak line widths consistent with a monomeric state (Fig. 1B) and the structure (PDB ID: 2B0H) showed a well folded 4-helix bundle domain [11]. However, in the light of our subsequent demonstration that the domains on either side (residues 1655–1822, 1974–2140, 2137–2294 and 2300–2482) are all 5-helix bundles [4,5,8], we sought to establish whether residues 1843–1973 could also be part of a larger 5-helix bundle. To test this we produced a series of constructs with different boundaries containing 4-helices (1843–1973), 5-helices (1815–1973) and 6-helices (1788–1973) (Fig. 1B). Each of these polypeptides expressed well and was soluble.

Addition of a further predicted helix at the N-terminus resulted in a polypeptide (residues 1815–1973) whose spectrum showed a similar pattern of chemical shifts to that of residues 1843–1973 but with better chemical shift dispersion and similar signal intensities for all the peaks, suggesting that the 5-helix construct has a more stable fold. Addition of a sixth helix (residues 1788–1973) did not affect the signals of the 5-helix bundle but introduced a cluster of sharp signals close to the middle of the [¹H, ¹⁵N]-HSQC spectrum, suggesting that the additional residues were unfolded (Fig. 1B).

The solution structure of the talin domain comprising residues 1815–1973 was calculated from 5117 NOE-based distance and 232 dihedral angle restraints. The structure consists of five anti-parallel amphipathic α-helices forming a bundle with up-down-up-down-up left-handed topology (Fig. 1C and D). Helices 2–5 bear a strong resemblance to the 4-helix bundle structure previously determined (PDB ID: 2B0H [11]), with an RMSD of 1.2 Å for the region 1843–1973 and the same topology (Fig. 2A–D). The 4-helix up-down, left-handed twist bundle is a common fold (SCOP 47161), and it forms the core of this and other domains in the talin rod. However in talin, uniquely, this fold is augmented by an extra N-terminal helix (here residues 1825–1842) connected by a long (9 residue) loop that allows this helix to pack against helices 3 and 4 of the bundle. Helix-1 is relatively short compared with the other helices (15 residues compared to 27–28 residues) and sits half way down the bundle, packing against the core of the bundle by an extensive hydrophobic interface (Fig. 2C). Like the other helices, helix-1 is amphipathic and has its more hydrophobic surface buried in the bundle, packing against the hydrophobic groove between helices 3 and 4. The aromatic ring of Y1893 partially protrudes from the side of the 4-helix core of the bundle but the small side chain of A1834 in helix-1 allows close packing of the helices. This 5-helix left-handed crossover connectivity (Fig. 2A) has been observed in several other talin bundles (482–655, 1655–1822, 1974–2140 and 2137–2294; SCOP ID 109879) but, to date, has not been observed in other proteins. We show here in the case of residues 1815–1973 that the extra helix has very striking effects on the properties of the domain.

The 4-helix up-down-up-down fold results in the N- and C-termini of the domain being at the same end of the bundle (Fig. 3A), whereas with a fifth helix the N- and C-termini are at opposite ends of the bundle. This is a key difference in a long rod such as talin; it is possible to arrange 5-helix bundles in the rod with the component helices parallel to the rod axis, whereas this is not possible for 4-helix bundles (Fig. 3B and C).

This clear difference in the orientation of the helices relative to the long axis of the rod will result in significant differences in the mechanical properties of these two arrangements, which may be important for the activation of vinculin binding (see below).

The 4-helix construct, residues 1843–1973, forms a stable bundle with a melting temperature of 58° (Fig. 2E). Whilst the addition of the extra helix has only minimal structural effect on the core of the bundle (Fig. 2B and D) it has a substantial effect on the stability of the bundle, the 5-helix bundle having a melting temperature of 74 °C (Fig. 2E).

The four helix bundle containing VBS3 can bind the vinculin Vd1 domain at room temperature [11] (Fig. 4). However, since the addition of the extra helix has a large effect on the stability of the bundle we anticipated that this would affect vinculin binding, since this requires unfolding of the bundle [9–11,16,17], and indeed the 5-helix bundle requires incubation at ⩾37 °C for binding to be observed (Fig. 4). This result highlights the importance of using complete domains for vinculin binding studies, and suggests that the vinculin-binding site (VBS) in the 1815–1973 5-helix bundle is unlikely to be constitutively active in full length talin.

The various vinculin binding sequences in talin are similar [3] and the corresponding peptides all bind tightly to vinculin in vitro. However, this and previous studies show that VBSs buried within stable 5-helix bundles are cryptic, whereas those in 4-helix bundles are constitutively active [9–11,16,18]. Experimental and theoretical studies show that vinculin binding to cryptic binding sites in talin can be activated by mechanical strain [19–22], and the present results indicate that this is likely to be essential for all the VBSs in the C-terminal part of the talin rod which are contained within 5-helix bundles (Fig. 3C). By contrast, two VBSs in the central and N-terminal part are likely to be constitutively active and may provide the initial attachment points for vinculin.