RIAM and Vinculin Binding to Talin Are Mutually Exclusive and Regulate Adhesion Assembly and Turnover^*

Protein Expression and Purification

Talin constructs were synthesized by PCR using a mouse talin1 cDNA as template and cloned into pET-151/d-TOPO (Invitrogen). Constructs were expressed in Escherichia coli BL21 Star(DE3) cells cultured either in LB, or for preparation of ¹⁵N-labeled samples, in minimal media containing 1 g of ¹⁵N-ammonium chloride per liter. Recombinant His-tagged talin polypeptides were purified as described previously (16), and their concentrations were determined using extinction coefficients at 280 nm. Recombinant His-tagged chicken vinculin domain 1 (Vd1; residues 1–258) was expressed from a pET-15b plasmid and purified as described by Ref. 18. Constructs encoding RIAM residues 45–127, 147–174, and 1–127 were synthesized by PCR using a mouse RIAM cDNA as template and cloned into pET-151/d-TOPO. Purified RIAM wild type (6-30, EDIDQMFSTLLGEMDLLTQSLGVDT) and mutant (6-30-4E, EDIDQEESTEEGEMDLLTQSLGVDT) peptides were a kind gift from Mark Ginsberg (University of California, San Diego).

Vinculin Binding

Analytical gel filtration chromatography using either Superdex-75 (10/300) GL or Superdex-200 (10/300) (Amersham Biosciences) was used to assay binding of talin polypeptides to the vinculin Vd1 domain. Polypeptides were incubated at various temperatures for 30 min prior to gel filtration. Columns were pre-equilibrated and eluted with 20 mm Tris, pH 8.0, 150 mm NaCl, and 2 mm DTT at a flow rate of 0.8 ml/min at room temperature. Fractions were analyzed using SDS-PAGE.

Size Exclusion Chromatography Multiangle Laser Light Scattering (SEC-MALLS)

Proteins (0.5 ml) were separated on a Superdex 75 10/300 gel filtration column (GE Healthcare Life Sciences) equilibrated in 20 mm Tris (pH 7.4), 150 mm NaCl, 2 mm DTT at 0.7 ml/min. Elution was monitored by a Wyatt EOS 18-angle laser photometer (Wyatt Technology), an Optilab rEX refractive index detector, and a Jasco UV-2077 Plus UV/Vis spectrophotometer (Jasco). Molar mass determinations were performed using the Astra software (version 5.3.2.16) (Wyatt Technology).

Circular Dichroism (CD) Spectroscopy

CD spectra were recorded using a JASCO J-715 spectropolarimeter equipped with a JASCO PTC-348WI temperature control unit. Far-UV CD spectra were recorded at 20 °C over the wavelength range 200–250 nm in a quartz cell of 0.1-cm path length (scan rate, 50 nm·min⁻¹). Proteins were dissolved in 20 mm sodium phosphate (pH 6.5), 50 mm NaCl at a concentration of 10 μm. For denaturation studies, the unfolding of α-helices was followed at 222 nm.

Isothermal Titration Calorimetry (ITC)

ITC data were collected using a VP-ITC microcalorimeter (MicroCal, Ltd., Northampton, MA) and analyzed by fitting to a single-site binding equation or a two-site binding equation using MicroCal Origin software.

NMR Spectroscopy

Proteins were prepared in 20 mm sodium phosphate (pH 6.5), 50 mm NaCl, 2 mm DTT with 10% (v/v) ²H₂O. NMR spectra were obtained at 298 K using Bruker AVANCE DRX 600 or AVANCE AVII 800 spectrometers both equipped with CryoProbes. Resonance assignment and structure calculations were carried out as described previously (9, 25); the structural statistics for each domain are presented in supplemental Table 1.

Protein Crystallization

Vinculin Vd1(1–258) was co-crystallized with RIAM residues 1–32 (2:1 peptide excess) in 0.05 m Tri-sodium citrate, 1.2 m ammonium sulfate, 3% isobutanol. Crystals were vitrified in the mother liquor containing 20% glycerol and 10% isobutanol. Data were collected to 2.3 Å at base line I24 at the Diamond Light Source. Data were processed and integrated using MOSFLM (26) and SCALA (27). Molecular replacement was performed using PHASER and the template structure (PDB ID 1XWJ) and refined using REFMAC (28). Statistics of the refinement are presented in supplemental Table 2.

Cell Culture, Transfection, Immunofluorescence, and Imaging

Human umbilical vein endothelial cells (HUVEC) were grown according to the manufacturer's instructions (PromoCell). For siRNA knockdown of human talin1, Stealth Select RNAi^TM oligo 804 (Invitrogen catalog no. 1299003) was used and a Universal Stealth RNA^TM siRNA was used as control (6). Cells (6 × 10⁶ per ml) were electroporated (Microporator; Invitrogen) with 80 pmol of talin1 siRNA or control RNA and grown on tissue culture plastic for 72 h. Cells (4 × 10⁴ cells) were then replated on 22-mm glass coverslips and analyzed 4 h later. Where appropriate, cells were co-transfected with 10 μg of plasmid DNA encoding GFP or GFP-tagged mouse talin1. Cells were fixed, stained for paxillin (AB-88 Sigma) or vinculin (monoclonal antibody F9, a gift from V. Koteliansky), imaged, and analyzed as described previously (6).

Ratio Imaging

Vinculin null mouse embryonic fibroblasts were cultured in Dulbecco's Modified Eagle's Medium (DMEM), supplemented with penicillin/streptomycin, 10% FCS, 2 mm l-glutamine, non-essential amino acids, and β-mercaptoethanol (all from Sigma-Aldrich). Cells were transfected with GFP-vinculin Vd1 and RIAM-mCherry (kind gift from Dr. Maddy Parsons; Kings College, London, UK) using Lipofectamine Plus (Invitrogen) according to the manufacturer's instructions and plated 3 h post transfection on glass-bottomed dishes (MatTek Corp.) coated with 10 μg/ml bovine plasma fibronectin (Sigma-Aldrich). After fixation with 3% paraformaldehyde at 24–36 h after transfection, cells were imaged using an inverted microscope (IX70; Olympus) controlled by a Deltavision system (Applied Precision) equipped with a Coolsnap HQ CCD (Photometrics) camera, suitable filter sets (Chroma), and a 100× numerical aperture 1.35 Uplan Apo objective. Ratio images of two components were calculated as described previously (29). Images showing localization of GFP-vinculin Vd1 and RIAM-mCherry were aligned, and ratio values were calculated for each pixel above threshold in both GFP and mCherry channels. Ratio images were displayed using a spectral color look-up table ranging from 0.2 (blue) to 5 (red). To compensate for different expression levels and the different photon yields of GFP and mCherry fusion proteins, the ratios were normalized by a constant that shifted their average values toward 1. For the presentation of ratio images were “smoothed” by averaging 2 × 2 pixels.

The Overall Structure of the Talin Rod

The 2,059 residues of the talin rod are predicted to form 62 α-helices separated by short unstructured regions (12). Using a combination of limited proteolysis, NMR, and x-ray crystallography, we previously identified and solved the structures of two adjacent domains at the N terminus (residues 482–786) (18) and seven sequential domains (residues 1357–2482) at the C terminus of the talin rod (13, 15–17, 19). All of these domains form compact four- or five-helix bundles stabilized by hydrophobic cores. However, the domain organization of residues 787–1356 (18 predicted helices) proved far more difficult to establish, and stably folded polypeptides could be obtained with coordinates incompatible with those of established neighboring domains (17, 30, 31). We therefore expressed multiple fragments containing either four or five helices (supplemental Fig. S1) and assessed their fold by NMR. Only the four non-overlapping fragments 787–911 (four helices), 913–1044 (four helices), 1046–1206 (five helices), and 1206–1357 (five helices) gave ¹H,¹⁵N HSQC spectra with high chemical shift dispersion and the uniform line widths characteristic of stable globular domains (supplemental Fig. S2). The four domains identified in this way complete the domain mapping of the talin rod, which consists of 13 globular domains (R1 to R13), terminating in a single helix responsible for talin dimerization (Fig. 1A). The structures of the four new domains R3–R6 are shown in Fig. 1, B–E, and supplemental Fig. S3 and are described in the supplemental data.

The talin rod begins with a five-helix bundle (R1) followed by three four-helix bundles (R2–R4). R2 makes extensive hydrophobic interactions with R1 to form a double-domain (18). Although no interactions have been detected between R2, R3, and R4, the R1–R4 region will have a compact arrangement (Fig. 1F) because the N and C termini of four-helix bundles are at the same end of the domain (e.g. see Fig. 1, B and C). The three four-helix bundles are followed by a series of eight five-helix bundles in which the N and C termini are positioned at opposite ends of the bundle (e.g. see Fig. 1, D and E), creating a linear chain (Fig. 1, A and F). The four-helix bundle R8 does not disrupt the chain because it is inserted into a loop in the R7 five-helix bundle (14). Importantly, four-helix and five-helix bundles are expected to differ in their response to mechanical stretch. Unfolding five-helix bundles by pulling on the termini positioned at the opposite ends of the bundle is restricted by extensive contacts throughout the length of the helices and requires a gradual breaking of hydrogen bonds. In four-helix bundles, the termini are at the same end of the bundle, and the applied force acts on the weak hydrophobic contacts, peeling helices away from the bundle. Molecular dynamics simulations on the N-terminal fragments of the talin rod demonstrate that the R1 five-helix bundle is more resistant to stretch than the four-helix bundle R2 and the bundle made by helices H9–H12 (32). Thus, the uneven distribution of four- and five-helix bundles in the talin rod creates two distinctly different zones: a compact N-terminal region sensitive to stretch and a linear C- terminal part that is optimal for force transmission.

Vinculin Binding to the Compact N-terminal Region of the Talin Rod

The talin rod contains multiple vinculin binding sites (VBS), with each defined by hydrophobic residues on a single helix that are normally buried within the helical bundles that make up the rod (Fig. 1A) (19). Thus, vinculin binding requires domain unfolding (33), and domain destabilization by force has been shown to promote binding (34). VBS are found in both four- and five-helix bundles, and when present, there is generally just one VBS per domain. However, the two four-helix bundles R2 and R3 are atypical, with each containing two VBS. This creates a cluster of VBS in the compact N-terminal region of the talin rod optimal for activation by force (Fig. 1, A and F).

Gel filtration experiments show that R3 binds the vinculin Vd1 domain (residues 1–258) readily at 25 °C, whereas R2 requires incubation at 37 °C for maximal binding (Fig. 2, A and B). The multidomain fragments R1–R3 and R1–R5 bind vinculin at 25 °C to the same extent as R3 alone (Fig. 2, C and D), demonstrating that the vinculin binding properties of R3 are not affected by adjacent domains. Size-exclusion chromatography multiangle laser light scattering data show that at 25 °C, R2R3 forms a 1:2 complex with vinculin, indicating that both VBS in R3 are engaged (supplemental Fig. S4). Incubation at 40 °C led to formation of a 1:4 complex, indicating that all the VBS in both R2 and R3 are now occupied.

Our new structural data reveal that R3 contains four threonines within its hydrophobic core creating a potentially weak point in the core of this domain (Fig. 2E). Replacing these threonines with larger hydrophobic residues (T809I/T833V/T867V/T901I, the IVVI mutation) markedly stabilized R3 with minimal effects on its overall structure (supplemental Fig. S5, A and B). Remarkably, the R3 IVVI mutant was no longer able to bind Vd1 at 25 °C (supplemental Fig. S5, C and D) although it bound efficiently at 45 °C (supplemental Fig. S5D), confirming that the mutations do not inactivate the VBS. Furthermore, introducing the IVVI mutation into R1–R3 (Fig. 2F) and R1–R5 rendered these larger N-terminal fragments unable to bind vinculin. We conclude that the presence of the four threonines in R3 enhances its ability to bind vinculin and that the VBS in R3 will be among the first to bind vinculin in response to force exerted on talin.

Multisite Interactions between Talin and RIAM

The Rap1 effector RIAM has been implicated in talin and thereby integrin activation (23, 24). The minimal talin-binding site (TBS) in RIAM has been mapped to residues 6–30 (TBS1), and a TBS1 peptide was shown to be sufficient to bind talin (24). The TBS1 sequence is highly conserved and is predicted to contain an amphipathic helix (24). To map the RIAM binding sites in the talin rod, we used the TBS1 peptide to conduct an NMR-based screen for possible interactions with each of the 13 authentic talin rod domains and identified four domains (R2, R3, R8, and R11) that bound RIAM-TBS1 (Fig. 3A and supplemental Fig. S6). Control NMR titrations showed that a mutant RIAM-TBS1 peptide (M11E/F12E/L15E/L16E), which does not bind full-length talin (24), did not bind any of these four domains, indicating that the interactions observed are specific.

In addition to RIAM-TBS1 (residues 6–30), two more regions in the N-terminal part of RIAM (residues 50–85, TBS2, and 149–173, α3) are predicted to form helical structures (Fig. 3B and supplemental Fig. S7A). We therefore tested constructs (45–127 and 147–174) containing these regions for an interaction with the talin rod domains that bind RIAM-TBS1 (R2, R3, R8, or R11). RIAM(147–174) containing α-helix 3 had no effect on the NMR spectra of any of these rod domains (Fig. 3C). However, RIAM(45–127) induced spectral changes that were similar to those caused by RIAM(6–30) (Fig. 3C), indicating comparable interactions. We conclude that the N-terminal region of RIAM contains two sequential talin binding sites TBS1 (residues 6–24) and TBS2 (residues 50–85) (numbers based on the secondary structure prediction).

The identification of two adjacent TBS in RIAM raises the possibility that simultaneous engagement of both sites might enhance binding to the talin rod. Moreover, two of the four RIAM binding sites in the talin rod are in adjacent domains (R2 and R3). Binding affinities determined by ITC show that R3 binds RIAM-TBS1 with a K_D of 36 μm (Fig. 3D), whereas the interaction with R2 was at least an order of magnitude weaker as judged by the relative chemical shift changes in the HSQC spectra of ¹⁵N-labeled R2 and R3 induced by RIAM-TBS1. In contrast, a RIAM polypeptide (residues 1–127) containing both TBS1 and TBS2 bound >10× more tightly to a talin R2R3 double domain fragment (K_D of 2 μm determined by ITC) (Fig. 3E). The number of binding sites derived from the ITC data (n = 0.8) indicates formation of a 1:1 talin R2R3·RIAM(1–127) complex. This was confirmed by size-exclusion chromatography multiangle laser light scattering; the molecular weight of the complex was 40 kDa, consistent with the calculated molecular weight (41.3 kDa) of a 1:1 complex (supplemental Fig. S7B). The HSQC spectrum of ¹⁵N-labeled R2R3 mixed 1:1 with RIAM(1–127) shows severe broadening of the resonances of both R2 and R3 (supplemental Fig. S7C), indicating that both domains are simultaneously engaged in the interaction with RIAM, restricting their relative mobility.

Competition between RIAM and Vinculin for Talin

Interestingly, all four talin rod domains that bind RIAM also contain VBS (Fig. 4A). However, vinculin and RIAM have markedly different modes of binding to talin. The talin binding sites in RIAM are single helices that interact with a folded talin domain, as shown by the localized chemical shift changes in the NMR spectra of R2, R3, R8, and R11 following addition of a RIAM(6–24) peptide (Fig. 3C and supplemental Fig. S6). The positions of many talin resonances do not show large changes even at high RIAM/talin ratios, indicating that no major structural rearrangements occur in regions not in direct contact with RIAM. In contrast, vinculin binding requires domain unfolding (33) and is therefore likely to be incompatible with RIAM binding (Fig. 4B). The different modes of binding are demonstrated by experiments with the IVVI mutation that stabilizes the R3 bundle. Although this mutation dramatically reduced vinculin binding (Fig. 2F), it had minimal effects on RIAM binding as determined by NMR (supplemental Fig. S8, A and B). Overall, the data suggest that RIAM and vinculin must compete for binding to the talin rod. Indeed, gel filtration experiments show that addition of RIAM(1–127) to a 1:1 mixture of Vd1/talin R2R3 results in the disappearance of the talin·vinculin complex and the appearance of a lower molecular weight peak close to the position of the talin·RIAM complex (Fig. 4C). This clearly demonstrates that vinculin and RIAM cannot bind to talin simultaneously and suggests that RIAM has a higher affinity for R2R3 than vinculin.

The NMR spectral changes in R3 following RIAM binding map predominantly to the sequential VBS helices that form a hairpin structure (Fig. 4D). We conclude that the RIAM helix likely binds to the groove on the surface of the R3 bundle made by the two contacting VBS helices. This type of interaction is likely to shield the VBS in R3 from vinculin, enhancing vinculin-RIAM competition for talin. In R2, the RIAM interactions map predominantly to the first helix that is also a VBS. Additional perturbations on the neighboring helix 2 and lack of perturbations on helix 4 suggest that RIAM occupies the groove between helices 1 and 2. Significantly, this surface is not involved in the close contact between the R1 and R2 domains (18), leaving it exposed in the R1R2 fragment, and NMR experiments show that R1R2 is able to interact with RIAM (supplemental Fig. S8C).

Interaction between RIAM and Vinculin

Although addition of a RIAM(1–127) peptide disrupts the talin·vinculin complex, a peak corresponding to free vinculin was not observed in the gel filtration experiments (Fig. 4C), raising the possibility that the displaced vinculin might bind RIAM. Indeed, the gel filtration profile of a 1:1 mixture of Vd1 and RIAM(1–127) clearly demonstrates complex formation (Fig. 5A). Furthermore, a Vd1·RIAM(1–32) complex readily crystallized under conditions similar to those used for the Vd1·talin peptide complex (18, 19); its structure was solved by molecular replacement using the Vd1/talin structure and refined to 2.3 Å (Fig. 5B and supplemental Table 2). The conformation of vinculin Vd1 is very similar in both talin and RIAM complexes (supplemental Fig. S9A). As in the Vd1/talin VBS peptide structures, the RIAM TBS1 helix is embedded in the hydrophobic groove formed by α-helices 1 and 2 of the Vd1 four-helix bundle (Fig. 5, C and D). The position and register of the RIAM helix is well defined by the electron density (supplemental Fig. S9, B and C), and its interaction with Vd1 is mediated by the hydrophobic side chains on one face of the amphipathic RIAM helix. These residues align well with the talin VBS residues contacting Vd1 (Fig. 5E), but most of the RIAM residues have bulkier hydrophobic side chains than those found in the talin VBS consensus sequence. As a result, the RIAM helix is positioned further away from the Vd1 core than talin VBS helices, reducing the packing density. In addition, Leu¹⁹⁶⁵ of the VBS in talin R10, which embeds in the Vd1 hydrophobic core, corresponds to Thr²³ in RIAM that makes no significant contacts with Vd1.

Isothermal titration calorimetry (ITC) data (supplemental Fig. S10) demonstrate that vinculin Vd1 has a significantly lower affinity for RIAM (1–32) (K_D, 5 μm) compared with the VBS (residues 820–844) in talin R3 (K_D, 0.8 μm). The lower affinity for RIAM is associated with a higher enthalpy of interaction than for talin (−1.5 and −5.9 kcal, respectively). This correlates with the displacement of the RIAM helix further away from the Vd1 bundle relative to the talin VBS, which reduces Vd1 contacts with RIAM.

The Balance between RIAM and Vinculin Binding to Talin in Cells

To explore whether our in vitro data on competition between RIAM and vinculin for talin applies in cells, we co-expressed RIAM-mCherry and GFP-vinculin fusion constructs in vinculin null mouse embryonic fibroblasts. In stationary cells, both vinculin and RIAM localization overlapped in mature FAs (Fig. 6, upper panels). Ratio imaging revealed that RIAM preferentially localized to the distal tip of FAs, whereas vinculin intensity was high in the proximal area. In cells with a migratory arc-shaped phenotype, the rim of the protruding cell edge was particularly rich in RIAM, whereas vinculin was prominent in FAs behind the leading edge, suggesting that RIAM binding to talin precedes the binding of vinculin to talin during FA formation (Fig. 6, lower panels).

The N-terminal Region of the Talin Rod Is Required for FA Formation

Although both the RIAM- and vinculin-binding sites are distributed throughout the talin rod (supplemental Fig. S11), the unusual features of the RIAM and vinculin binding sites located in the compact N-terminal region of the rod suggest that these may play a unique role in FA assembly. To explore this possibility, we used our detailed knowledge of the structures and domain boundaries of talin1 to design two mini-talin1 constructs (Fig. 7A). The first contained just the N-terminal FERM domain (residues 1–405), which includes the integrin binding and activation domains, coupled via the authentic unstructured linker region (residues 405–482) to the C-terminal part of the talin rod (residues 1974–2541; R11–R14). This region contains the C-terminal actin-binding site and dimerization domain, plus several RIAM and vinculin binding sites, and has recently been shown to be sufficient to support the signaling pathways that regulate cell cycle progression (35). The GFP-tagged mouse TlnΔR1-R10 construct was then tested for its ability to rescue defects in cell spreading and FA assembly in human endothelial cells depleted of talin1. However, although GFP-TlnΔR1-R10 was expressed to about the same extent as GFP-talin1 (supplemental Fig. S12A), it was unable to support efficient cell spreading on glass coverslips within a 4-h time frame, although the cells displayed numerous protrusions (Fig. 7, B–D, and supplemental Fig. S12B). Moreover, the cells contained very few paxillin- or vinculin-staining FAs, and GFP-TlnΔR1-R10 was diffusely distributed throughout the cytosol (Fig. 7D). In marked contrast, a second GFP-tagged mini-talin1 construct (GFP-TlnΔR4-R10) that includes the vinculin and RIAM binding sites in R2R3 (Fig. 7A) supported assembly of abundant paxillin- and vinculin-containing FAs (Fig. 7, B and E) and partially rescued cell spreading (supplemental Fig. S12, B and C). These results clearly establish that the N-terminal part of the talin rod plays a key role in cell spreading and the assembly of vinculin-containing FAs.