Structural and Mechanistic Insights into the Recruitment of Talin by RIAM in Integrin Signaling

Comparison of RIAM TBS1 and TBS2 regions in binding to cytoplasmic talin

The N-terminal flexible region of RIAM preceding the RA domain (residues 1-178) contains at least three predicted helical sequences (Fig. 1B: TBS1, TBS2, and CC) (Wynne et al., 2012; Zhang et al., 2014). Two of these sequences exhibit binding affinities to various talin domains and are named TBS1 and TBS2 (Goult et al., 2013b; Lee et al., 2009). Prior to its translocation to the PM, the homodimeric talin protein adopts a highly compacted autoinhibitory conformation (Goksoy et al., 2008; Goult et al., 2013a; Song et al., 2012). This conformation may partially mask the RIAM-binding domains. To assess the interaction of TBS1 and TBS2 with full-length cytoplasmic talin, we performed a pull-down assay using GST-tagged RIAM fragments (TBS1: residues 1-30; TBS2: residues 27-93; and TBS1-2: residues 1-93) with full-length talin expressed in mammalian cells. Both TBS1 and TBS1-2 exhibit a strong association with full-length talin at similar levels. However, the association of TBS2 with talin is undetectable in this binding assay (Fig. 1C).

Characterization of the primary RIAM binding sites in talin

Several talin rod domains including the R3 and R8 domains have been shown to interact with RIAM TBS1 (Goult et al., 2013b). Because it is suggested that the four helical-bundle R3 domain is intrinsically unstable and undergoes unfolding in response to force exerted on talin (Goult et al., 2013b), we assessed the binding affinities of RIAM TBS1, TBS2, and TBS1-2 fragments with talin R2R3 and R7R8 fragments. Recombinant R2R3 and R7R8 proteins were examined by CD spectra to confirm proper folding (Fig. S1A). The TBS1 and TBS1-2 fragments bind to both R2R3 and R7R8 in the in vitro binding assays, whereas the interactions of the TBS2 fragment with R2R3 and R7R8 are much weaker (Fig. 1D). We then assessed the binding affinities of TBS1 or TBS1-2 with R2R3 or R7R8 using quantitative pull-down assays. While the TBS1-binding affinities of R2R3 and R7R8 are both in the low micromolar range, TBS1 binds to R7R8 more strongly than R2R3, and the binding affinities of TBS1-2 with R7R8 or with R2R3 are similar to those of TBS1 (Fig. S1B-E). These results confirm that both R3 and R8 domains directly bind the TBS1 fragment and suggest that the R8 domain (in the form of R7R8) is a stronger TBS1-binding site.

Structure of the RIAM TBS1 in complex with the talin R7R8

To better understand the structural basis for the interaction between RIAM and talin, we determined the crystal structure of a RIAM TBS1 peptide (residues 5-25) in complex with the talin R7R8 domains (residues 1357-1657) at 1.5 Å resolution (Table 1). The asymmetric unit possesses one R7R8 molecule with a well-defined TBS1 fragment (Fig. S2A). The TBS1 peptide interacts with the talin R8 domain, but not with the R7 domain (Fig. 2A). Although TBS1 also forms hydrogen bonds with the symmetrically related R7 domain (Fig. S2B), removal of the R7 domain did not affect the association of TBS1 with the R8 domain, suggesting that these hydrogen bonds are likely the result of crystal packing. The TBS1 peptide binds to the R8 domain at the α2 and α3 helices via both hydrophobic and electrostatic interactions. The association is mediated primarily through a large hydrophobic contact interface formed by multiple side chains (Ile8, Met11, Phe12, Leu15, and Leu22 in RIAM TBS1 and Leu1492, Ala1495, Ala1499, Ala1529, Ala1533, Thr1536, Val1540, Cα of Arg1510 and Lys1544 in the R8 domain) (Fig. 2B), and is further fortified by several electrostatic interactions (Asp9_RIAM-Lys1544_talin, Glu18_RIAM-Arg1510_talin, and Glu18_RIAM-Asn1507_talin) (Fig. 2C).

Strikingly, in the structure of TBS1:R7R8 complex, the TBS1 peptide adopts an unexpected helical conformation with an approximately 55° kink at the backbone of residue Ser13 (Fig. 2D). An earlier study reported a structure of the RIAM TBS1 in complex with the vinculin Vd1 domain, in which a conventional α-helical TBS1 interacted with the α1, α2 and α4 helices of the Vd1 domain (Fig. 2D) (Goult et al., 2013b). The side-chain hydroxyl group of Ser13 in the R7R8-bound TBS1 forms a hydrogen bond with the backbone of Gln10 and a water-mediated hydrogen bond network with the backbones of Met11 and Thr14 (Fig. 2D). These interactions stabilize the helical kink at Ser13 in the TBS1:R7R8 complex. The kinked TBS1 helix allows additional hydrophobic side chains proximal to the N-terminus, including Ile8, Met11, and Phe12, to participate in the hydrophobic interactions with the R8 domain. This conformation also positions the Asp9 residue towards the α3 helix of the R8 domain to form the Asp9_RIAM-Lys1544_talin salt bridge. As a result, the average B-factor for the R7R8-bound TBS1 is 18.4 Ų for main chain atoms (18.7 Ų for all atoms), representing a much more stable conformation than that of the Vd1-bound TBS1 (average B-factor is 56.1 Ų for main chain atoms and 57.1 Ų for all atoms) (Fig. S2C). The shape complementary of the TBS1:R7R8 interaction (0.773) is also higher than that of the TBS1:Vd1 interaction (0.714) (Lawrence and Colman, 1993). The refined model was analyzed by MolProbity (Chen et al., 2010). 99.7% of all residues are in the favored regions, and 100% of all residues are in the allowed regions of the Ramachandran plot. The Clashscore is 1.24 (99^th percentile) and the overall MolProbity score is 0.84 (100^th percentile). Thus, the crystal structure of the R7R8:TBS1 complex suggests that binding to R7R8 induces a more stable kinked helical conformation of the TBS1 fragment, thus enhancing the TBS1:talin interaction.

Identification of the residues critical for TBS1:talin binding

To validate the side chain interactions observed in the crystal structure, we constructed several single point mutations in the RIAM TBS1 region (D9A, L15Y, and E18A) and a surface mutation V1540Y in the R8 domain, and evaluated their effect on the association of the RIAM TBS1 region with full-length talin. Additionally, to assess the contribution of the helical kink to the interaction, we also mutated Ser13 to a glycine to destabilize the kink in the RIAM TBS1 region. The D9A mutation in the RIAM TBS1 fragment and the V1540Y mutation in the R8 domain diminished the TBS1:talin association, and the other TBS1 mutations (L15Y, E18A, and S13G) completely abolished the interaction with full-length talin in the pull-down assays (Fig. 2E). In contrast, another talin surface mutation at a non-interacting residue (T1520Y) in the same helix as V1540Y exhibited no reduction in RIAM TBS1 binding (Fig. 2F). These data demonstrate that not only the hydrophobic and electrostatic interactions, but also the helical kink, are required for the high affinity TBS1:talin association.

The R3 domain may bind to RIAM TBS1 via a highly conserved interface

Although both the talin R3 and R8 domains contain four helical bundles, their primary sequences are significantly different with a sequence identity of less than 20%. Surprisingly, structure-based sequence alignment of the two domains reveals that the residues that interact with RIAM TBS1 in the R8 domain are completely conserved (75% identical and 25% similar) in the R3 domain (Fig. 2G). To examine whether the R3 domain also interacts with TBS1 via this conserved surface, we assessed the effects of the corresponding mutations on the association of TBS1 with the R2R3 domains and with full-length talin. Both single mutations, S13G and L15Y of TBS1, significantly diminished the TBS1:R2R3 interaction (Fig. 2H). Furthermore, the V871Y mutation in the R3 domain, equivalent to the V1540Y mutation in the R8 domain, also diminished the interaction of RIAM TBS1 with the full-length talin. Strikingly, the V871Y/1540Y mutant abolished the interaction (Fig. 2I), suggesting that the interactions of TBS1 with other talin rod domains are insignificant. These results suggest that RIAM TBS1 interacts with the talin R3 domain in a mode highly similar to the TBS1:R8 interaction via a conserved interface.

Binding determinants in the TBS1:R7R8 complex and the helical kink in TBS1 are required for RIAM:talin co-localization at the PM and for integrin activation

It has been shown that co-transfection of talin and RIAM TBS1, containing a C-terminal CAAX sequence directing it to the PM, in stably expressing αIIbβ3-integrin A5 cells, bears the capacity to both activate integrins and facilitate the cell surface co-clustering of these overexpressed proteins (Lee et al., 2009). To assess the role of TBS1 single mutants in regulating talin recruitment in integrin signaling, we compared the TBS1 single mutants S13G, L15Y, and E18A to GFP-RIAM-TBS1-CAAX control for their ability to co-cluster with mCherry-talin. As expected, wild type TBS1 containing CAAX and talin co-localize and form co-clusters at the lamellae, whereas RIAM TBS1 single mutants S13G, L15Y, and E18A failed to recruit talin into clusters (Fig. 3A,B). General Pearson’s correlation coefficient (PCC) showed modest, but significant higher correlations for wild type: RIAM-TBS1-CAAX (0.9434 ± 0.02) compared to S13G (0.9086 ± 0.07), L15Y (0.8620 ± 0.08) and E18A (0.8726 ± 0.07) (Fig. 3C). Nonetheless, when calculated solely at the cell’s lamellae (see Materials and Methods for details), differences in observed PCC values were more substantial: RIAM-TBS1-CAAX (0.9539 ± 0.02) compared to S13G (0.8400 ± 0.05), L15Y (0.7076 ± 0.06) and E18A (0.7785 ± 0.05) (Fig. 3D). These results, in full concordance with our crystal structure and with the in vitro binding studies, suggest that binding determinants and the helical kink in TBS1 are required for TBS1:talin co-clustering.

We then examined the effect of the TBS1 mutations on integrin activation in a well-accepted fluorescence-activated cell sorting (FACS) assay. Co-transfection of RIAM TBS1-CAAX and talin in A5 cells promotes activation of αIIbβ3 integrins, and this effect can be inhibited by EDTA and an αIIbβ3 integrin-specific inhibitor, Eptifibatide (Fig. 4A). The TBS1 mutants, including S13G, L15Y, and E18A, significantly diminish integrin activation (Fig. 4B). Full-length RIAM bearing these mutations also exhibit impaired function in promoting integrin activation when co-expressed with talin (Fig. 4C). To compare the effect of TBS1, TBS2, and TBS1-2 on mediating integrin activation, we deleted TBS1, TBS2, or both (ΔTBS1, ΔTBS2, and ΔTBS1-2) in RIAM and assessed their effects on integrin activity when co-expressed with full-length talin. Deletion of TBS1 and TBS1-2 leads to significant loss of integrin activity, whereas the effect of ΔTBS2 is much weaker on altering integrin activity (Fig. 4D). Furthermore, TBS1-CAAX and TBS1-2-CAAX, but not TBS2-CAAX, are capable of promoting the inside-out integrin activation (Fig. 4E). Together, our results suggest that binding determinants in the TBS1:R7R8 complex and the helical kink in the RIAM TBS1 are required for integrin activation, and TBS1, but not TBS2, is essential for talin recruitment in inside-out integrin signaling.

Substitution of RIAM TBS1 with corresponding residues in Lpd reduces talin binding and impairs integrin activation

RIAM and Lpd affect cell adhesion differently despite their similar structural architecture with 59% sequence identity in the TBS1-2 and the RA-PH regions (Krause et al., 2004; Lafuente et al., 2004). Lpd has been identified as an M-Ras effector protein but retains a moderate Rap1-binding affinity owing to an RA-PH functional module highly similar to that of RIAM (Lafuente et al., 2004; Tasaka et al., 2012; Yoshikawa et al., 2007; Zhang et al., 2014). Lpd also possesses all R8-interacting residues in RIAM TBS1 in its N-terminal talin-binding site (Lpd-TBS) except a single substitution of Leu15^RIAM with Trp31^Lpd (Fig. 5A). Interestingly, the Lpd-like L15W mutant retains substantial binding capacity to talin compared with the L15Y mutant (Fig. 5B). However, the association of RIAM TBS1 with cytoplasmic talin is much stronger than that of Lpd-TBS and talin (Fig. 5C). This result is also confirmed by an in vitro pull-down using a purified talin protein containing R7, R8, and R9 domains (R7R8R9) (Fig. S3A). Furthermore, the effect of Lpd-TBS-CAAX on promoting integrin activity is also much weaker than that of RIAM TBS1-CAAX (Fig. 5D). These results suggest that Lpd-TBS lacks other crucial factors for a tight talin association. We have shown that the Ser13 residue in RIAM stabilizes the helical kink in the R8-bound TBS1. This residue is substituted by a glycine residue in Lpd. The TBS1 fragment (residues 5-25) alone does not possess a stable secondary structure, and the S13G mutation does not alter the conformation of the isolated TBS1 fragment (Fig. S3B). However, the glycine residue is unable to form the hydrogen bond network that stabilizes the helical kink in the R8-bound TBS1. To test whether the lack of a stable helical kink leads to the weak association of talin with Lpd-TBS, we produced a gain-of-function mutation by mutating the corresponding Gly29 residue to serine in the Lpd-TBS fragment. Indeed, Lpd-TBS-G29S enhances the interaction of Lpd-TBS and talin (Fig. 5C). This mutation also improves integrin activation mediated by Lpd-TBS (Fig. 5D). Together, these results suggest that talin preferentially binds to RIAM over Lpd via a kinked helical configuration and favorable side chains of the TBS1 site.

Plasmid construction and protein purification

The TALIN rod domains were cloned into the modified pET28a expression vector with a His₆-tag and a TEV protease cleavage site. The RIAM talin-binding sites were cloned into the pGEX-5X-1 expression vector with a GST-tag. RIAM talin-binding site 1 (TBS1) consists of residues 1 to 30 and talin-binding site 2 (TBS2) consists of residues 27 to 93. Talin and RIAM point mutations were constructed using a site-directed mutagenesis method. Plasmids were transformed into Escherichia coli BL21(DE3) for protein expression. Cells were grown in LB medium containing kanamycin or ampicillin at 37°C until the A₆₀₀ reached 0.6-0.7, and induced by adding 0.4 mM IPTG with incubation continued overnight at 20°C for talin and for 3 hrs at 37°C for RIAM. Protein purification was carried out at 4°C. The cell pellets were resuspended in 20 mM Tris pH 7.5 with 500 mM NaCl for His-tagged proteins and 50 mM Tris pH 7.5, 150 mM NaCl, and 5 mM DTT for GST-tagged proteins. Cells were then lysed with an EmulsiFlex-C3 homogenizer (AVESTIN, Inc.). Protein samples were extracted from the supernatants using either a HisTrap FF or GSTrap FF column (GE Healthcare). For crystallization, the His-tag was removed via incubation with TEV protease and then further purified using a HiTrap Desalting column and a HisTrap FF column (GE Healthcare).

GST pull-down and Western blotting

For in vitro pull-down assays, purified GST-RIAM proteins were immobilized on glutathione agarose beads and then incubated with purified His-tagged talin proteins in reaction buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, and 2 mM DTT) to a total volume of 250 μl on a rotator for 1 hr at 4°C. Binding curves were fit to a single-site (saturating) binding model using SigmaPlot (Systat Software). For cell lysate pull-down assays, HEK 293T cells were transfected with various GFP-talin constructs using Lipofectamine 2000 reagent (Invitrogen) and lysed 48 hrs post-transfection with lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 1% (v/v) Triton X-100, 1 mM DTT, and complete protease inhibitor cocktail (Roche)). The cell lysates (400 μg per reaction) were clarified by centrifugation, and supernatants were mixed with the purified GST-RIAM proteins immobilized on 50 μl of 50% slurry of glutathione agarose beads and incubated on a rotator for 1 hr at 4°C. The bound proteins were washed three times in 500 μl of the lysis buffer and were eluted using an elution buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 2 mM DTT, and 10 mM reduced glutathione) at 4°C. The proteins were resolved by SDS-PAGE and detected by Coomassie staining or Western blotting. The Immobilon-P transfer membranes (EMD Millipore) were blocked with TBST buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1% (v/v) Tween 20) containing 5% (w/v) BSA for 1 hour and then incubated with anti-His (Sigma) or anti-GFP antibody (Clontech) for 1 hour at room temperature followed by a second incubation with HRP-conjugated secondary antibody (Santa Cruz Biotechnology). The blots were visualized with the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific) and detected using the FluorChem E imager (ProteinSimple).

X-ray crystallography

The RIAM peptide consisting of residues 5 to 25 (Genemed Synthesis, Inc.) was dissolved in 100 mM Tris pH 8.0 with 100 mM NaCl and the final pH was adjusted to pH 7-8. Talin R7R8 (11 mg/mL) was incubated with the RIAM peptide on ice at a 1:3 molar ratio prior to the crystallization setup. The talin R7R8 domains-RIAM peptide complex was crystallized using the hanging-drop vapor diffusion method at room temperature, with the droplets containing equal volumes of protein and reservoir solution. The crystals were grown in 100 mM NaCl, 20% (w/v) polyethylene glycol 3350 and 20% (v/v) ethylene glycol from microseeding. The crystals were grown for 2-3 days and then flash-frozen in liquid nitrogen prior to the diffraction experiments.

Final X-ray diffraction data were collected using beamline X29 of the National Synchrotron Light Source at the Brookhaven National Laboratory (Upton, NY). Data were processed using the HKL-2000 package (Otwinowski and Minor, 1997). The complex structure of the talin R7R8 domains and the RIAM TBS1 was determined by molecular replacement using talin rod residues 1359-1659 (PDB ID 2X0C) as the initial search model (Gingras et al., 2010). Structural refinement was performed using REFMAC (Murshudov et al., 1997) and Coot (Emsley and Cowtan, 2004). The models were validated with SFCHECK (Vaguine et al., 1999) and MolProbity (http://molprobity.biochem.duke.edu/). Data collection and refinement statistics are listed in Table 1. Structural figures were generated using the PyMOL program suite (http://pymol.sourceforge.net). The final atomic model contains residues 1357-1651 of chain A (talin) and residues 5-25 of chain B (RIAM). The atomic coordinates and structure factor have been deposited in the Protein Data Bank under the accession number 4W8P.

Confocal imaging & Pearson’s correlation coefficient analyses

Co-clustering of RIAM and talin was performed as previously described (Zhang et al., 2014). Briefly, CHO cells stably expressing αIIβ3 integrins (A5 cells) were co-transfected with mCherry-talin and either GFP-RIAM-TBS1-CAAX or the assorted mutants. 24 hrs post co-transfection, the cells were allowed to adhere and spread on glass cover slips coated with 10 μg/mL fibrinogen for 1 hr at 37°C and then fixed with 4% (w/v) paraformaldehyde for 30 mins at room temperature. The fixed cells were washed with PBS buffer and mounted with ProLong Gold antifade reagent (Molecular Probes). Images were acquired using a 60X 1.45 Pan APO TIRF oil immersion objective using a Nikon TE-2000U microscope (Optical Apparatus Co., Ardmore, PA) equipped with the Ultraview spinning-disc confocal (Perkin-Elmer Life Sciences, Boston, MA). Using Velocity 6.3, the software was instructed to acquire double channel simultaneous sequential images at 488nm (GFP) and 568nm (mCherry) wavelengths. Image files were captured as 0.1 μm-thick Z-slices and processed to obtain maximum projections. Ten cells per condition were analyzed corresponding to three independent experimental repetitions. Pearson’s correlation coefficients (PCCs) were obtained from whole cells as well as lamellar locations using MetaMorph offline 7.8 software (Molecular Devices, Downingtown, PA). Cellular lamellae areas were selected using RIAM’s GFP intensities to detect the cell edge, while cellular centroids were identified using the same MetaMorph software. The outermost 5 microns measured from the cell’s centroid constituted lamella positive areas.

Integrin activation assay

A5 cells were co-transfected with HA-talin and GFP-tagged RIAM or Lpd vectors. 24 hrs after transfection, the cells were resuspended by trypsinization and washed with Tyrode’s buffer (136.9 mM NaCl, 10 mM HEPES, 5.5 mM glucose, 11.9 nM NaHCO₃, 2.7 mM KCl, 0.5 mM CaCl₂, 1.5 mM MgCl₂, and 0.4 mM NaH₂PO₄, pH 7.4). PAC-1 mAb (Becton Dickinson Immunocytometry System), which specifically recognizes activated integrin, was then added to the cell suspension and incubated for 1 hr at 4°C. The cells were washed by Tyrode’s buffer and incubated with Alexa Fluor 647 conjugated goat anti-mouse IgM antibody for 1 hr on ice and resuspended in cold PBS. To specifically inhibit integrin activation, 10 mM EDTA or 10 μM Eptifibatide was incubated with cell suspension for 20 mins before adding PAC1 antibody. Stained cells were analyzed using the LSRII FACS instrument (BD Scientific). The collected data were processed using the FlowJo software package and expression levels of talin were examined by Western blot (Fig. S5). Only GFP positive cells were gated on to analyze PAC-1 binding. The mean fluorescence intensity (MFI) of PAC-1 channel of each transfectant was normalized to that of the GFP empty vector and HA-talin co-transfection to obtain the relative MFI. The data were represented by the means ± SD from at least three experiments. The equal expression of talin under each condition was verified by Western blot using anti-HA antibody (Fig. S5). The unpaired t test was performed to calculate the P value using GraphPad software.

CD spectroscopy

CD spectra were recorded in a 0.1-cm pathlength quartz cell on a CD spectrometer (model 62A DS; Aviv Associates, Lakewood, NJ). Proteins (R2R3 and R7R8) or peptides (TBS1 and TBS1-S13G) were measured at 0.5 mg/mL concentration. Spectra were recorded at 25°C over a wavelength of 200–300 nm.