Introduction

Talin was discovered nearly three decades ago as a highly abundant cytosolic protein important for cytoskeleton organization and cell-extracellular matrix (ECM) adhesion1. Extensive genetic and cell biological studies have established that talin is crucial for regulating a wide variety of integrin-mediated cell adhesion-dependent processes, such as cell-shape change, growth, differentiation, and migration2,3,4. Talin is large in size, with 2541 amino acids, and can be divided into two major segments, an N-terminal head (1-433, talin-H, 50 kDa) that contains a FERM (four-point-one-protein/ezrin/radixin/moesin) domain (86-400, talin-FERM) and a C-terminal rod (482-2541, talin-R, 220 kDa) that contains a series of consecutive helical bundles followed by an actin-binding motif2,3,4. Talin-FERM, which engages with heterodimeric (α/β) integrin adhesion receptors, can be further divided into F1, F2, and F3 subdomains, with F3 specifically interacting with integrin β cytoplasmic tails (CTs)2,3,4. Because of its capacity to bind both integrin and actin, talin has long been recognized as a mechanical linker between the ECM and actin cytoskeleton to regulate cell adhesion and morphology5. However, about a decade ago, it was found that the binding of talin-FERM to integrin β CTs can also promote the conversion of integrins from a low-affinity into a high-affinity ligand-binding state, a dynamic process termed integrin inside-out signaling or integrin activation6,7,8,9. This finding suggests that the talin-FERM/integrin interaction plays a multifunctional role in promoting integrin activation as well as integrin-actin coupling during dynamic cell adhesion processes10,11,12. Structural and biochemical analyses have shown that the talin-FERM binding to integrin β CT triggers the separation of a key integrin α/β CT clasp, thus facilitating the global conformational change and activation of the receptor7,8,13. Stretches of positively charged surfaces on F1, F2, and F3 were also found to be crucial for this process by targeting talin-FERM to the membrane and enhancing the talin-FERM/integrin β CT interaction13,14,15,16,17. However, how the membrane-targeting and integrin binding of talin-FERM are spatiotemporally regulated to control the dynamics of integrin adhesion is not understood. This issue is important, since uncontrolled talin activity may lead to dysfunction of integrins, which is linked to many human disorders, including thrombosis, stroke, bleeding, infections, and cancer metastasis18. Talin can adopt inactive and active states during dynamic cell adhesion events, but the underlying molecular basis for this transition is elusive2. Previous studies suggested that the inactive talin has an autoinhibited conformation, where a key integrin-binding site on talin-F3 is self-masked by a talin-R segment (talin-RS)19,20. Membrane lipid phosphotidylinositol-4,5-bisphosphate (PIP2)19,21, which is locally enriched by talin-recruited PIPKIγ kinase22,23, was shown to activate talin19,21 and promote integrin-mediated cell adhesion15,24,25,26,27,28, but the detailed structural basis of this is not clear. The prevailing hypothesis is that PIP2 may sterically induce the conformational change of talin2,15,16,19,21, a mechanism widely employed in the known PIP2-mediated activation of proteins, including FERM domain proteins29,30.

Here we describe the first high-resolution crystal structure of the principal autoinhibitory unit of talin. The structure reveals an unexpected dual inhibitory topology in which talin-RS not only sterically masks the integrin-binding site on talin-FERM via one large interface, but also electrostatically hinders the membrane-targeting of talin-FERM via another extensively negatively charged surface. Such a topology is consistent with the biochemical and functional data, but strikingly different from a previous model20. The topology further suggests a “pull-push” mechanism for talin activation by membrane enriched with PIP2, which differs distinctly from the classic steric-clash mechanism for the conventional PIP2-mediated activation of FERM proteins activation. We verify this “pull-push” mechanism by a combination of nuclear magnetic resonance (NMR), biochemical, and functional experiments. Our data lay down a new foundation for understanding talin inhibition and activation through novel regulatory mechanisms. They also provide significant insight into how the changes of the membrane surface promote an on/off switch of a cytosolic protein to control the receptor transmembrane signaling – an emerging area in signal transduction that is poorly explored30.

Results

Crystal structure of autoinhibited talin

Previous structural mapping experiments have shown that an interdomain complex between talin-F3 and the C-terminal rod fragment (1654-1848; talin-RS) represents the principal structural unit for talin autoinhibition19. Talin-R 1984-2344 fragment was also found to interact with talin-F3, but at a low affinity (∼20-fold weaker than the talin-RS/talin-F3 interaction) unrelated to talin autoinhibition19. To obtain a high-resolution view of the autoinhibited talin, we screened multiple constructs related to the talin-RS/talin-F3 complex for crystallization, among which talin-F2F3 (206-405) in complex with talin-RS was eluted by gel filtration and crystallized, allowing us to solve its structure (Figure 1A) at 2.0 Å resolution (see structural statistics in Table 1). Talin-F2F3 and talin-F3 bind to talin-RS identically as judged by NMR, indicating that F2 does not affect the talin-F3/talin-RS interface (Supplementary information, Figure S1A). In the complex, talin-F2F3 and talin-RS exhibit the same fold as their free forms20,31, but pack extensively with a buried interface of 1750 Å2 where the five-helix bundle of talin-RS is held in place by two protruding fingers of talin-F3 (Figure 1A). Superposition of talin-F2F3 bound to talin-RS with that bound to the integrin β1 CT13 (PDB code 3G9W) shows that a major portion of membrane-proximal region but not the membrane-distal region of integrin β1 CT is sterically blocked by the talin-F2F3/talin-RS interface (Figure 1B). The autoinhibitory interface is largely hydrophilic, but also contains hydrophobic interactions as summarized in Figure 1C. Notably, (i) the carboxyl group of E1770 of talin-RS forms a buried salt-bridge with K318-NH3+ and a strong hydrogen bond with the backbone of G321, and van der Waals contact with the backbone of M319 of talin-F3; (ii) the hydroxyl group of T1767 hydrogen bonds with K318 side chain -NH3+, N323 side chain -NH2, and backbone carbonyl of K322, respectively, and the methyl group of T1767 also makes van der Waals contacts with G321 and N323, respectively; (iii) the side chain of M1802 exhibits hydrophobic interactions with the methyl group of M319 and the hydrophobic part of K322 side chain; (iv) the carboxyl group of D1809 hydrogen bonds with the carboxamide of N323; (v) the carboxamide of Q1677 hydrogen bonds with the hydroxyl group of Y377; (vi) the carboxamide of N1669 hydrogen bonds with the carboxyl group of D375; (vii) the carboxyl group of D1676 hydrogen bonds with the backbone amide of Y377; and (viii) the side chains of V1683 and L1680 exhibit hydrophobic interactions with A360.

Figure 1
figure 1

Talin-RS has a dual inhibitory role on talin-F2F3. (A) Crystal structure of talin-F2F3/talin-RS complex. The individual subunits are labeled in green for talin-F2F3 and blue for talin-RS. Arrows indicate the two fingers in talin-F2F3 grabbing talin-RS. (B) Superimposition of talin-F2F3 bound to talin-RS with that bound to β1D CT (PDB code 3G9W) showing that the membrane proximal (MP, salmon) region, but not the membrane distal (MD, red) region in β1D CT is buried in the talin-F2F3/talin-RS interface. The arrow indicates the significant movement of K318-K324 loop in talin-F3 from the inactive state (bound to talin-RS) to active state (bound to β1D CT). (C) Close-up view of talin-F2F3/talin-RS interface shown in Figure 1A. Left panel shows the amino acid side chains involved in upper finger region (K318-K324) and right panel shows the ones in bottom finger region (D369-Y377) and the β5 and β6 regions. Interface residues from talin-F3 are colored in pink and labeled in black; interface residues from talin-RS are colored in orange and labeled in red; non-interface residues K324 and L325 in talin-F3 are colored in cyan. (D) Electrostatic surface representation of the complex viewed from the top (left panel) and from the side (right panel). The blue area indicates positive charge enriched region and red area indicates negative charge enriched region. It is clear that the extensively negatively charged surface on talin-RS would disfavor the talin association with membrane (right panel).

Full size image
Table 1 Talin-F2F3/talin-RS complex diffraction data and refinement statistics
Full size table

The autoinhibitory interface of the crystal structure is consistent with the previously reported result that the L325R mutation had no effect on the complex formation19, since L325 is not in the complex interface (Figure 1C). It is also consistent with extensive loss-of-function mutagenesis data19,20, including mutations of M319A, S365D, S379R, Q381V, T1767A/D/E, and E1770A, which all perturb the interface environment. Specifically, M319A, T1767A/D/E, and E1770A directly disrupt the interface based on the above described interactions. The S365D mutation reduces the talin-RS/talin-F3 interaction19, because it introduces a negatively charged group that would interact unfavorably with L1680 and V1683. Similarly, S379R and Q381V mutations introduce side chains with different properties, thereby perturbing the surrounding interface and reducing the talin-RS/talin-F3 binding19. The effect of the M319A mutation in relation to the structure is particularly notable, since this mutation was shown to constitutively activate talin19. On the basis of our structure (Figure 1C), the mutation would impair a hydrophobic network of the M319 side chain with M1802 and the hydrophobic portions of K322 and K324 side chains (Figure 1C). This network is crucial for stabilizing the K318-K324 finger conformation that accounts for ∼50% of the autoinhibitory interactions. Indeed, the position of the K318-K324 loop is significantly shifted in the autoinhibited form as compared to the active integrin-bound form (see the arrow in Figure 1B), indicating different conformations of the finger at different states. In Supplementary information, Figure S1, we further confirmed experimentally that the M319A mutation (Supplementary information, Figure S1B) or the K322A/K324A mutation (Supplementary information, Figure S1C) significantly impaired talin-F2F3 binding to talin-RS.

The crystal structure is drastically different from the previously reported NMR-based model that was primarily built by the HADDOCK program using ambiguous chemical shift constraints20. Overlay of talin-F3 between the crystal structure and the HADDOCK model reveals that the bound talin-RS in the crystal structure is rotated nearly 90° versus the model (Supplementary information, Figure S1D). The interface of the HADDOCK model would cause steric clash with PIPKIγ (Supplementary information, Figure S1E, right panel), which is incompatible with the previous competition experiment showing that PIPKIγ had little effect on the talin-F3/talin-RS complex formation19. By contrast, the competition data are fully consistent with the crystal structure which shows no steric clash with PIPKIγ (Supplementary information, Figure S1E, left panel). The interface of the HADDOCK model is also inconsistent with the mutagenesis data. Notably, L325 is in the core interface of the HADDOCK model (Supplementary information, Figure S1F)20, but its mutation to Arg had no effect on the talin-F3/talin-RS complex formation19. Also, M319 is not in the complex interface of the HADDOCK model and does not form any interaction network20 (Supplementary information, Figure S1F), which fails to explain why its mutation to Ala weakens the talin-F3/talin-RS interaction (Supplementary information, Figure S1B) and constitutively activates talin19. These data together indicate that the crystal structure represents a physiologically relevant form of inactive talin. The crystal structure also illustrates clearly how the inactive talin has a low capacity to bind and activate integrin (Figure 1B), and provides a template for understanding how talin may be activated (see below).

Talin-RS exhibits a negatively charged surface that electrostatically inhibits the talin-F2F3 binding to membrane

As mentioned above, talin-FERM contains a stretch of positively charged surface that associates with membrane to promote the talin-F3/β integrin-MPCT interaction13,14,15,16,17. Figure 1D (left panel) provides the detailed illustration of the positively charged surface in talin-F2F3, which mainly involves K268, K272, K274, R277, and K278 in talin-F2, and K322 and K324 in talin-F3. This positively charged surface appears to play a dominant role in the membrane anchoring of talin-FERM14,15,28. Interestingly, careful examination of the talin-F2F3/talin-RS complex structure revealed a large negatively charged surface on talin-RS, which is located at the same side as the positively charged surface of talin-F2F3 (Figure 1D, left panel). This finding immediately suggested to us that the membrane binding of talin-F2F3 in the latent talin would be inhibited by the negatively charged surface of talin-RS via charge-charge repulsion (Figure 1D, right panel). This prediction was readily verified by the NMR-based competition experiments showing that while free talin-F2F3 binds to large uniamellar vesicles (LUV) containing the negatively charged lipid, 1-palmitoyl-2-oleoyl-sn-glycero-3-[phospho-L-serine] (POPS; Figure 2, left panel), the binding is significantly inhibited upon addition of talin-RS (Figure 2, right panel). Since talin-RS does not bind LUV (data not shown), the only possibility for such inhibition to occur, based on our structure, is through charge-charge repulsion between talin-RS and the membrane when talin-RS is bound to talin-F2F3 (Figure 1D, right panel). Thus, the autoinhibited talin-F2F3/talin-RS complex has a unique conformation that is sterically inhibited for binding to integrin (Figure 1B) and electrostatically hindered for binding to membrane (Figure 1D, right panel).

Figure 2
figure 2

The autoinhibited talin is unfavorable to bind membrane. 2D-HSQC of 0.04 mM 15N-labeled talin-F2F3 (left) or 0.04 mM 15N-labeled talin-F2F3 plus 0.08 mM unlabeled talin-RS (right) in the absence (black) and presence (red) of 2.4 mM 4:1 POPC:POPS LUV. Significant line broadening (as reflected by the decreased peak intensity) occurred for many signals (see representative insert panel) when negatively charged vesicle was added to free talin-F2F3 (left panel). Such line broadening effect was dramatically reduced (right panel) when the same vesicles were added to talin-F2F3/talin-RS complex, indicating talin-RS effectively inhibits the access of talin-F2F3 to membrane.

Full size image

PIP2 binds to talin-F2F3 in a distinct bivalent mode

If the autoinhibited talin is unfavorable for membrane binding, how would talin be activated by PIP2 that binds to talin-F2F315,19,21? To understand this process, we first used NMR to compare the specificities of talin-F2F3 binding to LUV enriched with POPS, PIP2, and POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine), all of which are found on the inner surface of plasma membrane. Substantial line-broadening of the NMR resonances of 15N-labeled talin-F2F3 occurred upon adding POPS or PIP2 LUV, whereas much less change was detected when the neutral POPC was added (Supplementary information, Figure S2A). The line-broadening effects in talin-F2F3 induced by all three lipids was concentration-dependent with PIP2 exerting the most dramatic effect, that is, showing the steepest signal intensity drop at the same protein-lipid ratio (Supplementary information, Figure S2B). Thus, PIP2-containing LUV exhibit the strongest binding to talin-F2F3, which is consistent with the higher negative charge on its head group than POPS and POPC.

Next, we asked structurally how PIP2 recognizes talin-F2F3. NMR-based HSQC spectra were collected for 15N-labeled talin-F2F3 in the absence and presence of increasing concentrations of inositol-1,4,5-trisphosphate (IP3), the head group of PIP2, to obtain molar ratios of 1:0.0, 1:5.0, and 1:20.0. To examine the effect of negative charges, we also included inositol-1,3,4,5-tetraphosphate (IP4), the head group of PI(3,4,5)P3, at the same molar ratios. Figure 3A and 3B show the detailed chemical shift perturbation profiles of 15N-labeled talin-F2F3 obtained with IP3 and IP4, respectively. Both profiles revealed a major perturbation region (K268∼K278) in F2 that is rich in positively charged amino acids, consistent with previous functional analysis15. Another positively charged region (K318∼K324) in F3 is also perturbed, but less extensively. The effect of IP4 was more pronounced than that with IP3, presumably due to the additional negatively charged phosphate group. Both IP3 and IP4 are quite selective, because, under the same conditions, neither perturbed talin-F0F1 (Supplementary information, Figure S3A), which also has positively charged surface for potential interaction with membranes16. This specificity is consistent with the talin-F2F3 being the major fragment for the lipid binding14. To determine the lipid-binding sites on talin-F2F3 more precisely, we examined the IP4 binding of various talin-F2F3 mutants with mutations at sites that are positively charged and perturbed by lipid head groups, including K272A-K274A (on F2) and K322A-K324A (on F3). As a control to examine specificity, we also mutated a positively charged site (K254A-K256A on F2) that is not significantly perturbed. Interestingly, compared to wild-type talin-F2F3, K272A-K274A and K322A-K324A exhibited markedly reduced perturbation at the mutated sites, but little change at the non-mutated sites, indicating that these two sites are independent, each binding to a lipid molecule (Figure 3C and 3D). This interpretation is supported by the isothermal titration calorimetry (ITC) experiment, where talin-H was found to bind PIP2 at a 1:2 ratio (Supplementary information, Figure S3B) and NMR data showing that IP3 binds to these talin-F2F3 mutants in similar patterns as IP4 (Supplementary information, Figure S3C and S3D). As a comparison, the control mutant, K254A-K256A, showed comparable binding to IP4 (Figure 3E) or IP3 (Supplementary information, Figure S3E) as wild type with similar chemical shift perturbation profiles. This indicates that the K254-K256 site is not geometrically optimized for interacting with the lipid head group despite being positively charged. The bivalent binding mode of PIP2 to talin-F2F3 is consistent with the talin-F2F3 structure where (i) the closest distance between the two sites is at least 23 Å that is much larger than a single PIP2 binding pocket (4-11 Å); and (ii) both sites point towards the membrane, allowing the binding of two PIP2 (Figure 1D, right panel). Since talin-F2 is perturbed more substantially than talin-F3 by PIP2 (Figure 3A), it is possible that the former binds more tightly to the lipids than the latter, which is consistent with the ITC-based affinity estimation for PIP2 with KD1 at ∼0.4 μM and KD2 at ∼5 μM, respectively (Supplementary information, Figure S3B). The different affinities suggest that the two PIP2 molecules bind talin in a progressive manner in which talin-F2 with the higher affinity first anchors to PIP2 and then promotes the second PIP2 binding to talin-F3.

Figure 3
figure 3

Mapping of PIP2 and PIP3-binding sites on talin-F2F3 using the corresponding head groups IP3 and IP4, respectively. 1H/15N chemical shift changes of 0.04 mM talin-F2F3 in the presence of (A) 0.8 mM IP3 or (B) 0.2 mM IP4. Both ligands reveal two major similarly perturbed regions with IP4 having a stronger effect likely due to additional negatively charged phosphate group. (C-E) Perturbation profiles of 0.04 mM K272A/K274A (C), 0.04 mM K322A/K324A (D), and 0.04 mM K254A/K256A (E) talin-F2F3 in the presence of 0.2 mM IP4. We note that the N-terminal region of talin-F2F3 is also somewhat perturbed, which is insensitive to these mutations and thus the perturbation is probably due to some secondary effect.

Full size image

Membrane enriched with PIP2 activates talin via a pull-push mechanism

With the above data showing that PIP2 binds bivalently to talin-F2F3, we next set out to determine how such binding induces talin activation. As mentioned above, the existing hypothesis is that membrane containing PIP2 might sterically induce a conformational change of talin by binding to talin-H2,15,16,19,20,21, a scheme that is typically observed in other PIP2-binding proteins, including FERM proteins such as moesin29,30. However, the bipolar membrane-facing surface in our autoinhibited talin structure with the exposed PIP2-binding epitope (Figure 1D) suggests a rather different mechanism, which involves a “pull-push” process to relieve the dual inhibition of talin. In this process, the locally enriched PIP2 strongly “pulls” the positively charged surfaces of talin-F2 and F3 to the membrane surface and simultaneously “pushes” down the negatively charged surface of talin-RS (Figure 4A). The opposite movement of the two subunits would clearly cause the mismatch of the interface and the subsequent disruption of the talin-F3/talin-RS complex (Figure 4A). To test this hypothesis, we performed multiple independent experiments. First, we performed NMR-based competition experiment. Comparison of signals obtained for the free talin-RS (black) with those of the talin-F2F3-bound form (blue) shows that a significant number of signals were changed or broadened upon addition of talin-F2F3 (Figure 4B and Supplementary information, Figure S4A), indicative of the interaction between the two fragments. However, when PIP2-enriched LUV were introduced to this complex, essentially all the perturbed talin-RS signals either reappeared or returned towards the positions in free form (red spectrum in Figure 4B and Supplementary information, Figure S4A), thus providing evidence that PIP2 “pulled” talin-F2F3 to membrane surface, and simultaneously “pushes” away talin-RS to induce the mismatch and disruption of the talin-F3/talin-RS interface. By contrast, LUV containing the same percentage of POPS as PIP2 failed to exert the “pull” effect (Supplementary information, Figure S4B) due to its weaker binding to talin-F2F3 than PIP2. Second, to further evaluate this pull-push process, we repeated the experiments with the K272A/K274A mutant that loses the PIP2 binding in talin-F2, but retains the PIP2 binding capacity in talin-F3 and the talin-RS/talin-F2F3 interaction (Supplementary information, Figure S4C; note that the mutation site is far away from the talin-RS/talin-F3 autoinhibitory interface, see Figure 4A). As shown in Figure 4B, the same PIP2 LUV had a substantially reduced capacity to disrupt the talin-RS/talin-F2F3 mutant interaction (green), which is consistent with the weakened “pull” mechanism. Finally, to gain cellular functional evidence for the pull-push mechanism, we made a cluster of five charge-reversal mutations on talin-RS (E1714K/E1794K/E1797K/E1798K/E1808K, 5EtoK; see Figure 1D, left panel) alone and in combination with the constitutively active talin mutant M319A19 (M319A/5EtoK). The charge-reversal mutation sites are outside the autoinhibitory interface, but the mutations introduced would significantly reduce or reverse the negative charge on talin-RS, thereby facilitating the talin recruitment to membrane (Supplementary information, Figure S5, left panel) – a process that is inhibited for the WT talin (Figure 4A, left panel). Furthermore, upon the recruitment of talin-RS mutant (5EtoK) to the membrane surface, the positively charged surface of talin-F2F3 would further move upward by ∼7 Å to touch the membrane, because the membrane-facing plane of talin-F2F3 lags behind that of talin-RS based on our structure (Supplementary information, Figure S5). Such upward movement would also lead to the interface mismatch (Supplementary information, Figure S5, right panel) in a way that mimics the intermediate step of the PIP2-mediated pull-push process in which talin-F2F3 moves upward to the membrane while talin-RS is being pushed down (Figure 4A, right panel). As expected, Figure 4C shows that the 5EtoK mutant is more active than WT talin, but less active than the M319A/5EtoK mutant. The stronger effect of M319A/5EtoK is clearly due to the additional effect of M319A that disrupts the autoinhibitory interface (Figure 1C and Supplementary information, Figure S1B). Thus, we have obtained combined structural, biochemical, and functional evidence to strongly support the membrane-dependent mechanism of talin activation.

Figure 4
figure 4

Membrane enriched with PIP2 favorably interacts with talin-F2F3 and promotes talin unmasking via a pull-push mechanism. (A) A structural view to show how unstimulated membrane surface would repel the latent talin (left panel), whereas PIP2-enriched membrane lipids strongly “pulls” up the talin-F2F3 and “pushes” down talin-RS, leading to the interface mismatch and unmasking of the integrin site on talin-FERM (right panel). The blue area indicates region enriched with positive charge and red area indicates region enriched with negative charge. (B) Representative HSQC region of 0.06 mM 13C-15N-labeled talin-RS in the absence (black) and presence of 0.05 mM talin-F2F3 (blue) or 0.05 mM talin-F2F3 and 3 mM 4:1 POPC:PIP2 LUV (red) or 0.05 mM talin-F2F3 K272A/K274A mutant and 3 mM 4:1 POPC:PIP2 LUV (green). The arrows show that the addition of PIP2 vesicle competes effectively with talin-RS to talin-F2F3, but much less effectively with the talin-F2F3 K272A/K274A mutant. (C) Comparison of the activation of integrin αIIbβ3 by vector control, full-length talin (Talin), full-length talin E1714K/E1794K/E1797K/E1798K/E1808K (Talin 5EK), full-length talin M319A+5EK (Talin M319A-5EK), and full-length talin M319A (Talin M319A). Both Talin 5EK and Talin M319A are more active than Talin through different mechanisms (see the text). As expected, the combination of M319A and 5EK is more active than the individual ones (P < 0.007 for Talin vs Talin 5EK, P < 0.001 for Talin vs Talin M319, and P < 0.002 for Talin 5EK vs Talin M319A-5EK). Note that WT talin has some basal activity as shown before19, consistent with that there is an equilibrium between inactive and active talin. The data are representative of seven experiments.

Full size image

Discussion

Integrin activation and its coupling to actin are vital to almost all cell-ECM-dependent physiological and pathological processes, such as blood coagulation, tissue remodeling, and cancer metastasis10,11,12. The importance of talin in this dynamic process is undisputed based on the large amount of literatures over the decades5,10,11,12, but how talin is spatiotemporally turned on and off has remained elusive despite past significant efforts2. The crystal structure of the talin-F2F3/talin-RS complex now provides the first definitive view of the “off” state of talin, showing how the integrin binding to talin-FERM is sterically blocked by talin-R (Figure 1B). Moreover, the structure also reveals a previously unrecognized charge-charge repulsion mechanism via talin-RS that electrostatically inhibits the talin-FERM binding to membrane (Figure 4A, left panel). Since the membrane-binding of talin-FERM is essential for its access to integrin β CT membrane-proximal region and for enhancing the talin-FERM/integrin interaction13,14,15,16,17, the repulsion mechanism by talin-RS may represent an effective way to negatively regulate this targeting process, which in turn promotes the cytosolic retention of the latent talin. A recent report suggested that talin-R 482-787 fragment (talin-R1) may also interact with talin-F2F3 and inhibits the talin-FERM/membrane interaction32. However, as compared to the potent talin-RS/talin-F2F3 interaction, the binding of talin-R1 to talin-F2F3 appears to be very weak or non-specific, since no significant chemical shift changes were observed even when talin-R1 was in excess19,32 except for some NMR line-broadening that occurred when even more excess talin-R1 was added to 15N-labeled talin-F2F332. Further investigation such as obtaining the structures of full-length talin and talin-binding partners is clearly necessary to examine the existence of additional mechanisms that stabilize retention of the latent talin in the cytosol. Nevertheless, our results here have unraveled a distinct regulatory process where the binding of talin-FERM to membrane and integrin are dually inhibited by talin-RS. Such dual inhibition likely allows for a tight control of talin activity in physiological settings to minimize spontaneous integrin activation that otherwise could result in severe pathological states, such as thrombosis and tumor progression.

Another significant finding of our study is the mechanism of talin activation by PIP2. Compared to other phosphoinositides that may also bind to talin14,21, PIP2 is locally enriched by talin-recruited PIPKIγ kinase in vivo22,23 and is thus a physiologically important ligand for talin to promote integrin adhesion15,21,25. This selectivity is supported by the recent genetic studies where specific ablation of PIPKIγ in focal adhesions significantly impaired the talin recruitment to membrane and adhesion sites28. Using NMR, we have defined here a bivalent binding mode of PIP2 to talin F2F3, which is different from the classical phosphoinositide binding mode to other known modules33. The bivalent PIP2 sites on talin-F2 and F3 are also distinct from the previously reported monovalent PIP2 site in radixin that is located in a groove between F1 and F334, but would promote strong interactions of the membrane with talin-FERM. Thus, PIP2-binding modes to FERM domain are versatile. We then showed that the PIP2-enriched membrane attracts talin-F2F3 and simultaneously repels talin-RS, leading to the relief of the dual inhibition of talin-FERM via an electrostatic pull-push mechanism. This pull-push process differs from the previously proposed steric-clash mechanism2,15,16,19,20, and the latter is well-known for conventional PIP2-induced activation of proteins including FERM domain proteins moesin34 and radixin35, where the PIP2-binding site is buried and can be unmasked by PIP2 via steric occlusion29,30. Based on these observations, we envision a cascade of molecular events for the membrane-mediated spatiotemporal regulation of talin-FERM inhibition and activation (Figure 5): (a) In the unstimulated cells, inactive talin is maintained in the cytosol where its FERM domain is unable to bind membrane and integrin. (b) Upon cellular stimulation, PIPKIγ is recruited by talin, and promotes local PIP2 enrichment. Note that based on enzyme chemistry, only a tiny amount of PIPKIγ is necessary to produce large quantity of PIP2 locally once it associates with some fraction of the highly abundant talin in cells. (c) The membrane with enriched PIP2 recruits talin by strongly “pulling” talin-FERM to the membrane surface and “pushing” away talin-R. This then results in the unmasking of the integrin-binding site on talin-FERM and subsequent talin/integrin interaction and integrin activation. The stronger affinity of PIP3 for talin-F2F3 suggests the involvement of PI3K, which may convert PIP2 into PIP3 to activate talin more potently. This is consistent with the recent finding that knockdown of PIPKIγ reduced PIP2 levels by 40% and PIP3 to undetectable levels36. PI3K is known to promote integrin activation37, but detailed elucidation of this process will require future investigations in vivo. We note that in our competition experiment (Figure 4B), PIP2 did not completely convert inactive talin into the active state, suggesting an equilibrium between the two states that are balanced by the PIP2 composition of the membrane and possibly other factors. For example, RAP1-RIAM has been suggested to promote the membrane recruitment of talin and regulate integrin activation38. Although the exact mechanism remains to be determined, it is possible that RAP1-RIAM cooperates with PIP2-containing membrane in a combinatorial fashion to enhance the equilibrium shift from inactive state to active state of talin.

Figure 5
figure 5

A mechanism of spatiotemporal regulation of talin in mediating integrin activation. In unstimulated cells, talin is autoinhibited and hindered from accessing membrane via the charge-charge repulsion mechanism. Agonist stimulation promotes the co-localization of PIPKIγ with talin, thereby leading to the local enrichment of PIP2 for activating talin via the electrostatic pull-push mechanism. The electrostatic pull-push process is different from known steric occlusion by PIP2 that directly interferes with the autoinhibitory protein interfaces34,35.

Full size image

In summary, we have defined a novel on/off switch mechanism of talin-FERM in controlling integrin transmembrane signaling. In particular, our results unravel a unique role of membrane surface in this on/off switch process, i.e., membrane may act as a dynamic sensor, depending upon its surface chemical environment, to delocalize/inhibit or recruit/activate talin so as to rapidly and effectively communicate with integrins to mediate the integrin activation. Since talin binds to actin via the C-terminal end of the talin-R, such membrane-dependent talin/integrin interaction may also provide means for regulating the mechanical coupling of the integrin-actin linkage as well as membrane-actin linkage (Figure 5), thereby promoting cytoskeleton reorganization with a profound effect on diverse cellular events, such as cell migration, cell spreading, cell proliferation, and survival.

Materials and Methods

Peptide and protein preparation and purification

Talin fragment talin-F0F1 (1-206) was subcloned into the parallel His-1 vector containing a TEV-cleavable N-terminal His tag. The DNA sequence was verified and transformed into E. coli. The protein was purified according to the native purification protocol from Qiagen. The His tag was removed and the protein was further purified using size-exclusion chromatography on Superdex 75 (Amersham Bioscience). Other talin fragments including talin-F3 (309-405) with a solubility improving mutation C336S, talin-F2F3 (206-405), talin-RS (1654-1848), and remaining talin-R fragments were prepared as described previously19. 15N and/or 13C isotope labeling was achieved by employing 15NH4Cl (1.1 g/l) and/or 13C glucose (3 g/l) as the sole nitrogen and carbon sources in the cultures. 2H isotope labeling was achieved by using 2H glucose (3 g/l) and preparing the culture in 99.8% D2O. Double mutation of talin-F2F3 (K254A/K256A, K272A/K274A, and K322A/K324A) were made using QuickChange Site-Directed Mutagenesis (Stratagene) and prepared in the same way as wild type.

Lipid vesicle preparation

POPC, POPS, and PIP2 were purchased from Avanti Polar Lipids (Alabaster, AL, USA). LUVs of either single or mixed lipid were prepared by extrusion. Briefly, lipids were first dissolved together in chloroform. The chloroform was then removed under a stream of nitrogen followed by overnight vacuum pumping. The lipid film was hydrated in desired buffer, followed by homogenization with a few freeze-thaw cycles. The LUV was finally formed by extruding the lipid suspension ∼20 times through two stacked 0.1 μm polycarbonate filters (Avanti Polar Lipids).

Crystallization, data collection, and structure determination

Crystals of the talin-F2F3/talin-RS complex were grown at 16 °C by the hanging-drop vapor diffusion method in drops containing a 2:1 (v/v) ratio of protein solution (∼0.2 mM) and reservoir solution consisting of 100 mM MES, pH 6.5, 18.5% (w/v) polyethylene glycol 1500. Data sets were collected at 100 K at the beamline BL17U at the Shanghai Synchrotron Radiation Facility (Shanghai, China) using an ADSC Q315 CCD detector. All X-ray intensity data were integrated and scaled using the programs HKL200039. The structure was determined by the molecular replacement method with a crystal structure of talin-F2F3 (PDB entry: 3G9W) and an NMR structure of talin-RS (PDB entry: 2KBB) as starting models. Manual model building was carried out by the program COOT40 and completed by TURBO-FRODO41. The model was refined by iterative rounds of manual rebuilding followed by PHENIX42 and Refmac543. Stereochemistry of the structures was assessed by PROCHECK44. Shape correlation statistics were calculated by SC45. Structures are visualized with PyMOL (Schrodinger, LLC. http://www.pymol.org), and the surface electrostatic potential was calculated with the program APBS46.

NMR spectroscopy

All heteronuclear NMR experiments were performed at 25 °C on Bruker Avance 600 MHz and 900 MHz spectrometers, equipped with cryogenic triple resonance probes and shielded z-gradient units. All the NMR samples contain 5% (v/v) of 2H2O and 1 mM DSS. Titrations of wild-type and double-mutated talin-F2F3 with various ratios of different LUVs or IP3/IP4 were conducted with 0.04 mM 15N-labeled protein in MES buffer (20 mM MES, 20 mM NaCl, 2 mM NaN3, and pH 6.1). The HSQC experiments to examine the autoinhibitory talin complex were performed with 0.06 mM 13C/15N-labeled talin-RS and 0.05 mM-unlabeled talin-F2F3 in phosphate buffer (50 mM Na2HPO4/NaH2PO4, 50 mM NaCl, 2 mM NaN3, and pH 6.8). The effects of membrane lipids were examined with the addition of 3 mM desired LUV to the complex. The weighted chemical shift changes of amide proton and nitrogen were calculated using the equation: Δδobs[HN,N]=([ΔδHNWHN]2+[ΔδNWN]2)1/2, where WHN (1.0) and WN (0.154) are weighting factors based on the gyromagnetic ratios of 1H and 15N. NMR data were processed and analyzed using nmrPipe47, PIPP48, and Sparky49.

Isothermal calorimetry measurements

Isothermal titration calorimetry was performed using a VP-ITC instrument (MicroCal). Talin-H and talin-F2F3 were dialyzed against a buffer containing 100 mM NaCl and 50 mM Na2HPO4/NaH2PO4 at pH 6.8. D-myo-phosphatidylinositol 4,5-bisphoshate or PIP2-diC8 (Echelon Biosciences) was dissolved in the same buffer to a final concentration of 1.2 mM. Titrations consisted of 30 injections of 7.5 μl PIP2-diC8 stock into the sample cell containing 50 μM of 1.4 ml Talin-H protein. A control experiment was performed to estimate the heat of dilution and heat of mixing by injecting the ligand after saturation of the experiment. This was subtracted from all data points prior to curve fitting. The data were fitted into a two site binding model and thermodynamic parameters were calculated using Origin software provided by MicroCal. All experiments were performed at 25 °C.

Plasmids, mutagenesis, and transfections

The preparation of plasmids for mouse full-length talin1 and M319A mutant and their transfections were described previously19. Other mutations used for this study were performed using QuickChange Site-Directed Mutagenesis Kit (Stratagene).

Integrin activation assay

The effects of full-length talin, full-length talin mutants, and talin-H on integrin activation were analyzed using CHO cells stably expressing integrin (αIIbβ3-CHO) and an activation-specific anti-αIIbβ3 mAb (PAC1) as described previously19.