Mechanism of integrin activation by talin and its cooperation with kindlin

Abstract

Talin-induced integrin binding to extracellular matrix ligands (integrin activation) is the key step to trigger many fundamental cellular processes including cell adhesion, cell migration, and spreading. Talin is widely known to use its N-terminal head domain (talin-H) to bind and activate integrin, but how talin-H operates in the context of full-length talin and its surrounding remains unknown. Here we show that while being capable of inducing integrin activation, talin-H alone exhibits unexpectedly low potency versus a constitutively activated full-length talin. We find that the large C-terminal rod domain of talin (talin-R), which otherwise masks the integrin binding site on talin-H in inactive talin, dramatically enhances the talin-H potency by dimerizing activated talin and bridging it to the integrin co-activator kindlin-2 via the adaptor protein paxillin. These data provide crucial insight into the mechanism of talin and its cooperation with kindlin to promote potent integrin activation, cell adhesion, and signaling.

Fig. 1. Talin is an intracellular activator of integrin.

a Domain organization of talin. N-terminal head domain (talin-H) is composed of four subdomains and C-terminal rod domain (talin-R) contains 13 subdomains followed by dimerization domain (DD). Talin-H and talin-R are connected by a flexible linker. b A scheme of talin activation and talin-mediated integrin activation. Upon stimulation, inactive talin is released from an autoinhibited state to expose talin-H to bind integrin cytoplasmic tail and trigger conformational change of integrin to bind ligand. The autoinhibitory interface between talin F3 and talin R9 is displayed in the box showing how M319, T1767, and E1770 are involved in the interface and mutated to release the autoinhibiton. c Diagram of the various talin variants, including mutation sites to activate talin or disrupt talin dimerization. d Left panel: Integrin-activation level reflected by monomeric FN10 binding is not elevated much by constitutively activated full-length talin mutant, (tlnM3) compared with talin-H (tlnH). **p = 0.0081 with 95% confidence interval 0.2727–0.9914 (t-test), N = 3 biologically independent samples. All values are given as mean ± S.E.M. Right panel: FN10 staining was similar between tlnH (orange) and tlnM3 (light green). e Left panel: Integrin-activation level reflected by PAC-1 binding is substantially elevated by tlnM3 compared with tlnH. Talin rod (tlnR) alone has no effect on integrin activation. ****p < 0.0001 with 95% confidence interval 5.478–6.260 (t-test), N = 6 biologically independent samples. All values are given as mean ± S.E.M. Right panel: Significantly more PAC-1 staining was induced by intact active talin (tlnM3, light green) in comparison with talin head (tlnH, orange) without normalization. d, e Raw data are provided in Source Data file.

Fig. 2. Dimerization of active talin is required to promote potent ligand binding to integrin.

a A model of multivalent ligand binding to integrin mediated by talin dimerization. b Integrin activation triggered by active full-length talin (tlnM3) is partially reduced by deletion of dimerization domain (tlnM3 DDdel) as measured by the PAC-1-binding assays. All values are given as mean ± S.E.M. ****p < 0.0001 with 95% confidence interval −2.814 to −1.826 (t-test), N = 6 biologically independent samples. c Histogram of PAC-1 binding shows that DDdel (magenta) causes significantly reduced PAC-1 staining. d Cell adhesion to fibronectin, vitronectin, and laminin are all affected when we introduce endogenous level of talin DD deletion or R2526G mutation to talin^1/2dko fibroblasts to disrupt talin dimerization. WT values were set to 1. Relative cell adhesion was plotted as mean ±95% CI, N = 6 independent experiments. Fibronectin: ***p < 0.0001, Vitronectin: *p = 0.0233, **p = 0.0018, Laminin: **p = 0.0097. Statistical significance was tested by one-way ANOVA followed by Tukey’s multiple-comparison test. b, d Raw data are provided in Source Data file.

Fig. 3. Kindlin cooperates with intact active talin via paxillin to activate integrin.

a Kindlin-2 synergizes with intact active talin (tlnM3 + K2) to trigger PAC-1 binding to integrin-αIIbβ3. This effect is drastically lower in cells expressing similar level of talin head and kindlin-2 (tlnH+K2). ****p = 0.0001 with 95% confidence interval 13.02–22.32 (t-test), N = 6 biologically independent samples. Values are shown as mean ± S.E.M. b Representative histogram of (a) where tlnM3 + K2 (dark blue) enhances PAC-1 binding much more drastically than tlnH+K2 (light blue). c Talin WT does not cooperate with kindlin-2 in PAC-1-binding tests while tlnM3 does. ****p < 0.0001 with confidence interval 19.58–24.34 (t-test), N = 3 biologically independent samples; ns, p = 0.0815 with confidence interval −0.0459 to 0.5080 (t-test), N = 3 biologically independent samples. Values are shown as mean ± S.E.M. d Kindlin cooperation with full-length active talin (tlnM3 + K2WT) was significantly impaired when the paxillin-binding defective kindlin-2 mutant G42K/L46E (tlnM3 + K2GLKE) was expressed. **p = 0.0014 with 95% confidence interval −9.039 to −3.018 (t-test), N = 6 biologically independent samples. Values are shown as mean ± S.E.M. e Paxillin knockdown substantially reduced the synergy of tlnM3/kindlin-2 to activate integrin. Note that the effect of tlnM3 alone on PAC-1 binding was also significantly reduced (~15%), albeit less than that by tlnM3/kindlin-2 co-expression (~25%) and the latter had higher effect probably due to the overexpression of both tlnM3 and kindlin-2 vs tlnM3 alone in the former. **p = 0.0058 with 95% confidence interval −2.228 to −0.7115 (t-test); ***p = 0.0001 with 95% confidence interval −10.86 to −7.379 (t-test), N = 3 biologically independent samples. Values are shown as mean ± S.E.M. f Western blot showing that paxillin was efficiently knocked down. One out of two independent experiments is shown here. g Paxillin synergizes with tlnM3 to induce potent integrin activation. The synergy (tlnM3+paxiWT) was impaired by kindlin-2-binding defective paxillin mutant F577E (tlnM3 + paxiF577E). ****p = 0.0001 with 95% confidence interval −9.442 to −7.087 (t-test), N = 3 biologically independent samples. Values are shown as mean ± S.E.M. h Kindlin-2 knockdown substantially reduces the synergy of tlnM3/paxillin to activate integrin. ****p < 0.0001 with 95% confidence interval −21.65 to −19.14 (t- test), N = 3 biologically independent samples. Values are shown as mean ± S.E.M. i Western blot showing that paxillin was efficiently knocked down. One out of two independent experiments is shown here. Uncropped images (f, i) and raw data (a, c–e, g, h) are provided as a Data Source file.

Fig. 4. Full-length talin WT cooperates with kindlin through paxillin to induce stronger cell adhesion in comparison with talin-H.

a Cell adhesion of talin^1/2dKO fibroblasts expressing ypet, ypet-tagged talin head (tlnH), talin WT, or constitutively active talin (tlnM3) on fibronectin- or vitronectin-coated surfaces. WT values were set to 1. Data represent mean ±95% CI. N = 4/3 experiments. Fibronectin: **p = 0.0050, ***p = 0.0007, Vitronectin: *p = 0.0179, **p = 0.0068 (one-way ANOVA followed by Tukey’s multiple-comparison test). b Cell-spreading area of cells expressing tlnH, tlnWT, or tlnM3 seeded on fibronectin after 30 min. Data are presented as mean ±95% CI. N = 4 experiments. ***p < 0.0001. (c/d) Cell adhesion of talin^1/2dKO rescued with ypet alone, ypet–tlnH, or ypet–tlnWT in the presence (K2WT) or absence (K2KO) of kindlin-2. Kindlin-2 was knocked out by CRISPR/Cas9, K2WT cells were treated with a nontargeting control guideRNA. (c) Expression levels of ypet, full- length talin, talin head, and kindlin-2 assessed by Western blot. GAPDH served as loading control. d Static adhesion assays of the mentioned cell lines on fibronectin-coated surfaces. Data represent mean ±95% CI. N = 5 independent experiments. *p = 0.0339 (tlnH/K2WT vs tlnWT/K2WT), p = 0.0276 (tlnWT/K2WT vs tlnH/K2KO), p = 0.0199 (tlnWT/K2WT vs tlnWT/K2KO), **p = 0.0072 (one-way ANOVA followed by Tukey’s multiple-comparison test). e Western blots showing expression of mCherry, talin, kindlin-2, and paxillin in K2KO ypet–talinWT-expressing cells retrovirally transduced with expression constructs carrying mCherry or mCherry-tagged wild-type (K2WT) or paxillin-binding defective (K2GLKE) kindlin-2 and subsequently treated either with negative control (NC) or paxillin-targeting (pxn_s04) siRNA. GAPDH served as loading control. Representative blots from one out of four experiments are shown. f Static adhesion assay of transduced cells with or without paxillin-targeting siRNA treatment. Data represent mean ±95% CI. N = 7 independent experiments. **p = 0.0052, ***p = 0.0004 (K2WT/NC vs K2GLKE/NC), p = 0.0001 (K2WT/NC vs K2GLKE/pxn_s04) (one-way ANOVA followed by Tukey’s multiple-comparison test). Uncropped images (c, e) and raw data (a, b, d, f) are provided as a Data Source file.

Fig. 5. Paxillin binds to full-length active talin through multiple sites.

a Paxillin prefers binding to activated talin as shown by GST pulldown. GST-tagged full-length paxillin was immobilized to GST beads and incubated with the same amount of full-length wild-type talin (talinFL WT) and full- length activated talin (talinFL TM). Pull-down fraction of talin was detected by talin antibody on western blot. Immobilized GST proteins were detected by GST antibody on western blot after 10x dilution. Four independent experiments were performed. b N-terminal part of paxillin (paxi1–160) is mainly responsible for binding to full- length active talin, while paxillin C-terminal (paxi161–605) shows no binding to talinFL TM. Pull-down fraction of talin was detected by Anti-6xHis antibody on western blot. Immobilized GST protein was resolved by SDS-PAGE and shown by Coomassie blue staining. As can be recognized from Coomassie staining, GST-paxillin 161–605 became unstable upon the 1–160 deletion, which degraded further after gel filtration. To alleviate this problem, we loaded much more paxillin 161–605 to have the major band (see arrow) to match other paxillin inputs, yet there was still no talin pulled down, indicating that paxillin 161–605 does not contribute to talin binding in contrast to paxillin 1–160. Four independent experiments were performed. c The HSQC spectra of 50 µM ¹⁵N-labeled talin-R2 in the absence (black) and presence of 100 µM unlabeled paxillin-γ 1–160 (red) showing direct interaction between paxillin and talin R2. d Cartoon representation of the best model of Paxillin-LD2 (in cyan) and talin-R2 (in green) complex through Haddock calculation. Side chains of critical residues involved in the binding are displayed and labeled. Red-dashed lines indicate potential H-bonds. e Mutations in talin-R2 (A680E/K687E, AKEE) abolished talin-R2 binding to paxillin 1–160 at the same experimental condition as (c). f Summary of mutation testing results from HSQC. Uncropped images are provided in Source Data file.

Fig. 6. Paxillin-binding defective mutations in talin impair integrin activation and cell adhesion.

a Diagram of paxillin-binding defective talin mutants. b Paxillin-binding defective mutations on talin rod (R2mut/ R2R11mut/R2R8R11mut) reduced the tlnM3/paxillin synergy (tlnM3+paxiWT) to activate integrin. Combined paxillin-binding defective mutations on talin rod (tlnM3 R2R8R11mut) and kindlin-binding defective mutation on paxillin C-terminus (paxiF577E) lead to the lowest PAC-1 binding, comparable to that of only expressing active talin (tlnM3). **p = 0.0054 with 95% confidence interval −14.37 to −3.650 (t-test), N = 3 biologically independent samples; ***p = 0.0008 with 95% confidence interval −18.10 to −7.368 (t-test) or 0.0001 with 95% confidence interval −22.60 to −11.85 (t-test), N = 3 biologically independent samples. Values are given as mean ± S.E.M. c Immunoprecipitation experiments (talin^1/2dKO fibroblasts expressing ypet–talin WT/paxillin- binding mutants) show that talin R2 (AKEE) and R11 (TKEE) mutations both reduced paxillin binding and co-immunoprecipitation of kindlin-2. Relative paxillin and kindlin-2 intensities were plotted as mean ± S.E.M. Paxillin and kindlin-2 intensities were normalized to intensities of immunoprecipitated talin, N = 3 independent experiments. d Cell adhesion to fibronectin, vitronectin, and laminin are all affected by talin R2 (AKEE) and R11 (TKEE) mutations. WT values were set to 1. Relative cell adhesion was plotted as mean ± 95% CI, N = 8 experiments. Fibronectin: *p = 0.0198, **p = 0.0036 (WT vs R2R11mut), p = 0.0073 (R11mut vs R2R11mut), Vitronectin: *p = 0.0205 (WT vs R2mut), p = 0.016 (WT vs R2R11mut), **p = 0.0048, Laminin: *p = 0.0278, **p = 0.0045 (WT vs R2mut), p = 0.0035 (WT vs R2R11mut), p = 0.0013 (R11mut vs R2R11mut) (one-way ANOVA followed by Tukey’s multiple-comparison test). e Both R2 (AKEE) and R11 (TKEE) mutations lead to reduced cell spreading. Spreading areas of 30 cells per cell line and time point were measured per experiment, N = 6. **p = 0.0022, ***p = 0.0002 (one-way ANOVA followed by Sidak’s multiple-comparison test). f Both R2 (AKEE) and R11 (TKEE) mutations reduced focal adhesion (FA) number as well as total and relative FA area per cell, but only a combination of both rod mutations reduced FA size. FAs were characterized by quantification of paxillin immunofluorescence stainings shown in Fig. S12D. Data are shown as mean ±95% CI, N = 10 experiments (8 cells were analyzed per group and experiment). g Localization of talin and kindlin-2 to FAs was reduced in cells expressing paxillin-binding defective talin mutants. Active β1-integrin-specific antibody 9EG7 staining intensities within FAs were reduced in immunofluorescence stainings of cells expressing R2 and R2/R11 mutant talin variants. Bar plots represent mean ±95% CI, N = 10 or 4 or 6 experiments (8 cells were analyzed per group and experiment). f, g *p < 0.05, **p < 0.01, ***p < 0.001. Statistical significance was tested by one-way ANOVA followed by Tukey’s multiple-comparison test. Uncropped images (c), exact p-values (f–g), and raw data (b, d–g) are provided in Source Data file.

Fig. 7. Both talin dimerization and talin–paxillin–kindlin pathway contribute to integrin activation.

a PAC-1-binding assay showing that integrin activation by full-length active talin (tlnM3) was reduced by deletion of talin-dimerization domain (tlnM3 DDdel), and addition of paxillin-binding mutations (tlnM3 R2R8R11mut) further reduced PAC-1 binding to almost the level caused by talin head (tlnH) only. *p = 0.0416 with 95% confidence interval −4.267 to −0.0963 (t-test); ***p = 0.0003 with 95% confidence interval −6.453 to −2.529; N = 6 biologically independent samples. Values are shown as mean ± S.E.M. b Synergy between intact kindlin-2 and intact active talin (tlnM3 + K2FL) was significantly impaired by DDdel as shown by reduced PAC-1 binding. Combination of paxillin-binding deficient mutations and DDdel mutation further substantially reduced PAC-1 binding. *p = 0.0286 with 95% confidence interval −6.485 to −0.4448 (t-test), N = 3 biologically independent samples; ****p < 0.0001 with 95% confidence interval −13.40 to −7.084 (t-test), N = 6 biologically independent samples. Values are given as mean ± S.E.M. c A model for dynamic integrin-activation process where inactive talin first engages with PIP2 to initiate, via talin-H, the conformational opening of talin, which initiates the recruitment of paxillin (Step 1). At Step 2, activated talin dimer, while anchored to PIP2 membrane, clusters two conformationally open integrins (i1 and i2) via each talin monomer subunit. Meanwhile, talin, which is bound to integrin i1, links, via paxillin, to kindlin that is bound to another integrin i2. Such talin–paxillin–kindlin linkage (primarily driven by talin-R as shown in the text) strengthens the talin-dimer-induced microclustering of integrins, leading to their potent multivalent binding and efficient FA assembly for cell adhesion. a, b Raw data are provided in Source Data file.