Paxillin binding to the PH domain of kindlin-3 in platelets is required to support integrin αIIbβ3 outside-in signalling

Abstract

Background:

Kindlin-3 is essential for supporting the bidirectional signaling of integrin αIIbβ3 in platelets by bridging the crosstalk between integrin αIIbβ3 and the cytoplasmic signaling adaptors.

Objective:

In this study, we identified a previously unrecognized paxillin binding site in the pleckstrin homology (PH) domain of kindlin-3 and verified its functional significance.

Methods:

Structure-based approaches were employed to identify the paxillin binding site in the PH domain of kindlin-3. In addition, the bidirectional signaling of integrin αIIbβ3 were evaluated in both human and mouse platelets.

Results:

In brief, we found that a β1-β2 loop in the PH domain of kindlin-3, an important part of the canonical membrane phospholipid binding pocket, was also involved in mediating paxillin interaction. Interestingly, the binding sites of paxillin and membrane phospholipids in the PH domain of kindlin-3 were mutually exclusive. Specific disruption of paxillin binding to the PH domain by point mutations inhibited platelet spreading on immobilized fibrinogen while having no inhibition on soluble fibrinogen binding to stimulated platelets. In addition, a membrane-permeable peptide derived from the β1-β2 loop in the PH domain of kindlin-3 was capable of inhibiting platelet spreading and clot retraction, but it had no effect on soluble fibrinogen binding to platelets and platelet aggregation. Treatment with this peptide significantly reduced thrombus formation in mice.

Conclusion:

Taken together, these findings suggest that interaction between paxillin and the PH domain of kindlin-3 plays an important role in supporting integrin αIIbβ3 outside-in signaling in platelets, thus providing a novel antithrombotic target.

1 |. INTRODUCTION

Integrins are heterodimeric transmembrane receptors each consisting of an α and a β subunit.¹ Platelet-specific integrin αIIbβ3 can mediate platelet adhesion, aggregation, spreading, and clot retraction in both hemostatic and thrombotic conditions,^2,3 thus serving as an ideal target for developing antiplatelet drugs.^4–6 Integrin αIIbβ3 activation, a transition from a low- to a high-affinity state, is prerequisite for integrin αIIbβ3 binding to soluble ligands (inside-out signaling); and extracellular ligand binding further triggers outside-in signaling to facilitate platelet shape changes.⁷ A set of precisely regulated steps are required for modulating integrin αIIbβ3 bidirectional signaling in platelets,^8,9 allowing normal hemostatic activity while avoiding thrombotic risks, in which the crosstalk between integrin αIIbβ3 cytoplasmic tails (CTs) and their binding proteins in platelets plays a key role.^10,11 Talin and kindlin-3 are two essential integrin αIIbβ3 activators in platelets that function via directly binding to the integrin β3 CT.^11,12 When platelets are stimulated, agonist-triggered signaling drives the talin head domain to interact with the membrane-proximal NxxY motif in the integrin β3 CT and unclasp the integrin α/β CT complex, which induces integrin αIIbβ3 extracellular conformational changes to expose the extracellular ligand binding sites.^13–15 Meanwhile, kindlin-3 interacts with the membranedistal NxxY motif in the integrin β3 CT and cooperates with the talin head domain to support integrin αIIbβ3 bidirectional signaling in platelets.^16–18

Kindlin-3 is one of the kindlin family members, which are 4.1/ezrin/radixin/moesin (FERM) domain-containing adaptors. There are three kindlins in mammals, kindlin-1, kindlin-2, and kindlin-3, and they share high sequence homology and structural similarity.^19,20 In addition to having three typical FERM subdomains (F1, F2, and F3), kindlins also possess an N-terminal F0 subdomain and a pleckstrin homology (PH) domain inserted in the F2 subdomain.¹² Kindlin-1 is primarily expressed in epithelial cells, kindlin-2 is ubiquitously expressed, and kindlin-3 is mostly restricted to cells of hematopoietic origin.²¹ Deficiency of kindlin-3 in humans leads to type III leukocyte adhesion deficiency (LAD-III) and LAD-III patients show severe bleeding disorder and recurrent infections due to compromised integrin signaling in both platelets and leukocytes.^16,22,23 Mice with deficiency of kindlin-3 phenotypically mirror LAD-III patients.^17,18,24,25 Nonetheless, the detailed mechanism by which kindlin-3 mediates integrin αIIbβ3 bidirectional signaling in platelets remains largely unknown.

It has been believed that kindlins execute their functions by bridging crosstalk between the integrin β CTs and the cytoplasmic signaling adaptors.^20,26 Recently, several studies from different research groups including ours have shown that the paxillin family members bind kindlins and play an important role in regulating integrin-mediated cellular responses.^27–32 All three paxillin family members (paxillin, Hic-5, and leupaxin) are multifunctional adaptors,³³ each containing four or five N-terminal leucine-rich LD motifs and four C-terminal LIM domains.^34–36 The PH domain in kindlin-2 was first identified to mediate paxillin binding,³² but later it was demonstrated that the F0 subdomain in kindlins was involved in interacting with paxillin.^27–29 Disruption of kindlin binding to paxillin by introducing specific mutations in the F0 subdomain of kindlin-1 or kindlin-2 was able to suppress integrin-mediated cell adhesion, spreading, and migration.^27–29 In addition, we found that disconnection of paxillin with kindlin-3 by introducing specific mutations in the F0 subdomain in kindlin-3 impaired integrin αIIbβ3 bidirectional signaling in mouse platelets.²⁸ Employing the same mutations in kindlin-3, Klapproth et al. discovered that kindlin-3 was required to regulate stability and turnover of osteoclast podosomes through recruiting leupaxin.³¹ Interestingly, a recent study showed that kindlin-3 binding to paxillin/leupaxin via its F0 subdomain was capable of limiting myeloid cell mobility and phagocytosis.³⁰ Collectively, a body of solid evidence has established the functional involvement of kindlin binding to the paxillin family members via their F0 subdomains in supporting integrin-mediated cellular activities. However, it remains unknown if crosstalk between paxillin and the PH domain of kindlin-3 exists and plays a role in platelets. Therefore, in this study, we focus on their interaction and an evaluation of the functional consequence in integrin αIIbβ3 signaling in platelets.

2 |. RESULTS

2.1 |. Paxillin interacts with the kindlin-3-PH domain

The controversial findings in previous studies regarding the interaction between paxillin and the PH domain of kindlin-2 drove us to test if the PH domain in kindlin-3 (kindlin-3-PH) is involved in mediating paxillin binding.^29,32 First, pull-down assays were performed. Protein A/G beads coated with flag-tagged paxillin were used to incubate with glutathione S-transferase (GST)-fused PH domain of kindlins. As shown in Figure 1A, GST-fused kindlin-3-PH interacted with paxillin whereas binding of GST-fused kindlin-1-PH or kindlin-2-PH to paxillin was noticeably reduced. In addition, paxillin in cell lysates was found to interact with GST-fused kindlin-3-PH coupled on glutathione-agarose beads (Figure 1B); and consistently, binding of paxillin to GST-fused kindlin-1-PH or kindlin-2-PH was comparably weak. Second, biolayer interferometry (BLI) was employed to measure their interactions. Single-state association and dissociation kinetics of GST-fused kindlin-PH domain binding to paxillin were plotted as colored lines, and the equilibrium dissociation constants were also presented (Figure 1C–F). In line with the pull-down results, the BLI data showed that kindlin-3-PH exhibited the strongest binding affinity to paxillin compared to the other two counterparts, with a K_D of 7.6 μM for kindlin-3-PH versus 14.2 μM and 23.5 μM for kindlin-1-PH and kindlin-2-PH, respectively. The interactions between paxillin and each of the kindlin-PH domains are specific because they are significantly higher than the background as evaluated by GST alone (217.4 μM). These findings indicate that the PH domains of kindlins do possess the ability to bind paxillin, although paxillin preferentially interacts with the PH domain of kindlin-3.

FIGURE 1.

Paxillin interacts with the pleckstrin homology (PH) domains of kindlins. A, Protein A/G beads were pre-coated with flag-paxillin (PXN) and used to incubate with glutathione S-transferase (GST) or GST-fused kindlin-PH proteins (K1-PH, K2-PH, and K3-PH). After washing, the precipitates were evaluated with sodium dodecyl sulfate—polyacrylamide gel electrophoresis (SDS-PAGE) followed by western blotting. Loading of GST or GST-fused proteins were measured by Coomassie blue staining. B, Glutathione-sepharose beads coupled with GST or GST-fused kindlin-PH proteins were used to incubate with Chinese hamster ovary cell lysates expressing flag-PXN. After washing, the precipitates were evaluated with SDS-PAGE followed by western blotting and Coomassie blue staining. C–H, The biolayer interferometry assay was performed as described in the Methods section. A ligand was immobilized on the biosensor tip and its association and dissociation with an analyte in solution were measured using the Octet Red96 system. Kinetics of the association and the dissociation at different concentrations of analytes (30, 20, 15, 10, 7.5, and 3 μM) were shown. The K_D was calculated from the globally fit k_on and k_off (K_D = k_off/k_on)

Further, we sought to identify the kindlin-3-PH binding region in paxillin. Previous studies showed that both LIM3 and LIM4 domains at the C-terminus of paxillin were involved in kindlin-2 binding.^29,32 Therefore, we examined the respective binding of LIM3 and LIM4 of paxillin to kindlin-3-PH by BLI, and found that LIM3 but not LIM4 of paxillin was able to interact with kindlin-3-PH (Figure 1G,H). Next, we confirmed that kindlin-3-PH also interacted with the other two paxillin family members, Hic-5 and leupaxin (data not shown). Together, these findings suggest that the PH domain of kindlin-3 can interact with paxillin and its family members.

2.2 |. A β1-β2 loop in kindlin-3-PH is involved in mediating interaction with paxillin

The structures of the PH domains in three kindlins are highly conservative, characterized with a typical PH domain core formed by a seven-stranded β-barrel, with one end open responsible for phospholipid binding and the other end capped by two α-helixes^37–40 (Figure 2A,B). To identify the paxillin binding site in kindlin-3-PH, the exposed surface residues in kindlin-3-PH were selected based on their solvent accessibility (>30%) and grouped into seven clusters (Figure 2A, underlined); and accordingly, seven kindlin-3-PH mutants (M1-M7) were generated by alanine substitution for each of the clusters. These mutants were expressed and purified and their binding to paxillin was measured by BLI. Compared to wild-type kindlin-3-PH, both the M1 and the M2 mutants showed compromised binding while all other mutants exhibited comparable binding to paxillin (Figure 2C). As shown in Figure 2A,B, both the M1 and the M2 localize in the β1-β2 loop, suggesting that this loop in kindlin-3-PH domain is involved in mediating paxillin binding.

FIGURE 2.

Identification of the paxillin binding site in the pleckstrin homology (PH) domain of kindlin-3. A, Amino acid sequences of the PH domains of kindlins were shown. The selected residues exposed on the surface of the PH domain of kindlin-3 were underlined for alanine substitution. B, The published crystal structure of the PH domain of human kindlin-3 was shown (PDB: 5L81), in which the phospholipid binding pocket is pointed out with an arrow. C, Binding kinetics between paxillin and the PH domain of kindlin-3 (K3-PH) carrying the indicated mutations were measured by the biolayer interferometry assay. Values were obtained from global fitting of data using a 1:1 binding model. The K_D was calculated from the globally fit k_on and k_off

2.3 |. Binding of phospholipids and paxillin to kindlin-3-PH are mutually exclusive

Structurally, the β1-β2 loop in kindlin-3-PH participates in formation of the phospholipid-binding pocket.³⁹ Therefore, we next assessed the phospholipid-binding capacity of the M1 and the M2 mutants of kindlin-3-PH using a lipid-protein interaction assay. A phospholipid array spotted with a concentration gradient of seven types of phosphoinositides (PtdInsPs) and phosphatidylinositol (PtdIns) as a control was used to incubate with GST-fused kindlin-3-PH, and the binding was evaluated by western blotting with an anti-GST antibody. As shown in Figure 3A, GST-fused kindlin-3-PH but not GST had the ability to interact with PtdInsPs, except PtdIns(3,4)P2 and the control PtdIns. Interestingly, the M1 mutant of kindlin-3-PH failed to interact with PtdInsPs while the M2 mutant of kindlin-3-PH still exhibited the PtdInsP-binding ability that was similar to the wild type, suggesting that the M1 region but not the M2 region in kindlin-3-PH is involved in mediating PtdInsP binding. Because it is known that two lysine residues (K363 and K367) in the M1 region are required for PtdInsP binding,³⁹ we further generated an M1’ mutant by substituting these two lysine residues with alanine (KK/AA) in kindlin-3-PH and found that the M1’ mutant of kindlin-3-PH showed substantially decreased PtdInsP binding while retaining the paxillin binding ability (Figure 3A,B). These results disclose that the binding sites for paxillin and phospholipids in kindlin-3-PH are spatially overlapping but not identical, therefore providing specific mutations in kindlin-3-PH for disrupting its binding to phospholipids and paxillin, respectively.

FIGURE 3.

Binding of paxillin and phospholipids to the pleckstrin homology (PH) domain of kindlin-3 are mutually exclusive. A, Binding of the PH domain of kindlin-3 (K3-PH) to phospholipids was examined by the protein-lipid overlay assay. Glutathione S-transferase (GST), GST-fused K3-PH, or the indicated mutants (M1, M2, and M1’) were used to incubate with a hydrophobic membrane spotted with a concentration gradient of indicated phosphoinositides (PtdInsPs). The M1’ mutant of K3-PH contains two K/A substitutions in the M1 region. The lipid--protein binding was detected by western blotting using an anti-GST antibody. To examine if paxillin (PXN) blocks K3-PH binding to PtdInsPs, GST-fused K3-PH was pre-incubated with PXN at 1:10 molar ratio before adding to the membrane. B, Interaction of PXN with the M1’ mutant of K3-PH was measured by the biolayer interferometry (BLI) assay. C, BLI sensorgrams showed K3-PH binding to PXN in the presence of the indicated PtdInsPs

Further, we tested the mutual influence of paxillin and PtdInsPs on binding to kindlin-3-PH. On one hand, we found that presence of paxillin was able to block kindlin-3-PH binding to PtdInsPs (Figure 3A); on the other hand, we detected that pre-incubation of kindlin-3-PH with PtdIns(4,5)P2 or PtdIns(3,4,5)P3, but not the control PtdIns, was able to suppress kindlin-3-PH binding to paxillin (Figure 3C). Together, these results suggest that binding of paxillin and phospholipids to kindlin-3-PH are mutually exclusive.

2.4 |. Paxillin binding to kindlin-3-PH supports integrin αIIbβ3 outside-in signaling in mouse platelets

To evaluate how paxillin binding to the PH domain of kindlin-3 affects the bidirectional signaling of integrin αIIbβ3 in platelets, we expressed the M2 mutant of kindlin-3 in mouse platelets and performed functional analysis, as previously described.^28,41 In addition, the M1’ mutant of kindlin-3 was also employed to specify the functional effect of phospholipid binding to the PH domain of kindlin-3. In brief, bone marrow cells isolated from kindlin-3-deficient mice were transduced with lentiviral particles expressing EGFP-fused kindlin-3 (either wild type or the mutants) and then transplanted into lethally irradiated C57BL/6 recipient mice. After transplantation for 8 weeks, EGFP-positive platelets were selected for functional analysis. As shown in Figure 4A, after stimulation with either protease-activated receptor 4 (PAR4) agonist peptide or collagen-related peptide (CRP), platelets expressing the M1’ mutant of kindlin-3 exhibited significantly decreased soluble fibrinogen binding compared to platelets expressing wild-type kindlin-3; however, platelets expressing the M2 mutant of kindlin-3 showed comparable soluble fibrinogen binding to platelets expressing wild-type kindlin-3. Interestingly, platelets expressing either the M1’ mutant or the M2 mutant of kindlin-3 spread poorly on immobilized fibrinogen compared to platelets expressing wild-type kindlin-3 (Figure 4B,C and S1 in supporting information). These results suggest that paxillin binding to the PH domain of kindlin-3 in platelets specifically supports integrin αIIbβ3 outside-in signaling while membrane attachment of kindlin-3 via the PH domain is required for both inside-out and outside-in signaling of integrin αIIbβ3.

FIGURE 4.

Paxillin binding to the pleckstrin homology (PH) domain of kindlin-3 is required to support integrin αIIbβ3-mediated mouse platelet spreading. A, Washed platelets from transplanted mice expressing either EGFP-fused wild type kindlin-3 (K3) or the indicated kindlin-3 mutants (M1’ and M2) were stimulated with either protease-activated receptor 4 (PAR4) agonist peptide (PAR4; 150 μM) or collagen-related peptide (CRP; 5 μg/ml) and soluble fibrinogen binding to EGFP-positive platelets was quantified by flow cytometry analysis. B & C, Washed platelets from transplanted mice were allowed to spread on immobilized fibrinogen in the present to PAR4 agonist peptide (150 μM) for 60 min followed by fixation, permeabilization, and staining with Alexa Fluor 568 phalloidin. Spreading areas of EGFP-positive platelets were calculated by Image J software. Results present the mean ± standard deviation of three experiments. *P < .05; **P < .01; ns, non-significant

2.5 |. A peptide derived from the β1-β2 loop of kindlin-3-PH inhibits integrin αIIbβ3 outside-in signaling in platelets and possesses antithrombotic ability

To confirm the direct involvement of the β1-β2 loop in the PH domain of kindlin-3 in supporting integrin αIIbβ3 signaling, we synthesized a peptide encompassing the β1-β2 loop (designated K3PH-L) and a mutated peptide carrying the M2 mutations (designated K3PH-L-M2; Figure 5A). We found that K3PH-L, but not K3PH-L-M2, was able to inhibit kindlin-3-PH binding to paxillin (Figure 5B). However, neither of them inhibited kindlin-3-PH binding to phospholipids (data not shown), suggesting that the inhibitory effect of K3PH-L on paxillin binding to the PH domain of kindlin-3 is specific. To perform functional analysis for platelets, these peptides were N-terminally fused with a membrane-penetrating amino-acid sequence (TAT), as previously described⁴² (Figure 5A). As shown in Figure 5C, neither TAT-K3PH-L nor TAT-K3PH-L-M2 inhibited soluble fibrinogen binding to human platelets upon stimulation with either thrombin activator peptide 6 (TRAP-6) or collagen-related peptide (CRP). In addition, neither of them had striking effect on human platelet aggregation triggered by TRAP-6 or CRP (Figure 5D). Treatment with TAT-K3PH-L significantly inhibited human platelet spreading on immobilized fibrinogen and human platelet-mediated clot retraction upon stimulation with either TRAP-6 or CRP (Figure 5E–G). Comparatively, the inhibitory effect of TAT-K3PH-L-M2 on human platelet spreading or platelet-mediated clot retraction was substantially compromised. Together, these results suggest that the membrane-permeable peptide derived from the β1–β2 loop of the PH domain of kindlin-3 can specifically inhibit integrin αIIbβ3 out-side-in signaling in human platelets.

FIGURE 5.

A peptide derived from the paxillin binding site in kindlin-3-PH (pleckstrin homology) can inhibit integrin αIIbβ3 outside-in signaling in human platelets. A, The amino-acid sequences of selected peptide (K3PH-L) from the β1-β2 loop of kindlin-3-PH and its mutated form (K3PH-L-M2) are shown. For functional analysis, these peptides were conjugated with a membrane-permeable TAT peptide. B, Protein A/G beads pre-coated with flag-paxillin (PXN) were used to incubate with glutathione S-transferase-fused kindlin-3-PH (K3-PH) in the presence of the indicated peptides. After washing, the precipitates were evaluated by sodium dodecyl sulfate—polyacrylamide gel electrophoresis followed with western blotting. C, Washed human platelets isolated from healthy donors were stimulated with either thrombin activator peptide 6 (TRAP-6) (5 μM) or collagen-related peptide (CRP; 1 μg/mL) and soluble fibrinogen binding to platelets in the presence or absence of the indicated peptides (1 μM) was measured by flow cytometry. D, Washed human platelets were used to perform the aggregation assay under stimulation with either TRAP-6 (5 μM) or CRP (1 μg/ml) and the indicated peptides were added to evaluate their effects on aggregation. E, Washed human platelets were stimulated with either TRAP-6 (5 μM) or CRP (1 μg/ml) and allowed to spread on immobilized fibrinogen in the presence or absence of the indicated peptides for 45 min followed by fixation, permeabilization, and staining with Alexa Fluor 568 phalloidin. Platelet spreading areas were calculated by Image J software. F & G, Recalcified human platelet-rich plasma was allowed to retract under stimulation with either of TRAP-6 (5 μM) or CRP (1 μg/ml) in the presence or absence of the indicated peptides for 2 h and the clot retraction efficiencies were calculated. Results present the mean ± standard deviation of three experiments. *P < .05; **P < .01.

Next, we tested if TAT-K3PH-L affects thrombus formation in mice. With or without treatment with TAT-K3PH-L, platelet thrombus formation in mice was evaluated using fluorescent intravital microscopy. Mouse cremaster arterioles were explored and underwent laser injury, and platelet accumulation at the wounded site was visualized and recorded. As shown in Figure 6A,B, platelet thrombus formation quickly occurred in mouse cremaster arterioles after laser injury. Compared to mice without treatment, mice treated with TAT-K3PH-L exhibited reduced platelet accumulation in cremaster arterioles, indicating that TAT-K3PH-L is able to compromise thrombus formation in vivo. Interestingly, there was no significant difference in tail bleeding time between two groups of mice with or without peptide treatment (Figure 6C), suggesting that TAT-K3PH-L may possess antithrombotic capacity while having no significant impairment on normal hemostasis.

FIGURE 6.

The peptide derived from the paxillin binding site in kindlin-3-PH (pleckstrin homology) has an antithrombotic feature. A & B, Intravital microscopy was performed for examining laser injury–induced platelet thrombus formation in cremaster arterioles of male mice. Platelet thrombus formation at the sites of laser-induced arteriolar wall injury in mice, before and after injection of TAT-K3PH-L (1 mg/kg), was visualized by labeling platelets with a fluorescence-conjugated anti-GPIbβ antibody and quantified. Data were collected from 15 injuries in 5 mice per group before and after TAT-K3PH-L treatment. C, Mice were pre-treated with TAT-K3PH-L (1 mg/kg) or vehicle buffer by tail vein injection for 15 min. A small segment of mouse tail (3 mm) was transected, and the bleeding time was monitored. The time to first cessation in bleeding was recorded for each tested mouse. ns, non-significant

3 |. DISCUSSION

Our data show that paxillin preferentially interacts with the PH domain of kindlin-3 while also binding to the PH domain of kindlin-1 or kindlin-2. The relatively weak interaction between paxillin and the PH domain of kindlin-2 might account for the controversial findings regarding their interaction in the literature.^29,32 In addition to the PH domain, previous studies showed that the F0 subdomain in kindlins was also involved in mediating paxillin binding.^27,28 As a multidomain adaptor, paxillin has five N-terminal leucine-rich LD motifs and four C-terminal tandem double zinc finger LIM domains.³³ In a recent study, Zhu et al. identified that the LIM4 domain of paxillin was responsible for binding to the F0 subdomain of kindlin-2.²⁹ In this study, we demonstrate that the LIM3 domain but not the LIM4 domain of paxillin interacts with the PH domain of kindlin-3, which is consistent with a previous finding for kindlin-2.³² Collectively, these findings suggest that both the PH domain and the F0 subdomain of kindlins may participate in interaction with paxillin by recognizing its LIM3 domain and LIM4 domain, respectively, showing that the paxillin/kindlin interaction involves multiple binding sites in each of them.²⁷

An interesting question is how the PH domain in kindlin-3 is involved in binding to paxillin in platelets at different activation stages. To answer this question, we employed a peptide (K3PH-L) derived from the β1–β2 loop in the PH domain of kindlin-3 that possesses the ability to inhibit the interaction between paxillin and the PH domain of kindlin-3. We noticed that K3PH-L was only able to partially inhibit the interaction in pull-down assays even using higher concentrations (data not shown), indicating that structural support from the intact PH domain or some flanking residues may be required to optimize their interaction. Thus, further structural study is required to explore more interaction details. Nonetheless, the comparably less inhibitory effect of the M2 mutant peptide verifies the specificity of K3PH-L. For performing functional analysis, we conjugated these peptides with a member-permeable sequence (TAT). Compared to the M2 mutant peptide, TAT-conjugated K3PH-L was able to suppress paxillin binding to kindlin-3 in platelets adherent to immobilized fibrinogen while it had no apparent inhibition in either resting or statically active platelets (Figure S2 in supporting information). These observations suggest that the PH domain-mediated kindlin-3 binding to paxillin in platelets may preferentially occur during out-side-in signaling.

As previously known, the PH domain of kindlin-3 is responsible for binding to phospholipids to mediate membrane recruitment of kindlin-3.^39,43,44 Interestingly, we find that binding of phospholipids and paxillin to the PH domain of kindlin-3 is mutually exclusive, suggesting that these binding events may occur sequentially in platelets. Disruption of the PH domain of kindlin-3 binding to membrane phospholipids suppressed the bidirectional signaling of integrin αIIbβ3 in platelets, indicating that membrane recruitment of kindlin-3 via its PH domain is essential for supporting integrin αIIbβ3 activation in platelets. This mechanism has also been evidenced in neutrophils for supporting β2-integrin activation.⁴³ However, specific disconnection of the PH domain of kindlin-3 with paxillin in platelets only affected integrin αIIbβ3 outside-in signaling but not inside-out signaling, implying that this interaction may play a role in facilitating the transition of integrin αIIbβ3 signaling from inside-out toward the outside-in direction.

Based on these findings, we propose that when platelets encounter stimulation, recruitment of kindlin-3 to plasma membrane via its PH domain is a prerequisite for supporting integrin αIIbβ3 activation. Theoretically, enrichment of kindlin-3 near the plasma membrane can promote its interaction with the integrin β3 CT, thus allowing cooperation with the talin head domain to induce integrin αIIbβ3 conformational changes and/or clustering for enhancing ligand binding.^10,41 Meanwhile, kindlin-3 may recruit paxillin via its F0 subdomain binding to the LIM4 domain of paxillin, which may further facilitate talin binding to the integrin β3 CT by forming a multiprotein complex.^28,45 Upon ligand occupation, integrin αIIbβ3 outside-in signaling is initiated and additional cytoplasmic/cytoskeletal signaling adaptors are recruited to the integrin β3 CT, during which paxillin binding to the PH domain of kindlin-3 via its LIM3 domain may occur and play a key role in supporting integrin αIIbβ3 outside-in signaling in platelets (Figure S2). In fact, paxillin binding to the PH domain of kindlin-3 may detach kindlin-3 from the plasma membrane due to their mutually exclusive binding properties, which might play a role in triggering the switch of integrin αIIbβ3 signaling from the inside-out to the out-side-in direction. Obviously, more structural/functional studies are required for mechanistic understanding of these interactions in platelets.

The TAT-conjugated K3PH-L peptide (membrane-permeable) derived from the paxillin binding region in the PH domain of kindlin-3 was capable of inhibiting integrin αIIbβ3 outside-in signaling but it had no significant effect on integrin αIIbβ3 inside-out signaling in platelets, which is consistent with the observation that TAT-conjugated K3PH-L specifically suppressed paxillin binding to kindlin-3 in adherent platelets but not in resting or statically active platelets (Figure S2). In addition, we detected that, in adherent platelets, this inhibitory peptide was able to compromise activation of the p21-activated kinase (PAK) and p38 MAPK (Figure S3 in supporting information), two kinases that play important roles in supporting platelet spreading and clot retraction.^46–49 Interestingly, this peptide had no significant effect on paxillin phosphorylation. Due to the complexity of multi-protein communications proximal to the integrin αIIbβ3 cytoplasmic tails in adherent platelets, the underlying mechanistic details need to be further investigated. Importantly, treatment of mice with this inhibitory peptide efficiently suppressed thrombus formation while having no significant effect on hemostasis. The functional specificity of this peptide on thrombosis but not hemostasis may be due to its inhibitory effect preferentially on integrin αIIbβ3 outside-in signaling but not inside-out signaling. If so, targeting the interaction between paxillin and the PH domain of kindlin-3 may serve as a useful strategy for developing antiplatelet and antithrombotic drugs.

In summary, in this study we disclose an important interaction between kindlin-3 and paxillin that is specifically required for supporting integrin αIIbβ3 outside-in signaling in platelets, which not only furthers our understanding of integrin αIIbβ3 signaling but also provides a useful target for developing novel antithrombotic strategies.

4 |. EXPERIMENTAL PROCEDURES

4.1 |. Plasmid constructs

For mammalian expression, the cDNA fragments of paxillin, Hic-5, leupaxin, and the PH domain of kindlin-3 were subcloned into p3×flag-CMV-10 vector and pEGFP-C2, respectively. For bacterial expression, the PH domains of kindlin-1 (374–496), kindlin-2 (374–499), and kindlin-3 (351–476) were subcloned into pGST-parallel1 vector; the full-length paxillin was cloned into pMBP-parallel1 vector; and the LIM3 domain and the LIM4 domain of paxillin were also subcloned into pGST-parallel1 vector. EGFP-fused kindlin-3 and the mutants were constructed into lentiviral vector pLeGo-G2 for generating lentiviral particles. The kindlin mutants were generated using the Site-Directed Mutagenesis Kit (Agilent) and further verified by DNA sequencing.

4.2 |. Recombinant proteins and peptides

Recombinant proteins were expressed in Rosetta2 (DE3) E. coli by incubating with isopropyl β-d-1-thiogalactopyranoside (1 mM) overnight at 18°C. The bacterial pellets were harvested by centrifugation and lysed to collect the supernatants. The GST-fused proteins were purified using glutathione Sepharose 4B column. The maltose binding protein (MBP)-fused paxillin was purified using MBPTrap HP column and the MBP tag was cleaved by TEV protease. Purified proteins were dialyzed against 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 10% glycerol and stored in a freezer in aliquots. Peptides used in this study were synthesized, purified (>98%), and dissolved in dimethyl sulfoxide at 10 mM concentration for storage, and further diluted to the final working concentration for experiments.

4.3 |. Pull-down assays

Chinese hamster ovary (CHO) cells transiently expressing flag-tagged paxillin were lysed using CelLytic Cell Lysis Reagent (Sigma-Aldrich) in the presence of protease inhibitor cocktail (Roche). The lysates were incubated with protein A/G beads (Santa Cruz Biotech) pre-coupled with an anti-flag antibody (M2) for 2 h at room temperature. After washing, the beads were incubated with purified GST or GST-fused kindlin-PH for 2 h at room temperature. After that, the beads were washed three times using phosphate-buffered saline (PBS), and the precipitates on the beads were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by western blotting. Alternatively, purified GST or GST-fused kindlin-PH proteins were incubated with glutathione Sepharose 4B beads for 1 h at 4°C. After the wash step, the beads were incubated with CHO cell lysates expressing flag-tagged paxillin. Then the beads were washed again, and the precipitates on the beads were analyzed by SDS-PAGE followed by western blotting.

4.4 |. Biolayer interferometry binding assay

The binding kinetics of paxillin with the kindlin-PH domains were measured by biolayer interferometry using an Octet Red96 system (ForteBio). Microplate wells used were filled with 250 μl of samples and agitated at 1000 rpm at 30°C. Experiments were performed in the assay buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM ZnCl₂, 0.02% Tween 20, and 0.1% bovine serum albumin (BSA). An anti-flag antibody was immobilized on AMC biosensor tips (ForteBio) by dipping the biosensors into antibody solution (3 μg/ml) for 5 min. After washing, the biosensors coupled with the anti-flag antibody were dipped into diluted CHO cell lysate expressing flag-tagged paxillin. The biosensors loaded with flag-paxillin were washed with the assay buffer to remove non-specific binding. To measure association rate (k_on), the biosensors were dipped into the assay buffer containing GST-fused kindlin-PH or GST alone (analytes), during which the protein--protein association can increase the optical thickness of the biological layer at the biosensor tips and lead to a real-time shift in wavelength (Δλ). Six concentrations (3, 7.5, 10, 15, 20, and 30 μM) for each of the purified proteins were used for testing. Next, the associated proteins were allowed to disassociate (k_off) from the biosensors in the assay buffer. The changes of thickness of the biosensor layer during both association and dissociation phases were recorded. Interaction of each couple of proteins was quantitated to provide accurate data on the association/dissociation rates. Kinetic parameters (k_on and k_off) and affinity (K_D) were determined using the Octet data analysis software ver.9.1 (ForteBio). All measurements were corrected by subtracting the baseline contributed by the loaded biosensors in assay buffer. To confirm the mutually exclusive binding of phospholipids and paxillin to kindlin-3-PH, GST-fused kindlin-3-PH was pre-incubated with PtdIns(4,5) P₂, PtdIns(3,4,5)P₃, or PtdIns. To examine the interaction between kindlin-3-PH and the LIM domains of paxillin, EGFP-fused kindlin-3-PH was immobilized on the biosensors and GST-fused LIM3 or LIM4 of paxillin was prepared as an analyte.

4.5 |. Protein-lipid overlay assay

To determine binding of kindlin-3-PH and its mutants to phospholipids, the protein-lipid overlay assay was performed using a PIP array (Echelon Biosciences), which is a hydrophobic membrane spotted with a concentration gradient of different phosphoinositides. In brief, the membrane was first blocked with PBS buffer supplemented with 0.1% v/v Tween-20 and 3% BSA for 1 h at 22°C, and then GST-fused kindlin-3-PH (1 μg/ml) was added to incubate with the membrane for 1 h at 22°C. After incubation, the membrane was washed in PBS three times. The signal of kindlin-3-PH binding to the spotted phosphoinositides was evaluated by western blotting using an anti-GST antibody. To examine if paxillin inhibits kindlin-3-PH binding to phospholipids, paxillin was used to incubate with GST-fused kindlin-3-PH (molar ratio 10:1) before adding to the membrane.

4.6 |. Lentiviral transduction and bone marrow transplantation

Kindlin-3^fl/flMx1-Cre mice were intraperitoneally injected with poly(I:C) at 6–8 weeks to induce Mx1-cre-mediated kindlin-3 deletion in hematopoietic cells. Sca1+ bone marrow cells were collected from these mice for lentiviral transduction to rescue the expression of kindlin-3 or the mutants fused with an N-terminal EGFP tag. EGFP-positive bone marrow cells were injected into lethally irradiated wild-type C57BL/6 male recipient mice. Functional analyses of platelets in these mice were performed at 8 weeks after the transplantation. Lentiviral particles used to express EGFP-fused kindlin-3 or the mutants were prepared in the Viral Laboratory at Versiti Blood Research Institute. All animal experiments were approved by the Institutional Animal Care and Use Committee.

4.7 |. Functional analysis of mouse platelets

Mouse platelets were isolated from peripheral blood of transplanted mice, washed, and diluted in Tyrode’s buffer supplemented with 1 mM Ca²⁺ and 1 mM Mg²⁺. To measure integrin αIIbβ3 activation (inside-out signaling), mouse platelets were incubated with Alex Fluor 647-conjugated fibrinogen for 25 min at room temperature in the absence or presence of 150 μM of PAR4 agonist peptide (AYPGKF) or 5 μg/ml of CRP, followed by flow cytometry analysis. To determine integrin αIIbβ3-mediated platelet spreading (outside-in signaling), washed mouse platelets were allowed to spread on slides pre-coated with fibrinogen in the presence or absence of PAR4 agonist peptide (150 μM) for 60 min. The spreading areas of EGFP-positive platelets were imaged and analyzed using Image J software.

4.8 |. Functional analysis of human platelets

The study with human blood samples was approved by the Institutional Review Board of Versiti Blood Research Institute. Washed human platelets were prepared from citrated blood samples donated by healthy volunteers and suspended in Tyrode’s buffer supplemented with 1 mM CaCl₂ and 1 mM MgCl₂. Platelets were treated with 1 μM of each of the indicated peptides or vehicle (as a control) for 1 h at 22°C before the following examinations. (1) Soluble fibrinogen binding: Treated platelets were incubated with Alexa Fluor 647-conjugated fibrinogen for 25 min at 22°C in the absence or presence of either 5 μM of TRAP-6 or 1 μg/ml of CRP followed by flow cytometry analysis. (2) Platelet aggregation: Platelet aggregation assays were performed by Born Lumi-Aggregometer (Chrono-log Corporation) at 37°C with a stir speed of 1200 rpm. After stimulation with either TRAP-6 (5 μM) or CRP (1 μg/ml), the aggregation signal was monitored for 6 min. (3) Platelet spreading: Treated platelets were allowed to spread on slides pre-coated with fibrinogen in the presence of either TRAP-6 (5 μM) or CRP (1 μg/ml) for 45 min at 37°C. Adherent platelets were fixed, permeabilized, stained with Alexa Fluor 568 phalloidin, and photographed under a fluorescent microscope. The spreading areas of platelets were analyzed using Image J software. (4) Clot retraction: Platelet-rich plasma was incubated with 1 μM of each of the indicated peptides or vehicle for 1 h at 22°C followed by adding 20 mM CaCl₂ and one of the agonists (5 μM TRAP-6 or 1μg/ml CRP) to initiate clot retraction. The clots were allowed to retract at 37°C for 2 h and the images were analyzed using Image J software.

4.9 |. Laser injury-induced thrombus formation in cremaster arterioles of mice

Intravital imaging of platelet thrombus formation in mouse cremaster arterioles was performed as previously described.²⁸ Briefly, male mice were anesthetized and the body temperature was maintained at 37°C using a thermo-regulated heating pad. The cremaster muscle was exposed and superfused with warm saline (37°C) during the experiment. Three control arteriolar wall injuries, created by a laser ablation system (Intelligent Imaging Innovations), were first recorded in each mouse followed by tail vein injection of TAT-K3PH-L (1 mg/kg). After 15 min, more arteriolar wall injuries were made. Circulating platelets were labeled with a fluorescent anti-GPIbβ antibody (Emfret). Fluorescence images were captured using a high-speed camera (Orca-Flash4.0, Hamamatsu) for at least 5 min following the vessel injury.

4.10 |. Mouse tail-bleeding assay

Mice were anesthetized and injected with TAT-K3PH-L peptide (1 mg/kg) or vehicle buffer through the tail veins. Fifteen min after injection, a small portion of tail tip (~3 mm) was amputated and the tail was immediately immersed into sterile PBS buffer (37°C). Bleeding time was calculated from the start of bleeding to the first cessation of the bleeding. The wounded tail was cauterized to terminate the bleeding. All the animal experiments were approved by the Institutional Animal Care and Use Committee.

4.11 |. Statistics

P values were calculated by the 2-tailed Student’s t test. One-way ANOVA with post hoc tests was performed for multiple comparisons.

Essentials.

• Paxillin directly interacts with the pleckstrin homology (PH) domain of kindlin-3 (kindlin-3-PH).

• Paxillin/kindlin-3-PH interaction supports integrin αIIbβ3 outside-in signaling in platelets.

• An inhibitory peptide derived from kindlin-3-PH suppresses αIIbβ3 outside-in signaling.

• An inhibitory peptide derived from kindlin-3-PH possesses an antithrombotic effect.