Kindlin-3 Mediates Integrin αLβ2 Outside-in Signaling, and It Interacts with Scaffold Protein Receptor for Activated-C Kinase 1 (RACK1)*

Background: Kindlin-3 is a cytoplasmic protein that binds and modulates the ligand binding property of integrin αLβ2. Results: Kindlin-3 induces integrin αLβ2 clustering, and it interacts with the scaffold protein RACK1. Conclusion: Kindlin-3 is involved in integrin αLβ2 outside-in signaling. Significance: This study presents important findings in understanding the role of kindlin-3 in integrin signaling. Integrins are heterodimeric type I membrane cell adhesion molecules that are involved in many biological processes. Integrins are bidirectional signal transducers because their cytoplasmic tails are docking sites for cytoskeletal and signaling molecules. Kindlins are cytoplasmic molecules that mediate inside-out signaling and activation of the integrins. The three kindlin paralogs in humans are kindlin-1, -2, and -3. Each of these contains a 4.1-ezrin-radixin-moesin (FERM) domain and a pleckstrin homology domain. Kindlin-3 is expressed in platelets, hematopoietic cells, and endothelial cells. Here we show that kindlin-3 is involved in integrin αLβ2 outside-in signaling. It also promotes micro-clustering of integrin αLβ2. We provide evidence that kindlin-3 interacts with the receptor for activated-C kinase 1 (RACK1), a scaffold protein that folds into a seven-blade propeller. This interaction involves the pleckstrin homology domain of kindlin-3 and blades 5–7 of RACK1. Using the SKW3 human T lymphoma cells, we show that integrin αLβ2 engagement by its ligand ICAM-1 promotes the association of kindlin-3 with RACK1. We also show that kindlin-3 co-localizes with RACK1 in polarized SKW3 cells and human T lymphoblasts. Our findings suggest that kindlin-3 plays an important role in integrin αLβ2 outside-in signaling.

Integrins are a large family of cell adhesion molecules that have important functions in immunity, wound healing, hemostasis, and the development of metazoans (1). Besides mediating cell-cell or/and cell-extracellular matrix interactions, intracellular signals that are derived from integrins regulate cell proliferation, apoptosis, and differentiation (2). In humans there are 24 integrins, and each of these is a heterodimer formed by an ␣ subunit and a ␤ subunit that are non-covalently associated (1). Each subunit contains a large ectodomain and a transmembrane domain followed by a short cytoplasmic tail. Integrins are not enzymes, but they are transducers of bi-directional signals (1). Inside-out signaling involves intracellular signaling events triggered by cell surface receptors, for example chemokine receptor, that regulate the ligand binding affinity of the integrins (3,4), whereas outside-in signaling involves signaling pathways derived from ligand-bound integrins and/or integrin clustering.
The integrin cytoplasmic tails mediate both inside-out and outside-in signaling events. The integrin tails are docking sites for cytoskeletal and signaling proteins (4,5). The integrin ␤ tails contain two highly conserved NXX(Y/F) motifs. Talins (1 and 2) are large cytoskeletal proteins, each containing a 4.1-ezrin-radixin-moesin (FERM) 2 -containing head domain (HD) and a helical rod region (6). The talin FERM domain is composed of F1, F2, and F3 subdomains, and the latter contains a phosphotyrosine binding fold that allows talin to bind to the membrane proximal NXX(Y/F) motif in the integrin ␤ tails, leading to the separation of the integrin ␣ and ␤ tails (7)(8)(9)(10). This triggers further structural changes in the integrin that involves unpacking the transmembrane domains and unfolding of the ectodomain, leading to an increase in the ligand binding affinity of the integrin (11,12).
Kindlins (1, 2, and 3) are another family of FERM-containing cytoplasmic proteins that have been reported to regulate integrin ligand binding affinity (13)(14)(15). The F3 subdomain of kindlins binds to the membrane-distal NXX(Y/F) motif in the integrin ␤ tails (13,15,16). Each kindlin also contains an F0 and a pleckstrin homology (PH) domain (17,18). The F0 domain has an ubiquitin-like fold that targets kindlin-1 to integrin ␣IIb␤3containing focal adhesion sites (18). The PH domain has been shown to facilitate kindlin-2 binding to membrane phosphatidylinositol 3,4,5-trisphosphate (19). The detailed mechanism by which kindlin regulates the ligand binding affinity of integrins remains to be clarified. Besides integrins, the cytoplasmic * This work was supported by Nanyang Technological University AcRF Grant molecules integrin-linked kinase and migfilin have been reported to interact with kindlin-2 (20,21). Migfilin has been shown to be a positive regulator of integrin activation because it disrupts the binding of cytoskeletal protein filamin A to the integrin ␤ tail (22,23). Filamin A is a negative regulator of integrin activation because it competes with talin for binding the integrin ␤ tail (24). Hence, kindlin-migfilin interaction promotes integrin activation.
Kindlin-3 (UNC112-related protein 2 or FERMT3) is expressed in platelets, hematopoietic cells, and endothelial cells (25,26). Its importance is underscored by the disease leukocyte adhesion deficiency III in which defective kindlin-3 expression abrogated the functions of platelet ␣IIb␤3 and leukocyte ␣L␤2 integrins, leading to bleeding disorders and a compromised immune system in patients, respectively (27)(28)(29)(30). Mice with ablated KINDLIN-3 phenocopied leukocyte adhesion deficiency III but also had severe osteopetrosis and structural defects in the membranes of their erythrocytes (14,15,31). KINDLIN-3 Ϫ/Ϫ platelets adhered but failed to spread on immobilized fibrinogen despite activation of integrin ␣IIb␤3 by exogenous MnCl 2 which bypassed insideout signaling (15). This suggests that kindlin-3 is also required for integrin outside-in signaling, and it may interact with cytoplasmic protein(s) that modulate cell spreading.
Here we show that kindlin-3 is involved in integrin ␣L␤2 outside-in signaling, and it induces integrin ␣L␤2 micro-clustering. We also show that kindlin-3 can interact with the receptor for activated-C kinase 1 (RACK1), a small (36 kDa) sevenblade propeller WD-repeat scaffold protein with multiple cytosolic binding partners (32), including the integrin ␤ cytoplasmic tails (33)(34)(35). Using the human T lymphoma cell line SKW3, we show that integrin ␣L␤2 engagement by its ligand ICAM-1 promotes kindlin-3 and RACK1 interaction. Finally, we observed that kindlin-3 and RACK1 co-localized in polarized SKW3 cells and human T lymphoblasts that adhered to immobilized ICAM-1.
We have also generated rat anti-kindlin-3 mAbs clone 9 and clone 229A using recombinant full-length human kindlin-3 as the immunogen.
Generation of Stable K562 Cells with Reduced Kindlin-3 Expression-Stable K562 cells with reduced kindlin-3 expression were generated using a third generation lentiviral-based siRNA transduction system according to the manufacturer's instructions (Applied Biological Materials). Briefly, four kindlin-3-targeting piLenti-RNAi-GFP plasmids were generated. Each of these plasmids or the control siRNA encoding plasmid was co-transfected with the packaging plasmids into 293T cells to allow production of pseudo-virions, and the culture supernatant was collected. K562 cells (1 ϫ 10 6 ) were resuspended in 1 ml of complete RPMI 1640 medium supplemented with 6 g/ml Polybrene (Sigma) followed by the addition of 1 ml of the pseudo-virions-containing supernatant, and the cells were cultured for 1 day. Cells were then spun down and cultured in complete RPMI 1640 medium for at least 2 days. GFP expression in cells was visualized under a fluorescence microscope. Thereafter, cells stably expressing kindlin-3 siRNA were selected in the presence of puromycin (1.5 g/ml). Only one of the four kindlin-3-targeting piLenti-RNAi-GFP plasmids produced K562 clones that showed significant reduction in kindlin-3 expression. The sequence that encodes this kindlin-3 siRNA is TGGAGCAGATCAATCGCAA, which has been previously described (29).
Flow Cytometry Analyses-Flow cytometry analyses of cells were performed as previously described (39). Integrin transfectants were incubated with 20 g/ml each of mAb MHM24 (for ␣L␤2 or ␣L␤2N329S) or mAb 10E5 (for ␣IIb␤3 and ␣IIb␤3N339S) in medium on ice for 30 min. Cells were washed in medium and incubated in medium containing APC-conjugated goat anti-mouse IgG (BD Pharmingen) (1:400 dilution) on ice for 30 min. Sample acquisitions were performed on a FACSCalibur flow cytometer (BD Biosciences). Data were analyzed using the Flowjo software (Tree Star Inc. Ashland, OR).
Surface Plasmon Resonance Analyses-Sensor chip CM5, N-hydroxysuccinimide (NHS), N-ethyl-N-(3-dimethylaminopropyl)-carbodiimide hydrochloride (EDC), ethanolamine, and surfactant P20 were purchased from Biacore, GE Healthcare. Surface plasmon resonance assays were run at 30 l/min at 37°C on a Biacore 3000 system. Standard HBS-EP buffer (150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant P20, 50 mM HEPES, pH 8.0) was used for the analysis. Sensorgrams produced from Tris buffer used in protein purification were similar to those from HBS-EP buffer. GST-RACK1 was immobilized to 5000 response units onto sensor chip CM5 using a standard amine-coupling procedure as previously described (40). GST was also immobilized to the same level onto another flow cell as control. Kindlin-3-His 6 was verified not to interact with immobilized control GST. 2-Fold serially diluted kindlin-3-His 6 (1.56, 3.125, 6.25, 12.5, 25, 50 and 100 M) were separately injected across flow cells comprising the control GST surface or GST-RACK1 for 6 min, allowed to dissociate for 5 min, and regenerated with a short pulse (30 s) of 15 mM HCl. All raw sensorgrams were double-referenced (41) by subtracting all resultant sensorgrams with the con-trol GST surface (to eliminate nonspecific binding) and blank buffer injections (to eliminate equipment systematic error) and then globally fit to a simple bimolecular interac-tion model. Three replicates of the interaction were performed (with different batches of proteins) with similar results.
Pulldown Assays-GST, GST-RACK1, and GST-RACK1 W5-7 were affinity-purified on glutathione S-Sepharose 4B columns. The GST protein-bound beads were incubated in buffer A (150 mM NaCl, 10% (v/v) glycerol, 50 mM Tris, pH 8.0) containing 5% (w/v) BSA at 4°C for 1 h to block nonspecific binding sites. The beads were subjected to four washes in buffer A before use. Different amounts of purified kindlin-3 or its mutants were incubated with these beads in buffer A at 4°C overnight. Beads were subjected to four washes in buffer A, and bound proteins were eluted by boiling the beads in SDS-PAGE sample buffer containing DTT.
The N-terminal biotin-conjugated full-length integrin ␤2 tail peptide (Lys-702-Ser-747) and its mutant (N741A) were synthesized, purified, and ESI-MS-verified by ChinaPeptides, Shanghai, China. The peptide (100 g each) was incubated with 50 l of streptavidin-agarose beads suspension (Sigma) in buffer C (150 mM NaCl and 50 mM Tris, pH 8.0) at 4°C for 1 h. The beads were washed and incubated in the same buffer containing 0.5% (w/v) BSA at 4°C for 1 h to block nonspecific binding sites. Thereafter, washed beads were incubated with indicated amount of purified recombinant kindlin-3-His 6 or His 6 -RACK1.
FRET Analyses-COS-7 transfectants expressing CFP and YFP-RACK1, CFP-kindlin-3 and YFP-RACK1, CFP-kindlin-3 PH⌬ and YFP-RACK1, or CFP-kindlin-3 F3⌬ and YFP-RACK1 were seeded into coverslip glass-bottom culture dishes and cultured for 24 h. FRET analyses were performed as previously described (42, 43) on a Zeiss LSM510 confocal laser scanning microscope (Carl Zeiss, Inc., Thornwood, NY) to detect the interaction between RACK1 and kindlin-3 or mutants. The nucleus of the cell was excluded from the analyses. FRET efficiency was calculated as a percentage using the equation FRET efficiency ϭ (I 6 Ϫ I 5 ) ϫ 100/I 6 , where I n is the CFP intensity at the nth time point. Bleaching of YFP was performed between the fifth and sixth time points. For FRET analyses of integrin ␣L␤2 clustering, K562 cells were transfected with HA-tagged kindlin-3 or its mutants with ␣LmCFP, ␣LmYFP, and ␤2. Only the plasma membrane was selected as the region of interest for measurements of CFP signal before and after YFP photobleaching.
Real-time Electric Cell-Substrate Impedance Sensing Measurements-Dithiobis succinimidyl propionate (Pierce, Thermo Fisher Scientific, Rockford, IL) (4 mg/ml) in DMSO (40 l) was added to each well of a 16-well E-plate device (Acea Biosciences, Inc., San Diego, CA) with gold electrodes at the bottom of each well and incubated for 30 min at RT.
Wells were washed twice in deionized H 2 O. For the integrin ␣L␤2N329S assay, goat anti-human Fc (Sigma) (10 g/ml) in PBS was added to each well and incubated for 1 h at RT. Nonspecific binding sites were blocked with 0.1% (w/v) BSA in PBS for 15 min at RT. Wells were coated with 2 g/ml ICAM-1 Fc in PBS (50 l per well) and incubated for 2 h at RT. For the integrin ␣IIb␤3N339S assay, 1 g/ml fibrinogen (Sigma) in PBS was added to each well and incubated for 1 h at RT followed by blocking with 0.1% (w/v) BSA in PBS for 15 min at RT. Wells were washed once in RPMI 1640 complete medium, and each well was refilled with 100 l of medium. Background scans of the wells were performed on a Real Time Cell Electronic System TM (Acea Biosciences, Inc.). K562 transfectants (8 ϫ 10 4 cells per well) were seeded into each well, and AC impedance (cell index) measurements taken at 1-min intervals. Cell adhesion and spreading mediated by integrins ␣L␤2N329S and ␣IIb␤3N339S were verified by having function-blocking mAbs MHM24 and 10E5 (10 g/ml each), respectively. For clarity, cell index is plotted as a function of time at 5-min intervals.
Immunofluorescence Staining and Imaging-A coverslip glass-bottom tissue culture dish (MatTek, Ashland, MA) was coated with goat anti-human IgG Fc (5 g/ml) in bicarbonate buffer at 37°C for 1 h. Nonspecific binding sites were blocked with 0.5% (w/v) BSA in PBS at 37°C for 30 min. The culture dish was then coated with ICAM-1-Fc (1 g/ml) in PBS at 4°C overnight. T cells were resuspended in RPMI complete medium containing SDF-1␣ (100 ng/ml) (EMD4 Biosciences, Gibbstown, NJ), seeded into the dish, and incubated in a cell culture incubator for 15 min. Medium was discarded, and the cells were fixed in 3.7% (w/v) paraformaldehyde in PBS at RT for 10 min. Fixed cells were washed once in cytoskeleton stabilization buffer (100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 1 mM EGTA, 10 mM PIPES, pH 6.8) containing 0.3% (v/v) Triton X-100 followed by incubation in the same buffer at RT for 1 min. Cells were subjected to 3 washes in PBS, and nonspecific sites were blocked with 3% (w/v) BSA in PBS at RT for 30 min. Cells were incubated in PBS containing relevant primary antibodies at RT for 1 h. The primary antibodies used were mouse anti-RACK1 mAb (clone B-3), rat anti-kindlin-3 mAb (clone 9), and rabbit anti-MyH9 polyclonal antibody. Cells were then washed in PBS. For co-staining with actin, cells were incubated in PBS containing highly cross-adsorbed Alexa Fluor 594-conjugated goat anti-mouse IgG, donkey anti-rat IgG, or goat antirabbit IgG antibodies (1:10,000 dilution each) (Molecular FIGURE 1. Kindlin-3 mediates integrin ␣L␤2 outside-in signaling and promotes integrin ␣L␤2 micro-clustering. A, cell lysates from K562 stably expressing control siRNA or kindlin-3 (K3)-targeting siRNA were subjected to immunoblotting to assess the expression levels of kindlin-3 and other cytoplasmic proteins. Actin was used as loading control. B, K562 cells were successfully and stably transduced with the lentiviral-based siRNA with GFP as the reporter was transfected with integrin mutant ␣L␤2N329S. Expression level of integrin ␣L␤2N329S was determined by staining transfectants with mAb MHM24 followed by APC-conjugated secondary antibody. Two-color flow cytometry analyses were performed. FL1 was for GFP expression, and FL4 was for integrin ␣L␤2N329S expression. Expression index (EI) was calculated by the percentage of cells gated positive (GP) ϫ geo-mean fluorescence intensity (GM). C, shown is electric cell-substrate impedance sensing measurements of these K562 transfectants on immobilized ICAM-1. The function-blocking mAb MHM24 was included to demonstrate integrin ␣L␤2-mediated binding specificity. Each data point is the mean Ϯ S.D. (gray bar) of technical triplicates. A representative plot of three independent experiments is shown. Measurements were taken at 1-min intervals, but for clarity cell index is plotted as a function of time at 5-min intervals. D, the drawing illustrates the principle of FRET-based detection of integrin ␣L␤2 micro-clustering using the ␣L subunit having a C-terminal fusion of mCFP or mYFP as previously reported (9,43). E, shown are immunoblots of cell lysates of K562 transfected with integrin subunits ␣LmCFP, ␣LmYFP, ␤2, and HA-tagged kindlin-3 constructs. F, shown are FRET analyses of integrin ␣L␤2 micro-clustering in K562 transfectants. Each data point represents the mean Ϯ S.D. of Ն40 cells analyzed. *, p Ͻ 0.05, Student's t test. A representative plot of three independent experiments is shown.
Probes, Invitrogen) with Alexa Fluor 488-conjugated phalloidin (0.27 ng/ml) and DAPI (0.1 g/ml) at RT for 1 h. Cells were then subjected to 3 washes in PBS. Images were acquired on a Zeiss LSM 510 confocal laser scanning microscope with a 63ϫ oil objective lens. Images were analyzed using the LSM 510 META software Version 3.2 SP2. Thermal LUT surface plots (intensity plots) were generated using the Image J software.
See supplemental Experimental Procedures for expression plasmids, cell culture and transfection, immunoprecipitation assays and immunoblotting, and expression and purification of recombinant proteins from Escherichia coli.

Kindlin-3 Is Required for Integrin ␣L␤2
Outside-in Signaling-Kindlins are co-activators, together with talin, of integrins. Kindlin-3 is well reported to be important for ␤2 integrin-mediated leukocyte adhesion and spreading (15,25,28,30,44). However, direct evidence of kindlin-3 in ␤2 integrin outside-in signaling is lacking. To this end we made use of a lentiviralbased siRNA transduction system with GFP reporter to generate stable K562 cell line that expressed either siRNA-targeting kindlin-3 or control siRNA. Cells transduced with siRNA-targeting kindlin-3 showed reduced kindlin-3 expression compared with cells transduced with control siRNA (Fig. 1A). The expression levels of cytoplasmic proteins RACK1, FAK, talin, and PKC␤ in siRNA-targeting kindlin-3 and control siRNA cells were comparable. K562 cells do not express endogenous ␤2 integrins. K562 cells were subsequently transfected with the integrin mutant ␣L␤2N329S, which has been shown previously to be activated constitutively (45). The rationale to use a constitutively activated mutant is to bypass the inside-out activation process that requires kindlin-3. This allows us to examine the role of kindlin-3 in integrin ␣L␤2 outside-in signaling in K562 cells with reduced kindlin-3 expression. The expression levels of integrin ␣L␤2N329S on the transfectants were comparable based on mAb MHM24 staining and flow cytometry analyses (Fig. 1B). Both transfectants were also GFP-positive compared with wild-type K562 cells that were not transduced with the lentiviral siRNA.
We performed real-time electrical cell-substrate impedance sensing measurements (46) of these cells on immobilized ICAM-1. Impedance sensing is reported as cell index herein. Cell index increases when cells adhere and spread on the electrode. The cell index of control siRNA cells expressing integrin ␣L␤2N329S increased markedly over time, whereas that of kindlin-3 siRNA cells expressing integrin ␣L␤2N329S was marginal (Fig. 1C). The specificity of integrin ␣L␤2N329S-mediated adhesion and spreading was demonstrated by including the function-blocking mAb MHM24 (anti-␣L). These data suggest that kindlin-3 is involved in integrin ␣L␤2 outside-in signaling as K562 cells, with reduced kindlin-3 expression failed to adhere and spread effectively despite expressing constitutively activated integrin ␣L␤2. The generality of kindlin-3 in mediating integrin outside-in signaling was also shown using cells that expressed constitutively activated integrin ␣IIb␤3N339S (45) (supplemental Fig. S1).
These data suggest that kindlin-3 could induce integrin ␣L␤2 micro-clustering and both its PH and F3 domains are required for this function. The F3 domain is required for kindlin-3 binding to the integrin ␤ tail, accounting for the lack of integrin ␣L␤2 micro-clustering in cells overexpressing HA-kindlin-3F3⌬ (13,15,16). Although it remains to be shown that kindlin-3 PH domain has similar phosphatidylinositol phosphate(s) binding properties to kindlin-2 (19), it is possible that the PH domain allows kindlin-3 to tether at membrane microdomains enriched in phosphatidylinositol phosphate(s), thereby promoting integrin ␣L␤2 micro-clustering.
Kindlin-3 Interacts with Scaffold Protein RACK1-The seven-blade propeller WD-repeat scaffold protein RACK1 has been shown to bind integrin ␤ tails, including the integrin ␤2 tail (33,35). RACK1 blades 5-7 apparently binds the membrane proximal sequence Lys-702-His-706 of the integrin ␤2 tail, and it co-immunoprecipitated with integrin ␣L␤2 from lysate of phorbol 12-myristate 13-acetate-treated JY cells (33). It has been reported that RACK1 localized at the site in Jurkat T cell that was in contact with chemokine SDF-1-coated latex bead and at the migratory front of formylmethionylleucylphenylalanine-treated HL-60 cells (47). RACK1 has also been shown to bind PH domains (48). These data prompted us to examine the association of kindlin-3 with RACK1.
The co-immunoprecipitation assay was performed to examine the interaction of endogenous RACK1 with ectopically expressed HA-kindlin-3 (wild type and full-length) or its mutants or HA-kindlin-2 (wild type and full-length) in 293T cells ( Fig. 2A). Kindlin-2 was included in the assay because it has been reported to be important in cell adhesion and spreading in many biological processes, including angiogenesis, cancer, and wound healing (49 -51). RACK1 co-precipitated with HAkindlin-3 and HA-kindlin-3 F3⌬ but not HA-kindlin-2 and HA-kindlin-3 PH⌬. The lack of interaction between kindlin-2 and RACK1 prompted us to perform pulldown assays using recombinant purified GST-RACK1, kindlin-2-His 6 , and kindlin-3-His 6 (supplemental Fig. S2). The results from the pull-down assays are consistent with the co-immunoprecipitation data.
Next, we performed FRET-based interaction studies in cells. COS-7 cells were transfected with YFP-RACK1 and CFP-kindlin-3, CFP-kindlin-2, or CFP empty vector followed by YFPphotobleach FRET analyses (Fig. 2B). A significant level of FRET was detected in cells expressing YFP-RACK1 and CFPkindlin-3 compared with others. COS-7 cells were also transfected with YFP-RACK1 and CFP-kindlin-3 or its mutants or CFP empty vector followed by FRET analyses (Fig. 2C). The levels of FRET detected are on the order of CFP-kindlin-3 Ͼ CFP-kindlin-3 F3⌬ Ͼ CFP-kindlin-3 PH⌬ Ͼ CFP. These observations suggest that the PH domain and to a lesser extent the F3 domain of kindlin-3 is involved in RACK1 interaction.
To obtain binding constants for the interaction between kindlin-3 and RACK1, we performed surface plasmon resonance by immobilizing GST-RACK1 on a CM5 sensor chip followed by injections of kindlin-3-His 6 at different concentrations (Fig. 2D). The experimental data (colored lines) conformed well to a simple bimolecular model (orange lines) with k a ϭ 0.2 ϫ 10 4 M Ϫ1 s Ϫ1 , k D ϭ 4.1 ϫ 10 Ϫ4 s Ϫ1 , K D ϭ 1.8 M. Three replicates using different batches of proteins gave an average affinity constant of 1.60 Ϯ 0.08 M.
In the next series of experiments, GST pulldown assays using recombinant purified proteins were performed to determine the interaction sites of kindlin-3 and RACK1. Consistent with the previous data, kindlin-3-His 6 and kindlin-3 F3⌬-His 6 , but not kindlin-3 PH⌬-His 6 , interacted with GST-RACK1 (Fig.  3A). No interaction between the kindlin-3 proteins with control GST was detected. RACK1 folds into a propeller containing seven blades, which are docking sites for different cytoplasmic proteins (32). To examine which of these blades of RACK1 interacts with kindlin-3, we attempted to express and purify these blades with an N-terminal GST tag. However, in our hands only GST-RACK1 blades 5-7 Gly-190 -Arg-317 could be expressed in soluble form (data not shown). Kindlin-3-His 6 was detected in pulldown samples of GST-RACK1 and GST-RACK1 W5-7 but not control GST (Fig. 3B).
We also attempted to express kindlin-3 PH domain to test its association with RACK1 W5-7. The yield of kindlin-3 PH domain in E. coli expression system was low, and it was prone to degradation. To circumvent this problem, we constructed and were able to express kindlin-3 PH domain with its flanking F2 subdomains (His 6 -kindlin-3 PH-F2), which associated with GST-RACK1 or GST-RACK1 W5-7 but not control GST (Fig. 3C). Taken together, these data suggest RACK1 has been reported to associate with the integrin ␤2 tail (34). We asked if the integrin ␤2 tail, RACK1, and kindlin-3 can form a ternary complex. To this end, N-terminal biotinlabeled full-length wild-type ␤2 or mutant ␤2 N741A integrin tails were conjugated to streptavidin-agarose beads. Kindlins have been reported to bind to the membrane distal NXX(F/Y) motif of the integrin ␤ tails (13, 52); hence, we included the ␤2 N741A mutant in which Asn of the membrane distal 741 NPKF motif in the ␤2 tail was replaced with Ala. Pulldown assays were performed using these integrin ␤2 tail-peptide-bound beads and different amounts of purified recombinant full-length kindlin-3-His 6 or His 6 -RACK1 (Fig. 3D, top panel). The association of kindlin-3-His 6 with wild-type ␤2 tail was concentration-dependent. By contrast, the level of kindlin-3-His 6 that associated with ␤2 N741A was diminished. The levels of His 6 -RACK1 with wild-type ␤2 tail and His 6 -RACK1 with N741A ␤2 tail were comparable, suggesting that RACK1 interaction with integrin ␤2 tail is not dependent on the membrane distal  Pulldown assays using recombinant kindlin-3 (K3) and RACK1 proteins. A, B, and C, shown are GST pulldown assays using purified recombinant proteins of K3 and RACK1 and their mutants. Immunoblotting (IB) of K3 was performed using commercial anti-kindlin-3 antibody. D, shown are pulldown assays using N-terminal biotin labeled wild-type ␤2 tail or ␤2 N741A integrin tail-peptide bound to streptavidin-agarose beads and purified recombinant K3-His 6 and His 6 -RACK1. Coomassie-stained gels of wild-type ␤2 or ␤2 N741A integrin tail peptide bound to streptavidin-agarose beads are shown. Top panel, shown are immunoblots of K3-His 6 or His 6 -RACK1 associated with wild-type ␤2 or ␤2 N741A integrin tail-peptide. Middle panel, shown is an illustration of ternary complex formation of integrin ␤2 tail, kindlin-3, and RACK1. Because kindlin-3 interacts with RACK1, a ternary complex of integrin ␤2 N741A, kindlin-3, and RACK1 could be formed. The dotted line denotes interaction. Bottom panel, shown are immunoblots of K3-His 6 associated with wild-type ␤2 or ␤2 N741A integrin tail-peptide in the presence of His 6 -RACK1. 741 NPKF motif. This is in line with the observation that RACK1 interacts with the membrane proximal sequence of the integrin ␤2 tail (33).

Kindlin-3 Mediates Integrin ␣L␤2 Outside-in Signaling
Next we asked whether integrin ␤2 tail, kindlin-3, and RACK1 can form a ternary complex. We incubated the integrin ␤2 tail-peptide-conjugated beads with the same amount of His 6 -RACK1 and with different amounts of kindlin-3-His 6 (Fig.  3D, bottom panel). The rationale is that if integrin ␤2 tail, kindlin-3, and RACK1 form a ternary complex, a ternary complex could still be formed using the integrin ␤2 N741A mutant because RACK1 could serve as the bridging molecule for kindlin-3 and integrin ␤2 N741A despite the fact that kindlin-3 interacts poorly with ␤2 N741A (Fig. 3D, schematic). Indeed, kindlin-3-His 6 was detected in both samples of ␤2 tail peptides in the presence of His 6 -RACK1, and the levels of association were comparable. These data suggest that the integrin ␤2 tail, kindlin-3, and RACK1 can form a ternary complex.
Interaction and Co-localization Analyses of Kindlin-3 and RACK1 in T Cells-To examine the interaction of endogenous kindlin-3 and RACK1, we performed co-immunoprecipitation assays. SKW3 human T lymphoma cells (53) were seeded into microtiter wells with or without ICAM-1-coating in the presence of SDF-1␣. After incubation for 30 min under culture conditions, cells were collected and lysed, and immunoprecipitation of kindlin-3 was performed using a kindlin-3-specific mAb (clone 9) that we generated. Immunoblotting of precipitated kindlin-3 was performed using another kindlin-3-specific mAb (clone 229A). High levels of RACK1 co-precipitated with kindlin-3 from lysates of SDF-1␣-treated cells that were allowed to adhere to immobilized ICAM-1 compared with cells treated with SDF-1␣ alone (Fig. 4). Integrin ␣L␤2 also co-precipitated with kindlin-3 in both samples.
To further investigate the physiological relevance of kindlin-3 and RACK1 interaction, we performed immunofluorescence microscopy of T cells. First we examined the localization of these molecules in SDF-1␣-treated SKW3 cells that adhered and spread on immobilized ICAM-1. Actin and nuclei staining were also performed. MyH9 staining was also included because it has been reported to localize at the rear of migrating T cells (54). Kindlin-3 and RACK1 localized at the migrating front, whereas MyH9 localized predominantly at the rear of the cells (Fig. 5). Representative intensity plots showing the distributions of these molecules in the polarized cells are shown (Fig. 5,  insets). In double-staining experiments, kindlin-3 and RACK1 were observed to co-localize at the front of the polarized cells (Fig. 6). Similar observations were made in human T lymphoblasts that adhered and spread on immobilized ICAM-1 in the presence of SDF-1␣ (Fig. 7). To rule out interdonor variation, we also performed additional double-staining experiments with T lymphoblasts from two other donors (supplemental Fig.  S3 and Fig. S4), and the results were consistent.

DISCUSSION
Kindlins are widely reported as co-activators with talin of integrins, but there is gaining evidence that suggests kindlins as mediators of integrin outside-in signaling. In human keratinocytes, kindlin-1 has been reported to form a complex with the ␤1 integrin, ␣-actinin, migfilin, and FAK, and it controls lamellipodia formation via RhoGTPase signaling pathways (55). In mouse osteoblastic cells, kindlin-2 knockdown abrogated ␤1 integrin-mediated activation of Rac1, Akt, and AP-1, which are involved in cell adhesion and proliferation (56). Platelets from mice deficient in kindlin-3 adhered to fibrinogen in the presence of Mn 2ϩ that activates integrin ␣⌱⌱b␤3, but they failed to form lamellipodia (15). Here we show that kindlin-3 is required for outside-in signaling of integrins ␣L␤2 and ␣IIb␤3. The requirement for kindlin-3 in inside-out activation of these integrins was bypassed using mutants ␣L␤2N329S and ␣IIb␤3N339S, which are constitutively activated with high ligand binding affinity (45). Electric cell-substrate impedancesensing measurements showed that cells with reduced kindlin-3 expression were defective in spreading despite expressing these activated mutants. Whether kindlin-3, like its paralogs kindlin-1 and kindlin-2, is involved in mediating integrin ␣L␤2or ␣IIb␤3-derived RhoGTPase signaling remains to be determined.
A recent study of integrin ␣L␤2-mediated T lymphocyte adhesion on ICAM-1 has shown that kindlin-3 is required for T cell receptor-induced clustering of activated ␣L␤2 in early focal assemblies rather than inducing integrin ␣L␤2 conformational changes (44). Our data suggest that kindlin-3 may induce micro-clustering of integrin ␣L␤2. The overall strength of cell adhesion or avidity is dependent on the affinity of individual adhesion molecule to its ligand and the number of adhesion molecules at the contact site (57). Clustering of integrins, therefore, plays an important role in adhesion strengthening. At present, we do not have information on the mechanism by which kindlin-3 induces integrin ␣L␤2 micro-clustering. Its PH domain may have an important role in this process by binding to membrane phosphatidylinositol phosphate(s). It is interesting to note that talin, which is a well established direct activator of integrins, can induce ␤3 integrin clustering, and this is dependent on its F2 subdomain binding to phosphatidylinositol 4,5-diphosphate (58, 59). SKW3 human T lymphoma cells were seeded into empty or ICAM-1-coated microtiter wells in the presence of SDF-1␣. After incubation for 30 min, cells were harvested, and immunoprecipitation (IP) was performed to assess the interaction between kindlin-3 and RACK1. Endogenous kindlin-3 was immunoprecipitated with anti-kindlin-3 mAb (clone 9). An immunoblot (IB) of precipitated kindlin-3 was performed using anti-kindlin-3 (clone 229A). Immunoblots of co-precipitated endogenous RACK1 and integrin ␣L␤2 were performed using anti-RACK1 mAb (clone B-3) and anti-␣L mAb (clone 27), respectively. IgG, control rat IgG.
Aforementioned, we and others have shown that kindlins are involved in integrin outside-in signaling. We have provided evidence that kindlin-3 interacts with RACK1, a seven-blade propeller WD-repeat scaffold protein (32). Two other RACKs are the PKC⑀ and PKC␣ receptors ␤Ј-COP and Pick1, respectively (60,61). RACK1 binds many cytoplasmic molecules (32), and the interactions involved different blades of RACK1. For example, c-Src (blade 6), RAID (blades 5, 6, and 7), PKC␤II (blades 3 and 6), and integrin ␤1, ␤2, and ␤5 tails (blades 5, 6, and 7) (32, 33). RACK1 stabilizes the activated forms of PKCs and enables the translocation of PKCs to different sites in the cell (62), but it inhibits the activity of Src family kinases (63). However, it has been reported that RACK1 does not interact with PKC␤ or Src in chemokine-treated Jurkat T cells (47). There are also conflicting reports as to whether RACK1 inhibits or promotes cell migration. RACK1 overexpression in integrin ␣IIb␤3-expressing CHO cells increased the number of actin stress fibers and focal contacts, and the cells showed impeded migration on

Kindlin-3 Mediates Integrin ␣L␤2 Outside-in Signaling
fibrinogen (34). RACK1 has been shown to localize to nascent focal complexes and regulates cell protrusion (64). It has also been reported that RACK1 localized to the leading edge of chemoattractant-induced polarized Jurkat and HL-60 cells, but it associates with G␤␥ and act as a negative regulator of directed cell migration (47). Thus, the functions of RACK1 may be cell type-specific and dependent on the type of cellular stimuli. Adding to the complexity is that RACK1 has been reported as a eukaryotic ribosomal protein (65).
The functional consequences of kindlin-3-RACK1 interaction remain to be delineated. Our data suggest that the interaction is dependent on kindlin-3 PH domain and RACK1 blades 5-7. We cannot exclude other blades of RACK1 that may be important for this interaction because we were unable to obtain soluble expression of other RACK1 blades using the E. coli. expression system. Because blades 5-7 of RACK1 contain binding sites for other proteins, it remains to be determined if RACK1 can interact with these molecules simultaneously or with selective molecule(s) at any one time. RACK1 has been shown to bind the membrane proximal sequence Lys-702-Leu-721 of the integrin ␤2 tail (33), whereas kindlin-3 binds to the C-terminal sequence of the integrin ␤2 tail that contains the 741 NPKF motif (14). Our data suggest that the integrin ␤2 tail, kindlin-3, and RACK1 can form a ternary complex. It is possible that RACK1 recruits kinases such as c-Src or activated PKCs to this complex, which may modulate both ligand binding and signaling activity of the ␤2 integrins.
The kindlin-3 and talin binding sites in the integrin ␤2 tail are in close proximity of each other. Whether kindlin-3 and talin binding to the integrin ␤2 tail influence the interaction of kindlin-3 with RACK1 remains to be determined. Currently it has yet to be shown that kindlin-3 and talin head domain can bind a single integrin ␤2 tail simultaneously. There is little evidence suggesting direct interaction between kindlins and talin head domain. The fate of these two molecules after integrin ligand binding is also unclear. Several models of integrin activation by kindlins and talin have been proposed (25). The cis cooperation model positions kindlin and talin on a single integrin tail, but the sequential binding and trans co-operation models do not. Our findings using the SKW3 T cells suggest that integrin ␣L␤2 engagement by ICAM-1 promotes kindlin-3 and RACK1 interaction. This may be important in integrin ␣L␤2 outside-in signaling. More studies are needed to delineate the function of kindlin-3-RACK1 interaction in this context. Unlike kindlin-3, which interacts with RACK1, we were unable to detect significant interaction between kindlin-2 and RACK1. RACK1 has been reported to be important for nascent focal adhesion complex formation (64,66), and its localization was detected in preforming adhesion sites known as spreading initiation centers but not in mature focal adhesion sites (67). Kindlin-3 has been shown to localize at the spreading lamellipodia, but not focal adhesion sites, of human umbilical vein endothelial cells cultured on integrin substrates, whereas kindlin-2 was detected in both cellular structures (26). Indeed, we observed that kindlin-3 and RACK1 localized at the migrating front of polarized T cells. By contrast, MyH9 was localized at the rear of these cells, which is in line with previous observations (54). We did not perform kindlin-2 staining of these cells because we could not detect kindlin-2 protein in these cells by immunoblotting (data not shown), although kindlin-2 mRNA expression was detected at a low level in lymphoid organs (68). The physiological significance of kindlin-3, but not kindlin-2, interacting with RACK1 remains to be determined. It may be interesting to examine this in human umbilical vein endothelial cells that express both kindlins (26).
In summary, we have shown that kindlin-3 plays a role in integrin ␣L␤2 outside-in signaling, and it can induce integrin ␣L␤2 micro-clustering. We also report that kindlin-3 interacts with the scaffold protein RACK1 that is known to have multiple cytoplasmic binding partners. The biological importance of this interaction in integrin outside-in signaling awaits further studies.