Platelets lacking PIP5KIγ have normal integrin activation but impaired cytoskeletal-membrane integrity and adhesion

Critical experiments by Lowell and Mabel Hokin >50 years ago demonstrated that the inositol head group of phosphatidylinositol can be transiently phosphorylated at the hydroxyl groups of the 3, 4, or 5 position to generate 7 distinct members of the phosphoinositide family.¹  Although only 1% of cell membrane phospholipids are phosphoinositides, they play a key role within all eukaryotic cells because of their unique ability to be phosphorylated.^2,3  Much interest has been focused on a specific phosphoinositide, phosphatidylinositol-4,5-bisphosphate (PIP₂). This phosphoinositide is predominantly synthesized by phosphatidylinositol-4-phosphate 5-kinase (PIP5KI)-mediated phosphorylation of phosphoinositide 4-phosphate at the D5 position on its inositol ring. PIP₂ is widely known as being a substrate for the production of second messengers because of its hydrolysis by phospholipase C and its phosphorylation by phosphatidylinositol 3-kinase. PIP₂ also directly binds to proteins, which alters the function of these proteins and ultimately helps to regulate GTP-binding proteins, actin-binding proteins, phospholipases, and vesicle secretion.

All mammals have three genes that encode the 3 isoforms of PIP5KI known as PIP5KIα, PIP5KIβ, and PIP5KIγ.^4-6  All 3 isoforms can be activated by small GTPases (ρ, Rac, Cdc42, and ARF), as well as by phosphatidic acid.^7,8  Although PIP5KIα, PIP5KIβ, and PIP5KIγ are all capable of synthesizing PIP₂, these isoenzymes have significantly dissimilar primary structures, different expression levels in different tissues, and different locations within the subcellular compartments.^9-15  Primary sequence alignments and studies of genetically engineered mice suggest that PIP5KIα and PIP5KIβ have similar structures and functions.^5,16-18  However, PIP5KIγ remains distinct from these 2 other isoforms. First, outside of its catalytic kinase core, PIP5KIγ is much larger than the other isoforms and shares very little sequence homology with them. Second, PIP5KIγ is the only isoform that contains alternative splice variants. These splice variants have different subcellular distributions, which suggests their different functions. Finally, expression studies suggest that PIP5KIγ is the isoform that contributes to focal adhesion formation.^12-14 

The critical region of PIP5KIγ for association with talin is absent in a naturally occurring p87 splice variant of this enzyme. In contrast to the p87 shorter splice variant of PIP5KIγ, the p90 splice variant can coimmunoprecipitate and colocalize with talin. This talin association is attributed to 26 amino acids (encoded in exon 17) at the carboxy-terminus of the longer p90 splice variant.^12,13  These studies have suggested that the interaction between PIP5KIγ p90 and talin is an important modifier of talin-mediated integrin activation.¹⁹  Furthermore, additional fibroblast expression studies suggest that focal adhesions can only form in the presence of the p90 splice variant of PIP5KIγ.

A model proposes that only talin-bound p90 PIP5KIγ locally generates PIP₂, which subsequently binds to talin and enables it to regulate integrin activation. Although this model is widely quoted, the selective deletion of the longer p90 splice form of PIP5KIγ produces only subtle integrin defects in both T cells and fibroblasts.^20,21  This argues that the p87 splice form might be able to compensate for the loss of p90 and support integrin dynamics. We addressed this question by genetically modifying mice so that they lacked either the p90 PIP5KIγ isoform alone or both of the splice forms of PIP5KIγ in platelets. Our studies demonstrate that, although PIP5KIγ is essential for normal platelet function, individual isoforms of PIP5KIγ fulfill unique roles for the integrin-dependent integrity of the membrane cytoskeleton and for the stabilization of platelet adhesion.

An 8.45-kb region used to construct the targeting vector was first subcloned from a positively identified C57BL/6 (RPCI23: 446D15) BAC clone (genetargeting.com). The region was designed such that the short homology arm extends about 2.03 kb 5′ to exon 17. The long homology arm ends 3′ to exon 17 and is 5.70 kb long. The loxP/FRT flanked Neo cassette was inserted on the 5′ side of exon 17, and the single loxP site is inserted at the 3′ side of exon 17. The target region is 721 bp and includes exon 17. The homology recombination in embryonic stem cells was confirmed by Southern blotting combined with sequence analysis. The mice with conditional allele were genotyped with a primer pair of SDL2: 5′-TGTACTCCCGCTTCACTATAGCG-3′ and LOX1: 5′-CAGGTGATGTCGCTGAGCTC-3′. They were then crossed with PF4Cre transgenic mice to induce the deletion of exon 17 specifically in platelets and in megakaryocytes.

To confirm the deletion of exon 17 in murine platelet mRNA, a primer pair flanking the exon 17 was used for reverse transcription-polymerase chain reaction (RT-PCR) analysis. The sequence of the PCR primers were 5′-CAGGTGGAGCCAGTGTGCGG-3′ and 5′-AGGAAGTGGCTGGGGTGGCA-3′. The PCR reaction amplified a 319-bp band for the p90 PIP5K1γ and a 241-bp band for the p87 PIP5K1γ. Total platelet lysates were used for immunoblotting to confirm the loss of the 26 amino acids encoded by exon 17 by using a PIP5K1C polyclonal antibody (catalog no. 3296; Cell Signaling Technology, Danvers, MA).

The generation of gene trap PIP5KIγ heterozygous mice has been previously described.²²  These mice were crossed with MLC-2v Cre transgenic mice (a generous gift of Kenneth Chien, University of California, San Diego, CA) to generate PIP5KIγ^+/− MLC-2v Cre⁺ mice. Pairing of these mice produced some PIP5KIγ^−/− MLC-2v Cre⁺ offpsring. The Institutional Animal Care and Use Committee approved this study.

The yolk sacs of E10.5 embryos were dissected and disaggregated with collagenase, and the hematopoietic progenitor cells were differentiated with thrombopoietin-enriched conditioned media over 5 days into pro-megakaryocytes and megakaryocytes as previously described.²²  Confocal microscopy and staining of megakaryocytes expressing green fluorescent protein (GFP)-tagged PIP5KIγ were imaged at 37°C under a 60×/1.4 NA oil immersion lens (Nikon, Melville, NY) using an Ultraview-LCI spinning disk confocal microscope (PerkinElmer, Shelton, CT) attached to a Nikon TE-300 (Nikon, Melville, NY) inverted light microscope as described.²² 

Immunogold labeling of talin and GFP antibodies was done as described previously.²³  Megakaryocytes were attached to fibrinogen-coated coverslips by centrifugation at 280g for 5 minutes followed by incubation at 37°C for 1 hour. Adherent megakaryocytes were permeabilized using 0.75% Triton X-100 in PHEM buffer (60 mM Pipes, Hepes [N-2-hydroxyethylpiper-azine-N'-2-ethanesulfonic acid], EGTA, and magnesium [PHEM], 25 mM Hepes, 2 mM MgCl₂, 10 mM EGTA, and 1 µM phallacidin) containing 0.05% glutaraldehyde for 2 minutes. Cytoskeletons were washed with PHEM buffer and fixed with 1% formaldehyde in PHEM buffer for 10 minutes at 37°C, and unreacted aldehydes were reduced with a wash of 1 mg/mL sodium borohydride in phosphate-buffered saline (PBS) for 1 minute. The cytoskeletons were washed in PBS containing 1% bovine serum albumin (BSA) and labeled with rabbit anti-GFP and mouse anti-talin antibodies for 3 hours. They were then washed three times in 1% BSA in PBS and incubated with 1:20 dilutions of 10 nm goat anti-rabbit immunogold (Ted Pella, Inc, Redding, CA) and 5 nm goat anti-mouse IgG overnight. The samples were then washed three times with 1% BSA in PBS, washed three times with PBS, fixed with 1% glutaraldehyde in PBS for 10 minutes, and extensively washed in distilled water. The samples were rapidly frozen, freeze-dried at −90°C, and metal cast with 1.4 nm of tungsten-tantalum at 45°C with rotation and 5 nm of carbon at 90°C without rotation (Cressington CFE-60 Freeze Fracture Machine; Cressington, Watford, England). The metal casts were separated from the coverslips using 25% hydrofluoric acid, washed with distilled water, picked up using 200 mesh formvar-coated copper grids, and photographed at 100 kV in a JEOL 1200-EX electron microscope.

Acid citrate dextrose-anticoagulated platelet-rich plasma was isolated, and washed platelet aggregation was measured by the turbidimetric method at 37°C in a Lumi-dual aggregometer (CHRONO-LOG Co., Havertown, PA) as previously described.¹⁶  Jon/A binding was done on prewarmed washed platelets that were diluted to a density of 5 × 10⁶/mL and then stimulated at 37°C with thrombin in the presence of phycoerythrin-conjugated Jon/A. For the actin assembly experiment, platelets were analyzed on a FACSCalibur (BD PharMingen) using FlowJo software (Tree Star, Ashland, OR) as described previously.²⁴  Analysis of focal adhesion kinase (FAK) was performed by layering 4 × 10⁸ platelets over 0.1% fibrinogen-coated 10-cm tissue culture dishes in the presence of 0.5 U/mL thrombin. After incubation at 37°C for 15 minutes, the cells were lysed in RIPA buffer and immunoblotted with either anti-pFAK (Cell Signaling Technology) or anti-FAK (Millipore Corp., Billerica, MA) antibodies. Carotid artery injury induced by FeCl₃ was previously described.¹⁶ 

To enable murine platelet binding to human vWF-A1, histidine at position 1326 in human vWF-A1 was mutated to arginine (H1326R).²⁵  The cDNA encoding human vWF-A1-6His tag in the pQE9 vector (a generous gift from Dr Miguel Cruz, Baylor College of Medicine)²⁶  was used as a template to alter H1326 to R using the Quick Change mutagenesis kit (Stratagene, La Jolla, CA). The mutation was confirmed by DNA sequencing.

Histidine-tagged wild-type or H1326R vWF-A1 was produced and purified according to the manufacturer’s instructions. Briefly, Escherichia coli M15 (Qiagen, Germantown, MD) were transformed with pQE9 vectors containing vWF-A1 and H1326R mutants. Protein expression was induced with 1 mM isopropyl β-D-1 thiogalactopyranoside. Bacterial cells were harvested and resuspended in 5 mL of lysis buffer (50 mM Na₂H₂PO₄, 300 mM NaCl, and 10 mM imidazole) containing protease inhibitor (Thermo Scientific) and 1 mg/mL of lysozyme. The bacterial suspension was sonicated (6 cycles of 10-second bursts followed by 10 seconds of cooling), and proteins were purified from bacterial lysates using Ni-affinity chromatography.

Platelets were isolated from murine peripheral blood as described previously.²⁷  Murine platelet rolling on human H1326R vWF-A1 under shear flow was measured as was described for human platelets rolling on vWF-A1.²⁸  Membrane tethers of platelets were visualized in real-time images by differential interference contrast videomicroscopy, using a 100× oil immersion objective lens under the indicated shear stress.²⁹  The images were recorded on a SVHS recorder, transferred onto a digital video recorder (Sony, Tokyo, Japan), and digitalized at 30 frames/s.

Platelet adhesion to collagen type I or to fibrinogen (Sigma-Aldrich, St. Louis, MO) was performed as described previously.³⁰  Culture dishes (35 mm) were coated with 500 μg/mL collagen type I or 200 μg/mL fibrinogen (Sigma-Aldrich) at 4°C overnight and then blocked by 2% human serum albumin at 4°C for 2 hours. Platelets (1 × 10⁸/mL) in Hepes-Tyrode’s buffer with 2 mM MgCl₂ were perfused and allowed to accumulate on collagen- or fibrinogen-coated dishes at a wall shear stress of 1.0 dyn/cm² for 10 minutes. After changing to platelet-free buffer, wall shear stress was increased every 20 seconds up to 40 dyn/cm², and the platelet covered area was measured using Element software (Nikon, Tokyo, Japan).

A custom-built optical trap³¹  was used to form membrane tethers by pulling on a fibrinogen-coated bead that was touched repeatedly to the surface of a cell. Mouse fibrinogen (Sigma-Aldrich) was covalently bound to 1.87-μm carboxylate-modified latex beads (Bangs Laboratories, Fishers, IN). The mixture of the cells and the beads was inserted into a microscope flow chamber in the culture medium containing 1 mM MnCl₂ to activate the integrins, and the cells were allowed to settle on the polylysine-coated bottom of the chamber for 5 to 10 minutes. Near a cell that was firmly attached to the bottom of the chamber, a bead was trapped, and the microscope focus was adjusted so that the bead and cell centers were approximately the same distance from the bottom surface. The position of the laser trap was then oscillated in a triangular waveform with a frequency of 1 Hz and a constant peak-to-peak amplitude of 2 μm (0.2 pN/nm trap stiffness; loading rate 800 pN/s), allowing the bead to touch the cell repeatedly until they stuck to each other. When a cell-attached fibrinogen-coated bead was pulled apart by the optical trap, it either remained unmovable on the surface (firm attachment, no tether formation) or moved away from the cell while being connected by a thin elastic, highly extensible membrane tether, as shown in Figures 3A and 5A. The frequency of tether formation was calculated by pooling together data from all experiments performed with cells having the same genotype. The results are presented as an average ± standard deviation from the n series.

A modification of the FeCl₃-induced thrombosis assay originally described by Kurz et al³²  was performed. Briefly, mice were anesthetized using sodium pentobarbital (90 mg/kg, intraperitoneal), a midline incision was made in the neck, and the carotid artery was exposed by blunt dissection. Blood flow in the artery was measured using a Doppler flow probe (Transonic Systems, Inc, Ithica, NY). Following a baseline flow measurement, a 1 × 1-mm piece of filter paper soaked with 10% FeCl₃ was placed on the outside of the blood vessel for 2 minutes. The filter paper was then removed, and blood flow was monitored for an additional 20 minutes. The time to complete vessel occlusion was recorded.

To elucidate the functions of PIP5KIγ, we previously described an embryonic stem cell line that contained a β-geo gene trap within the first intron of the PIP5KIγ gene.^22,33  The strategy was designed to create an abnormal mRNA transcript from the trapped allele that would produce a fusion protein corresponding to the first 32 amino acids of PIP5KIγ fused to β-galactosidase. However, this targeting strategy in the R129/C57Bl6 mixed genetic background led to embryonic lethality at day 11.5 of development, which is attributable to a failure of myocardial cells to form tight junctions.²² 

This early prenatal lethality of the PIP5KIγ^−/− embryos precluded studies of hematopoietic cells derived from the bone marrow or liver.²²  Therefore, to analyze the loss of PIP5KIγ on the cytoskeleton of platelets and in megakaryocytes, we analyzed cultured yolk sac progenitor cells that were treated with thrombopoietin ex vivo and that were then differentiated into megakaryocytes. After 5 days in culture, ∼70% of the cells had a multilobulated nucleus and expressed CD41 (integrin αIIb), a marker for the megakaryocyte lineage. This demonstrated that the majority of the viable cultured cells had differentiated into platelet precursor cells and that PIP5KIγ is not required for this differentiation ex vivo. Electron microscopy showed that PIP5Kγ-null megakaryocytes had a normal architectural appearance of their actin network and formed normal appearing proplatelets and platelets (data not shown).

To determine whether p90 PIP5KIγ codistributes with talin in primary hematopoietic cells such as in megakaryocytes, we infected cells with retrovirus expressing p90 PIP5KIγ fused to GFP. Confocal microscopy demonstrated that the GFP-PIP5KIγ fusion protein was predominantly concentrated on the cellular membranes (Figure 1A, top). The distribution of talin was quite similar (Figure 1A, middle). Merging these distributions confirmed that p90 PIP5KIγ and talin colocalize within the same regions of megakaryocytes (Figure 1A, bottom). However, similar results were obtained with p87 PIP5KIγ. Therefore, the resolution of light microscopy limits our ability to determine whether there is a direct interaction between p90 PIP5KIγ and talin.

The localization of PIP5KIγ relative to talin in the cortex of spread megakaryocytes expressing GFP-PIP5KIγ splice variants was further studied using electron microscopy. The cytoskeleton of the megakaryocytes was stained with rabbit anti-GFP IgG and mouse anti-talin IgG coupled with 5- and 10-nm gold particles, respectively. The gold-coupled immunoglobulin against PIP5KIγ was found in clusters composed of 2 to 5 gold particles (Figure 1B, bracketed with red squares) that were situated at points of the cytoskeletal filament-substrate attachment. The gold-coupled immunoglobulins against talin localized at similar positions in the cytoskeleton with the location of the antibody against PIP5KIγ (Figure 1B, blue circles). Comparison of the location of the 2 different gold sizes demonstrated ∼50% (28/54) colocalization of talin with p90 PIP5KIγ within 50 nm, a distance that takes into account the width of the gold particles and the shell size of the 2 antibodies (Figure 1C). In contrast, the p87 PIP5KIγ rarely colocalized with talin. This result, along with previously published PIP5KIγ-talin coimmunoprecipitation and crystallization studies,^12,13,34  suggests that within megakaryocytes, there is a direct interaction between talin and the p90 PIP5KIγ splice variant but not a direct interaction with the p87 splice variant.

To analyze whether the p90 (talin-binding) splice variant of PIP5KIγ serves specific functions within platelets, we genetically engineered mice to lack only the p90 PIP5KIγ variant. This was achieved by introducing loxP sites into regions of the murine PIP5KIγ gene that flank exon 17, which is the exon that encodes the 26 amino acids critical for talin binding (Figure 2A). Heterozygous embryonic stem cells were confirmed by Southern blotting shown in the bottom panel of Figure 2A and were used to generate chimeric and ultimately heterozygous targeted mice. Mice containing this germ-line mutation were crossed with PF4 Cre or CMV Cre transgenic mice. This breeding strategy ultimately yielded PIP5KIγ E17^−/− PF4 Cre⁺ mice that specifically lacked the p90 PIP5KIγ splice variant exclusively in their platelets, or it yielded PIP5KIγ E17^−/− CMV Cre⁺ mice that lacked the p90 PIP5KIγ splice variant in all tissues including in platelets (Figure 2B).

These mice were morphologically normal, had normal platelet counts, and exhibited no signs of spontaneous hemorrhage. We analyzed the integrity of the integrin-cytoskeleton connection in PIP5KIγ E17^−/− PF4 Cre⁺ platelets. Our previous studies indicated that megakaryocytes derived from PIP5KIγ-null yolk sac progenitor cells have a defect in anchoring their integrins to the underlying cytoskeleton.³⁵  We also observed that this defect was dependent on the lipid kinase activity of PIP5KIγ. Although our prior work suggested that PIP5KIγ was important for anchoring integrins, this study did not analyze membrane dynamics in a cell that required integrins for their principal function. In addition, it has been postulated that platelets regulate the elasticity of their cell membranes during the initial phases of platelet adhesion.³⁵  Therefore, we reasoned that PIP5KIγ might contribute to the regulation of integrins and cell elasticity in platelets. We observed by using an optical trap that PIP5KIγ E17^−/− PF4 Cre⁺ platelets had a defect in anchoring their integrins to the underlying actin cytoskeleton (Figure 3B). This is consistent with our previous observations that the stable attachment of the integrins to the cytoskeleton is dependent on both talin and the p90 PIP5KIγ splice variant.³⁶ 

We further analyzed whether platelets that expressed p87 PIP5KIγ, but failed to express p90 PIP5KIγ, had an ex vivo adhesion defect. Platelets derived from PIP5KIγ E17^−/− CMV Cre⁺ mice were flowed over collagen or fibrinogen. Platelets lacking only p90 PIP5KIγ adhered normally to collagen at low wall shear stresses and resisted detachment at wall shear stresses of up to 40 dyn/cm² (Figure 3C). Both wild-type and knockout platelets adhered less well to fibrinogen. Platelets lacking p90 PIP5KIγ also adhered normally to fibrinogen at low wall shear stresses, although they detached modestly more frequently at high wall shear stresses (Figure 3D).

The results of platelets lacking only the p90 PIP5KIγ splice variant demonstrate that it is dispensible for platelet adhesion. Therefore, we questioned whether the p87 PIP5KIγ isoform might fulfill this role. Our prior studies indicated that the PIP5KIγ knockout embryos have defective myocardial development but normal development of the foregut and endocardium.²²  This cardiac defect was caused by the lack of PIP5KIγ myocardial expression. Therefore, by preserving expression in myocardiocytes, it was predicted that we could correct this phenotype and allow the embryos to progress beyond the 12th day of development.

The PIP5KIγ gene trap targeting construct used for this targeting strategy has lox recombination sites flanking its splice acceptor (Figure 2C). Consequently, the PIP5KIγ mRNA trap will be spared in tissues that express Cre recombinase (this is opposite of the typical use of the Cre-lox system that induces the null mutation only in the Cre expressing tissues.) Chen et al³⁷  have shown that MLC-2v Cre⁺ transgenic mice exclusively express Cre recombinase in the myocardium. We bred PIP5KIγ^+/− MLC-2v Cre⁺ mice, and we predicted that the pairing of these mice would generate some embryos that have the null mutation for PIP5KIγ in all tissues except for the myocardium. Given the pleomorphic developmental abnormalities in the PIP5KIγ-null embryos, our expectation was that embryos lacking PIP5KIγ in all tissues except for the myocardium would still die prenatally because of widespread and profound developmental defects.

Surprisingly, of the 42 mice produced by the pairing of PIP5KIγ^+/− MLC-2v Cre⁺ mice, we found 6 living PIP5KIγ-null mice. RT-PCR and anti-PIP5KIγ immunoblotting confirmed that Cre recombinase under the control of the MLC-2v promotor successfully reverted PIP5KIγ expression only in the heart. In all other analyzed tissue samples including platelets, there was a complete lack of PIP5KIγ expression (ie, the mice lacked PIP5KIγ in all cells except for the myocardiocytes; Figure 2D). Quantitative RT-PCR of platelet mRNA confirmed the absence of the PIP5KIγ transcript and also demonstrated that there was no compensatory change in the expression levels of the transcripts that encode for PIP5KIα or PIP5KIβ mRNA (data not shown). Presumably, reverting the PIP5KIγ knockout in the heart allows the embryo to survive long enough for compensatory pathways to arise that permit development of the brain and other vital organs.

We used several assays to analyze integrin activation and actin dynamics in PIP5KIγ-null platelets. First, we investigated the impact of several different platelet agonists on integrin activation by analyzing platelet aggregation. Agonist stimulation initiates an “inside-out” signaling cascade that induces a talin-dependent conformational change in the platelet integrin αIIbβ3. This conformational change allows the integrin to bind to its ligand fibrinogen and thereby supports the aggregation of platelets. Platelets lacking PIP5KIγ aggregated normally at low or high concentrations of 4 different analyzed agonists (collagen, PAR4 agonist peptide AYP, U46619, and thrombin; Figure 4A). Second, using the Jon/A antibody that only binds to the activated αIIbβ3 integrin, we found that the activation of αIIbβ3 was completely normal in platelets lacking PIP5KIγ (Figure 4B). Third, we determined that agonist-stimulated actin assembly was normal in PIP5KIγ-null platelets (Figure 4C). Fourth, we analyzed FAK phosphorylation that is induced when αIIbβ3 binds to its cognate ligand (“outside-in” signaling). As shown in Figure 4D, no significant changes in FAK phosphorylation were found in platelets lacking PIP5KIγ . Taken together, these studies demonstrate that inside-out signaling and integrin-induced FAK phosphorylation are normal in platelets lacking PIP5KIγ.

Finally, we used an optical trap to pull the integrin αIIbβ3 on platelets lacking PIP5KIγ to determine whether they had a failure to anchor their cell membranes to the actin cytoskeleton. This was done by pulling on a fibrinogen-coated bead firmly attached to a platelet membrane after Mn²⁺-induced integrin activation (Figure 5A). As shown in Figure 5B, we found that a portion of PIP5KIγ knockout platelets had cell membranes that could easily be pulled apart from the cytoskeleton, whereas this phenomenon was uncommonly seen in wild-type platelets. Similar to the PIP5KIγ-null cells, the wild-type platelets treated with latrunculin A displayed a marked ability to form membrane tethers. Together, these results demonstrate that platelets lacking PIP5KIγ have a membrane-related cytoskeletal defect that might account for the ex vivo adhesion defect.

To further analyze the contribution of PIP5KIγ to integrin dynamics, we studied whether the loss of this PIP5KI isoform affected stable adhesion of platelets exposed to shear pressure. To determine whether PIP5KIγ was essential for integrin-mediated adhesion, we examined adhesion of platelets to collagen, which is mediated by integrin α2β1. At a wall shear stress of 1.0 dyn/cm², a similar area of collagen-coated dishes were covered by wild-type and PIP5KIγ-null platelets. After increasing wall shear stress up to 40 dyn/cm², significant platelet detachment and reduction of the platelet covered area were observed in PIP5KIγ-null platelets but not in wild-type platelets (Figure 5C-D). The equivalent adhesion of PIP5KIγ-null and wild-type platelets at low shear stress implies that integrin α2β1 in PIP5KIγ-null platelets binds normally to collagen. However, stability of integrin binding in PIP5KIγ-null platelets is impaired in conditions of high shear. Furthermore, wild-type platelets adhered to collagen in small clusters, whereas adherent platelets lacking PIP5KIγ failed to aggregate with other platelets (Figure 5C).

These results could indicate that PIP5KIγ is required for all types of platelet adhesion or only for efficient integrin-mediated cell attachment. Platelet interactions with von Willebrand factor are mediated by the GPIb-V-IX complex, an adhesion receptor that is not a member of the integrin family. Therefore, we analyzed adhesion of wild-type and PIP5KIγ-null platelets to immobilized von Willebrand factor. On this substrate, wild-type and PIP5KIγ-null platelets exhibited similar flow-enhanced rolling velocities (Figure 5C) and extruded membrane tethers at the trailing edges in similar numbers and with similar lengths (data not shown). These results suggest that the adhesion defect observed in PIP5KIγ-null platelets is restricted to integrins.

Finally, we also analyzed the PIP5KIγ E17^−/− PF4 Cre⁺ and PIP5KIγ^−/− MLC-2v Cre⁺ mice for their ability to form thrombi in response to a chemical injury applied to their carotid arteries.^32,38  Ferric chloride was applied to an exposed carotid artery of anesthetized mice, and the formation of thrombi was monitored using a Doppler flowmeter. Similar to the results of the ex vivo adhesion experiments (Figures 3 and 5), we observed that PIP5KIγ E17^−/− PF4 Cre⁺ mice had normal platelet adhesion in vivo, but PIP5KIγ^−/− MLC-2v Cre⁺ mice formed stable thrombi less efficiently (Figure 6).

Under the experimental conditions, wild-type mice always formed an occlusive thrombus in response to the chemical irritant (such as shown in supplemental Figure 1 on the Blood website). In contrast, the PIP5KIγ-null mice usually formed thrombi as rapidly as wild-type mice, but these vascular occlusions were unstable (Figure 6A-C). Together, these results suggests that PIP5KIγ plays a critical role in stable platelet adhesion ex vivo and in vivo.

Reports by several groups have shown that PIP5KIγ can associate with talin under certain circumstances.^12,13,39-42  The atomic structure of a fragment of mouse talin-1 (amino acids 306-429) bound to a peptide corresponding to amino acids 641-648 of PIP5KIγ has also been published.⁴³  Our data demonstrate that p90 PIP5KIγ can localize with talin in a primary cell, but surprisingly is not required for integrin-mediated adhesion. Instead, it affects the linkage between membrane bound integrins with the underlying cytoskeleton in platelets. Curiously, we also observed that mice lacking both p87 and p90 PIP5KIγ have impaired platelet adhesion ex vivo and impaired thrombus formation in vivo.

Our data are in agreement with a model first proposed by Di Paulo et al and Ling et al that p90 PIP5KIγ associates with talin.^12,13  We demonstrated p90 PIP5KIγ-talin colocalization by immuno-electron microscopy and by genetic evidence that cells lacking PIP5KIγ or talin³⁶  share a similar membrane phenotype. Our finding is not consistent with publications that suggest that p90 PIP5KIγ is essential for integrin activation.^12,13  Instead, we determined that all aspects of integrin signaling analyzed were completely normal in platelets lacking this isoform of PIP5KIγ. The apparent discrepancies in our findings might be attributable to our studies of a primary cell as opposed to cell lines or might be unique to the biology of platelets. Our data are consistent with work that demonstrates that PIP5KIγ-synthesized PIP₂ in fibroblasts is necessary for the coupling of integrins with the actin cytoskeleton.^20,21  However, our work demonstrates that in platelets, neither of the PIP5KIγ isoforms is required for the binding of integrins to their ligands. Instead, PIP5KIγ likely functions to provide integrin-dependent signaling events once adhesion occurs.

A surprising finding in this study is the observation that the loss of both p87 and p90 PIP5KIγ impairs stable integrin-mediated platelet adhesion, yet the loss of p90 PIP5KIγ alone does not have this effect. This indicates that p87 PIP5KIγ is actually the isoform that is critical for stable adhesion or that platelets need only 1 of the 2 PIP5KIγ isoforms to support adhesion. Intriguingly, stable integrin-mediated adhesion does not correlate with the ability of integrins to anchor to the underlying cytoskeleton (compare Figure 3B with Figures 3C and 6D). Microscopy of adherent platelets indicates that platelets lacking both p87 and p90 PIP5KIγ fail to form small clusters in the flow chamber (Figure 5C). This suggests that PIP5KIγ is mediating a signaling event that promotes shear-resistant aggregates of platelets. The nature of this signaling cascade is not evident at this point, although it clearly does not involve the activation of FAK (Figure 4D).

Stable thrombus formation is impaired in PIP5KIγ^−/− MLC-2v Cre⁺ mice that lack both PIP5KIγ isoforms. Although the thrombosis defect could be attributable to a lack of PIP5KIγ in a cell other than in a platelet, it is notable that their platelets display a shear-resistent integrin-mediated adhesion defect ex vivo.⁴⁴  Therefore, our experiments suggest that the in vivo adhesion defect in the PIP5KIγ^−/− MLC-2v Cre⁺ mice is attributable to their platelets.

Our findings suggest that the p90 PIP5KIγ splice variant contributes to the flexibility of the platelet membrane. These data are consistent with work by Raucher et al^45,46  that demonstrated that marked disruption of PIP₂ by overexpression of phospholipid-sequestering polypeptides or inositol phosphatases or by pharmacologic activation of phospholipase C reduces the adhesion energy between the cell membrane and the cytoskeleton. Our data also support the findings by Sun et al,⁴⁷  which showed that overexpression of PIP5KIγ increased the rigidity of the membrane in T cells. The structural mediators of this PIP₂-driven signal are less clear, but candidate PIP₂-binding proteins are vinculin, talin, and filamin. We and others^36,48,49  have demonstrated that cells lacking any of these 3 proteins have impaired anchoring of their membranes to the underlying cytoskeleton. Because of the close proximity between p90 PIP5KIγ and talin, impaired talin-mediated linking of the cell membrane with the cytoskeleton is the most plausible explanation for the phenotype observed in PIP5KIγ-null platelets. Thus, in contrast to other PIP5KI isoforms, PIP5KIγ may be uniquely poised to locally generate high concentrations of PIP₂ that regulate this specific function of talin. Curiously, failure to anchor platelet integrins to the underlying cytoskeleton does not affect platelet adhesion ex vivo or in vivo.

We have previously published that the loss of PIP5KIγ leads to a perinatal lethality at day 11.5 of development.²²  This was associated with pleomorphic abnormalities that were visible in the heart, brain, and limb buds. Here we demonstrated that reverting this null mutation exclusively in the myocardium allows for the survival in mice otherwise lacking PIP5KIγ in other tissues. This finding implies that the lack of PIP5KIγ delays normal in utero development of multiple organs, but ultimately PIP5KIγ is nonessential for the development of nonmyocardial cells. It is notable that Legate et al²⁰  found that mice lacking the p90 splice variant of PIP5KIγ develop normally as well. Together, these data suggest that p87 PIP5KIγ is essential for the development of the cardiovascular system.

In summary, we found no evidence that the loss of p90 PIP5KIγ impairs inside-out integrin signaling in platelets. We found that platelets lacking p90 PIP5KIγ have impaired linkage between integrins and the underlying cytoskeleton. However, p90 PIP5KIγ is not required to synthesize the pool of PIP₂ necessary for integrin activation in this type of primary cell. Instead, we found that p87 PIP5KIγ is actually the isoform critical for stable adhesion or that either of the 2 isoforms can perform this function. Our work therefore challenges an existing model of integrin activation. These studies demonstrate that in contrast to PIP5KIα and PIP5KIβ, PIP5KIγ uniquely synthesizes a spatially localized pool of PIP₂ that preserves the integrin-dependent events that are important for stable platelet adhesion ex vivo and in vivo.

This work was supported in part by funds from the National Institutes of Health, National Heart, Lung, and Blood Institute (grants HL40397, HL083392, and HL110110).

Contribution: Y.W. designed, performed, and interpreted the experiments and assisted in the preparation of the manuscript; L.Z. designed, performed, and interpreted the experiments; A.S. interpreted the experiments and assisted in the preparation of the manuscript; L.L. designed, performed, and interpreted the experiments; S.H.M. interpreted the experiments; Z.W. performed the experiments; R.I.L. designed, performed, and interpreted the experiments and assisted in the preparation of the manuscript; T.J.S. designed, performed, and interpreted the experiments; T.Y. performed and interpreted the experiments; A.G.K. performed and interpreted the experiments; Y.M. performed and interpreted the experiments; D.W.S. designed, performed, and interpreted the experiments; H.Y. interpreted the experiments and assisted in preparation of manuscript; J.K.C. interpreted the experiments and assisted in the preparation of the manuscript; R.P.M. interpreted the experiments and assisted in the preparation of the manuscript; J.W.W. interpreted the experiments and assisted in the preparation of the manuscript; J.H.H. designed, performed, and interpreted the experiments and assisted in the preparation of the manuscript; and C.S.A. designed and interpreted the experiments and assisted in the preparation of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Charles S. Abrams, Hematology-Oncology Division, University of Pennsylvania School of Medicine, 421 Curie Blvd, Biomedical Research Building II/III, #912, Philadelphia, PA 19104; e-mail: abrams@mail.med.upenn.edu.