GPIbα–filamin A interaction regulates megakaryocyte localization and budding during platelet biogenesis

Bernard–Soulier syndrome (BSS) is an autosomal recessive disorder caused by quantitative or qualitative abnormalities of the glycoprotein Ib-IX-V complex (GPIb-IX-V), the receptor for von Willebrand factor (VWF). BSS is a rare inherited bleeding disorder with an incidence of <1 case per million¹ and associated with thrombocytopenia, giant platelets, and increased number of megakaryocytes (MKs) in the bone marrow.² Although mutations involving GPIX are the single commonest cause of BSS, mutations involving GPIbα or GPIbβ are responsible for up to half of all cases.³ Mouse models of BSS recapitulate the human disease and are characterized by defective platelet production from structurally abnormal MKs within the bone marrow.⁴ 

Megakaryocyte (MK) maturation relies on a process of cytoplasmic maturation resulting in the formation of the demarcation membrane system (DMS),⁵ thought to function as a membrane reserve for platelet production.⁶ The current prevailing model of platelet formation from the MK relies on the use of this membrane reserve to transform the mature MK into a unique cell structure with pronounced ‘pseudopod-like’ membrane projections. These structures, known as proplatelets, are generated by the outflow and evagination of the DMS internal membrane, with the detachment of individual platelets from the tip of proplatelet extensions first observed more than a century ago.⁷ Although original descriptions of proplatelet production were based on in vitro observations of cultured MKs, in vivo observations paint a more complex picture. The emigration of entire MKs,⁸ singular long membrane extensions,^9,10 and membrane buds^11,12 have all been demonstrated within the bone marrow of mice, suggesting no single model encompasses all the intricacies of in vivo platelet biogenesis. The demonstration that MK-derived buds are identical in size to platelets, have similar ultrastructural features to platelets, and are released directly into bone marrow sinusoids, indicates that budding is a bona fide mechanism of in vivo platelet biogenesis. Whether dysregulation of the MK budding process is linked to the development of macrothrombocytopenias is yet to be established.

In mice lacking GPIb, defective formation of the DMS has been demonstrated,¹³ with amelioration of the macrothrombocytopenia after insertion of a chimeric Interleukin-4Rα (IL-4Rα) possessing the GPIbα cytoplasmic domain,¹⁴ highlighting the importance of the GPIbα cytoplasmic tail in DMS formation. GPIbα interacts with the cytoskeleton through the actin binding protein filamin A (flnA) and the role of flnA in platelets and MKs has been evaluated in mice with genetic deletion of flnA.¹⁵ FlnA-null mice are thrombocytopenic with strikingly accelerated platelet clearance, primarily due to spontaneous membrane fragmentation. Ultrastructural examination of flnA-null MKs reveals a poorly defined DMS, without characteristic platelet-forming territories.¹⁶ FlnA has numerous functions in megakaryocytes and platelets, including integrin stabilization and signaling roles mediated through RhoA¹⁷ and Syk.¹⁸ However, given that flnA has up to 100 binding partners,¹⁹ this makes it challenging to understand how specific molecular interactions govern platelet biogenesis. Moreover, mice expressing a mutant form of GPIbα lacking the terminal 24 amino acids of the GPIbα intracellular tail, with altered GPIbα signaling function through 14-3-3 and propidium iodine (PI) 3-kinase, but preserved flnA binding, have a mild reduction in platelet count and enlarged platelet size.²⁰ This raises the possibility that interactions other than those through flnA may regulate GPIbα-dependent platelet biogenesis.

The importance of the GPIbα-flnA interaction in regulating GPIbα adhesive function has been investigated in various cell models and mouse platelets.²¹ Experiments on GPIbα cytoplasmic tail mutants and crystallography studies have demonstrated a critical role of GPIbα hydrophobic residues in mediating the interaction between GPIbα and flnA.^21-23 Substitution of Phe568 (F) and Trp570 (W) with alanine (GPIb-FW mutant) prevented the association of flnA with GPIbα in Chinese hamster ovary cells.²⁴ Coexpression of the human GPIbα-FW transgene in mouse platelets also expressing endogenous mouse WT GPIb (hGPIbα^FW) have confirmed an essential role of the GPIb-flnA interaction in anchoring the receptor complex to the membrane skeleton and maintaining membrane stability under high shear stress.²⁵ This critical role for GPIbα in regulating membrane stability has been confirmed in platelets from individuals with BSS²⁶ and in mouse platelets lacking flnA.¹⁵ 

How GPIbα regulates the size of formed platelets remains unknown. Abnormalities in the formation of the DMS, defective proplatelet formation,¹³ and dysregulated MK localization within the marrow compartment⁸ have provided important insights into the potential defects in platelet production, however, it is unclear how these abnormalities in MK biology dysregulate platelet size. Using our hGPIbα^FW mutant mouse model, in which there is a selective disruption in the ability of GPIbα to bind flnA, we demonstrate that the GPIbα-flnA interaction is not essential for proplatelet formation in vitro and in vivo; however, the cytoskeletal anchorage of GPIbα is critical for the development of a well-organized MK DMS. This abnormality in the MK internal membrane reserve did not affect the ability of MKs to form proplatelets or membrane buds in vivo. However, MK buds were enlarged, and their release was misdirected into interstitial bone marrow compartment by ectopically localized MKs, suggesting that dysregulation of MK budding is likely to be an important cause of macrothrombocytopenia.

Mice were generated at Monash Mouseworks according to ethics committee approvals at the Alfred Medical Research & Education Precinct (AMREP) and Monash University School of Biomedical Sciences (application E/0471/2006/M and SOBSA/MW/2008/14BC0). All experimental procedures were approved by ethics committees at the AMREP and The University of Sydney (Project 2017/1175 and 2018/1461). To generate mice expressing either wild-type (WT) human GPIbα or GPIbα mutated at F and W, we used our previously established transgenic mice expressing either hGPIbα^WT or hGPIbα^FW on a C57BL/6 background.²⁵ To generate mGPIbα^–/–/hGPIbα^FW/cIL-4Rα mice, hGPIbα^FW mice were crossed with GPIb chimera cIL-4Rα mice whose extracellular domain of GPIbα is replaced with cIL-4Rα as previously described (hGPIbα^FW/cIL-4Rα).¹⁴ All strains were born healthy with no observable defects and had a normal life span. All mice were held in a 12-hour night/day cycle and provided ad libitum access to chow and water. Male mice aged 8 to 16 weeks were used for experimental procedures.

Biotin anti-mouse CD41 (MWReg30), CD105 (MJ7/18), AlexaFluor647–anti-mouse/rat CD61, and PE–anti-mouse CD41 and DyLight649-Donkey anti-rabbit antibodies were from BioLegend. Paraformaldehyde (PFA) and EDTA were purchased from Sigma-Aldrich. Phalloidin iFluor488 and rabbit anti-mouse flnA antibodies were from Abcam. Mouse anti-human CD42b (SZ-2) was from Beckman Coulter. Allophycocyanin (APC) and AlexaFluor568 streptavidin and Sulfo-NHS-LC-Biotin were from ThermoFisher. Rat anti-mouse GPIbβ IgG, Dylight-488 or -649 (X488, X649), CD41-PE, and CD42a-FITC monoclonal antibodies were from Emfret Analytics. Rabbit anti-mouse laminin antibody was from Novus Biologicals. WM23 (mouse anti-human GPIbα antibody) was a gift from Michael Berndt (Alfred Medical Research and Education Precinct, Monash University).

Mouse blood was collected via a submandibular bleed into EDTA-anticoagulated tube (EDTA coated pediatric microtainer tubes, BD Biosciences). Platelet count and size were analyzed using an automated blood analyzer (Sysmex KX-21N, Kobe, Japan).

Platelet surface expression of hGPIbα and IL-4Rα was measured using EDTA-anticoagulated blood, diluted 1:10 in platelet washing buffer (PWB; 4.3 mM K₂HPO₄, 4.3 mM Na₂HPO₄, 24.3 mM Na₂PO₄, 113 mM NaCl, 5.5 mM D-glucose, 10 mM theophylline, and 0.5% BSA, pH 6.5) and stained with anti-CD42b-phycoerythrin (PE) or anti-CD124-PE and analyzed by flow cytometry (BD Accuri C6 Plus) using CellQuest software (BD Biosciences) or FlowJo software (TreeStar). Platelet size was also assessed by flow cytometry using the forward scatter parameter.

Mouse platelets were isolated from EDTA-anticoagulated blood as previously described²⁷ with minor modifications, in which blood was centrifuged at a lower g force (90g × 2’) to obtain giant hGPIbα^FW mouse platelets in platelet-rich plasma (PRP). To accurately determine the size of ex vivo platelets, mice were anesthetized with intraperitoneal injection of ketamine (150 mg/kg, Ceva Animal Heal, Glenorie, Australia) and xylazine (15 mg/kg, Trot Laboratories, Glendenning, Australia), and blood was drawn via the inferior vena cava into a syringe containing enoxaparin (400 mg/mL) diluted in 100 μL sterile saline. Anticoagulated whole blood was fixed in 4% PFA at 1:1 ratio (v/v) for 30 minutes. Fixed whole blood was allowed to settle on VWF-coated glass coverslips and pretreated with botrocetin for 1 hour. Unbound cells were removed by gentle washing with PWB. Cells were imaged using differential interference contrast (DIC) microscopy using a Leica DMIRB (Leica microsystems, Wetzlar, Germany), 63× water objective, NA 1.2, and fitted with a Hammamatsu Orca camera (Tokyo, Japan). Images were acquired using the Micromanager plugin (version 1.4.2) for ImageJ.

In vivo platelet life span was determined using the previously described method.²⁸ Mice were injected with 1.5 mg Sulfo-NHS-LC-Biotin dissolved in 100 μL sterile saline via the tail vein. Blood samples were collected via tail vein into EDTA-containing syringes at the indicated times, including immediately after injection (0 h), 3 hours after injection, and daily thereafter. Whole blood (5 μL) was washed to remove plasma with PWB and platelets incubated with Streptavidin-APC (1/100) to detect biotin-labeled platelets and CD41-PE (1/100) to label total platelets, for 30 minutes at room temperature, shielded from light. The number of biotin- and CD41-labeled platelets was determined using flow cytometry (BD Accuri C5 plus), in which at least 5000 CD41⁺ events were collected and analyzed using FlowJo. The proportion of dual-stained platelets relative to the total CD41⁺ platelets over time was analyzed and expressed as platelet life span.

Mouse lineage negative (Lin^–) bone marrow cells were isolated as previously described.²⁹ Briefly, bone marrow from femora and tibiae bones was flushed out with Dulbecco modified Eagle medium (DMEM) containing 10% fetal calf serum and antibiotics (1:100 of 10 000 U/mL penicillin and 10 mg/mL streptomycin, Sigma). Lin^– cells were obtained by magnetic-activated cell sorting using a Mouse Lineage Cell Depletion Kit (Miltenyi Biotec) according to the manufacturer’s instructions. Cells were cultured at 1 × 10⁶ per mL in Dulbecco modified Eagle medium/10% fetal bovine serum/antibiotics supplemented with 50 ng/mL recombinant mouse thrombopoietin (TPO) (Miltenyi Biotec) at 37°C, 5% CO₂ for 4 to 5 days.

Proplatelet formation in vitro was performed as previously described.³⁰ Briefly, Lin^– cells isolated from murine bone marrow were seeded in duplicate in 24-well plates at 1 × 10⁵ cells per well and cultured as described above in the presence of TPO. Proplatelet bearing MKs were counted on day 4 of culture at 20× magnification using an inverted light microscope (Leica DMIRB). After proplatelet analysis, the cells were collected from culture plates and stained with anti–CD41-PE (Emfret Analytics, Eibelstadt, Germany), and CountBright counting beads (ThermoFisher) were added in accordance with the manufacturer instructions. The total number of CD41⁺ cells in each well was determined by flow cytometry. Results were expressed as proplatelet bearing MKs per 10 000 total MKs.

To measure the size of proplatelet terminal swellings, bone marrow explants were cultured as described previously.⁹ Intact marrows were flushed out from mouse femurs with Tyrode buffer containing 0.35% bovine serum albumin (BSA). Ten 0.5 mm thick transversal sections of each bone marrow sample were placed in tissue culture wells, cultured in Tyrode buffer containing 0.35% BSA and 5% mouse serum, in a 37°C CO₂ incubator in the absence of thrombopoietin. After ∼20 hours in culture, Alexa 488 anti-CD42c Ab (Emfret) was added to the living tissue sections, and each was examined for proplatelet bearing MKs under confocal and DIC microscopy (Zeiss, LSM800, 40× oil objective, NA 1.3). The size of proplatelet terminal swellings was quantified using CD42c staining and Zeiss inbuilt software and/or Image J, by quantifying the maximal transverse length of individual swellings.

Day 5 cultured, bone marrow–derived Lin^– cells were harvested, washed, and stained with rat anti-mouse CD41-PE and CD42a-FITC monoclonal antibodies (Emfret Analytics, Eibelstadt, Germany) for 30 minutes at 4°C. Free antibody was removed and surface expression of GPIX (CD42a) and GPIIb-IIIa (CD41) determined using flow cytometry (FACSCanto II, BD Biosciences). We elected to use CD42a as the principal MK marker, because CD42a is expressed at higher levels than CD42b on MKs.³¹ Furthermore, in preliminary studies, we confirmed that GPIb expression patterns on hGPIbα^FW mice were similar to that of flnA knockout (KO) mice, in that GPIb levels were normal on MKs but reduced on platelets.¹⁵ 

MK ploidy was determined as previously described.^31,32 Briefly, day 5 cultured MKs in phosphate-buffered saline (PBS) buffer (0.5% BSA/2 mM EDTA, pH 7.2) were stained with anti-mouse CD41-FITC (30 minutes at 4°C) and washed once with the same buffer. The cells were then permeabilized and stained with the nuclear dye propidium iodide (PI) by resuspending cells in hypotonic citrate buffer (1.25 mM sodium citrate, 2.5 mM sodium chloride, and 3.5 mM dextrose) containing 0.05% Triton-X 10 and 20 μg/mL of PI for 15 minutes at 4°C in dark. Ribonuclease (20 μg/mL) was then added to the cell reaction mixture and incubated for a further 30 minutes at 4°C. Ploidy was determined by the number of PI positive events within the CD41⁺ population, with a minimum of 30 000 events collected by flow cytometry (FACSCanto II, BD Bioscience).

After euthanasia, mouse femurs were isolated, fixed overnight in 2% PFA, and decalcified in 0.5M EDTA for 7 days, exchanging EDTA daily, before being embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Images (40× air objective NA 0.65) were taken on a Zeiss Axioscope upright microscope (Axio Scan 5 with Axiocam 105 camera). Images were analyzed using Zen software and ImageJ.

Mouse femurs were harvested, and bone marrow MKs were examined using TEM as previously described.³³ Briefly, mouse femurs were flushed with 2.5% glutaraldehyde in PBS overnight and embedded in Epon. Transversal thin sections (100 nm) of the entire bone marrow were obtained, stained with uranyl acetate and lead citrate, and examined under a Jeol 2100-Plus transmission electron microscope (120 kV). The number of MKs were counted under the TEM in the entire transversal section and expressed as the density per unit area (defined as one square of the grid, ie, 16 000 μm²). MKs at maturity stages I, II, and III were identified based on distinct ultrastructural characteristics, in which stage I corresponded to a cell 10 μm in diameter with a large nucleus; stage II, to a cell 10 to 20 μm in diameter containing platelet-specific granules; and stage III, mature MKs having a well-developed DMS with clearly defined platelet territories and a peripheral zone.

After euthanasia, mouse femurs were removed by dissection, fixed overnight in 2% PFA at 4°C, decalcified for 48 hours in 0.5M EDTA, and incubated overnight in 30% sucrose before embedding in optimal cutting temperature (OCT) compound (ProSciTech) and freezing at −80°C.

The lungs of the euthanized mice were inflated using 1.5% low melting-point agarose (Sigma-Aldrich) with a catheter inserted into the trachea. Lungs were then isolated, fixed overnight in 4% PFA at 4°C, and stored in PBS. Before sectioning, lungs were mounted in 5% agarose.

Decalcified bones were cut into 10 μm thick cryosections (Cryotome FSE, ThermoScientific), and lungs were sectioned (100 μm thick) using a VT 1200 vibratome (Leica Biosystems). Tissue sections were mounted on SuperFrost Plus glass slides (ThermoFisher), blocked for 1 hour with 10% fetal calf serum in PBS containing 0.5% Triton-X100, and stained overnight with primary antibodies at 4°C. After 3 washes with PBS, secondary antibodies, streptavidin, and phalloidin were incubated for 1 hour at room temperature. Hoechst (1 μg/mL) was incubated for the final 10 minutes where indicated. Slides were washed 3 times in PBS and mounted with ProLong Gold (Confocal microscopy) or ProLong Diamond (for STED).

Confocal Imaging was performed using a Zeiss LSM 880 confocal microscope. Bone marrow regions free from processing artifacts were selected for examination using a grid of 3D z-stack. For analytic purposes, a 6×6 grid (overlap 10%) z-stack was generated to cover a region ∼750 μm² (63× oil objective, NA 1.30). Because of the relative infrequency of lung MKs, larger fields were assessed for lung tissue. A similar 6×6 grid (10% overlap) was produced using a 20× objective (air, NA 0.8). Z-stacks (step size – 0.2 μm for bones, 0.8 μm for lungs) encompassed the full thickness of the bone marrow and lung tissues. Individual z-stacks were stitched using Zeiss Zen software (version 2.3). For the purposes of analysis, the entire image was divided into 4 quadrants and analyzed separately. The average value of the 4 regions was determined for each mouse. Images were analyzed using ImageJ (version 1.53c) or Imaris (Bitplane AG, Schlieren, Switzerland; version 9.8). Stitched images were examined for the number of MKs, platelets, and their localization using ImageJ.

Bone marrow intravascular platelets were counted as platelet-sized, CD41⁺ structures, within vessels delineated by laminin. Interstitial platelet buds were determined using the same criteria, located clearly in the bone marrow interstitium and within a vascular structure. The diameter of MKs and interstitial buds was determined as the maximal length of a line drawn through individual buds. MK buds were defined by its link to the MK outer membrane based on CD41 staining in at least 1 plane. Lung proplatelets were counted as CD41⁺ structures with thin, irregular protrusions as previously described.³⁴ 

Bone marrow cryosections were prepared as for confocal imaging and mounted with ProLong Diamond (ThermoFisher). Images were acquired using a Leica (Wetzlar, Germany) SP8 imaging platform (93× glycerol objective, NA 1.30). Stimulated emission depletion (STED) z-stacks (step size, 0.12-0.20 μm) were collected using LAS X software (version 3.5.7). Image deconvolution was performed using Huygens Software (version 20.04, Scientific Volume Imaging, Hilversum, The Netherlands). Three dimensional renders were created using Imaris (Bitplane AG; version 9.8).

Cryosections of mouse femurs of size 3 μm were prepared for confocal imaging. After staining, slides were postfixed with 0.5% PFA for 5 minutes. The imaging buffer consisted of standard imaging buffer (tris/HCl 100 mM, NaCl 20 mM, and glucose 10% vol/vol), mercaptoethylamine (0.1 M), and glucose oxidase/catalase (600 μg/mL and 60 μg/mL, respectively). Ground-state depletion (GSD) images were acquired on a Leica ST GSD (160× oil objective, NA 1.43). d-STORM reconstructions were performed using the ThunderSTORM (version 1.3)³⁵ plugin for ImageJ.

Statistical analysis was performed using GraphPad Prism. Statistical significance was determined using either 2-tailed Student t test or analysis of variance as appropriate (with Bonferroni post hoc test for multiple comparisons). P values <.05 were considered statistically significant (∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001).

We have previously generated a mutant human GPIbα transgene that has a selective defect in flnA binding (Phe568Ala and Trp570Ala; FW), termed hGPIbα^FW. This transgene and a WT human GPIbα (hGPIbα^WT) control have been expressed in C57BL/6 mice, generating mice with both human and endogenous mouse GPIbα (mGPIbα^+/+). To generate mice expressing human GPIbα alone on mouse platelets, we crossed GPIbα deficient mice (mGPIbα^–/–) with mice expressing either hGPIbα^WT or mutant hGPIbα^FW to generate mGPIbα^–/–/hGPIbα^WT or mGPIbα^–/–/hGPIbα^FW mice. The first generation of mGPIbα^+/−/hGPIbα^WT mice and mGPIbα^+/−/hGPIbα^FW were bred to obtain homozygous mGPIbα^–/–/hGPIbα^WT or mGPIbα^–/–/hGPIbα^FW mice (supplemental Figure 1, available on the Blood website); throughout the remainder of this study, we refer to these mice as hGPIbα^WT or hGPIbα^FW, respectively. Both strains were born healthy with no observable defects and had a normal life span.

BSS is typically associated with mutations in either the GPIbα, GPIbβ, or GPIX genes, leading to macrothrombocytopenia and deficiency of the GPIb-IX-V receptor complex. Genetic deletion of flnA also leads to macrothrombocytopenia; however, the expression of GPIb-IX-V on the surface of platelets is preserved, albeit at ∼50% of normal levels.¹⁶ Consistent with this, platelets from hGPIbα^FW mice had a 57.6% reduction in GPIb expression levels relative to hGPIbα^WT controls (Figure 1A). GPIb expression was broad and heterogeneous within the hGPIbα^FW platelet population (Figure 1B). To distinguish the effects of reduced hGPIbα expression levels on the phenotypes of mouse platelets and MKs from alterations related to defective hGPIbα-cytoskeletal anchorage, we selected transgenic WT controls with lower levels of expression of the human hGPIbα^WT transgene, similar to hGPIbα^FW mice (supplemental Figure 2A). The circulating platelet count in transgenic WT mouse controls was between 400 × 10⁶ and 500 × 10⁶ per mL, ∼50% lower than nontransgenic C57BL/6 mice (950 × 10⁶ to 1050 × 10⁶ per mL; supplemental Figure 2B).

Platelet size was determined using mean platelet volume and changes in forward scatter of CD41-labeled platelets. Both methods revealed enlarged platelets from hGPIbα^FW mice (Figure 1C and supplemental Figure 2C). DIC microscopy using washed platelets also confirmed the presence of abnormally large, discoid platelets (Figure 1D). hGPIbα^FW mice were thrombocytopenic (Figure 1E), with platelet counts in whole blood ranging from 150× 10⁶ to 200 × 10⁶ per mL. To exclude the possibility that the low platelet count was due to inaccurate counting of very large platelets by the analyzer, we confirmed the low platelet count by performing manual counting with a hemocytometer using isolated mouse platelets. There was no correlation between GPIbα expression level and platelet count in hGPIbα^FW mice (supplemental Figure 2D), indicating that lower levels of GPIbα was unlikely to be the primary cause of the macrothrombocytopenia. All other blood counts were similar between strains, with respect to white blood cell and red blood cell parameters (supplemental Table 1).

The similarity in platelet phenotype between hGPIbα^FW mice and flnA-deficient platelets raised the possibility that hGPIbα^FW platelets may be more rapidly cleared from the circulation. Previous studies have demonstrated that flnA-deficient platelets are intrinsically fragile and undergo microvesiculation in blood and in vitro and are cleared by macrophages.¹⁵ To ascertain whether the thrombocytopenic state in hGPIbα^FW mice is due to impaired platelet production or excessive clearance, we performed in vivo labeling and determined the lifetime of circulating platelets. Interestingly, we found hGPIbα^FW mice had a normal circulating lifetime (Figure 1F), suggesting the defect in platelet numbers is most likely related to defective thrombopoiesis, rather than intravascular microvesiculation and accelerated clearance, as seen with flnA-null platelets.

To evaluate whether dysregulated platelet production or release into the circulation was the underlying cause of thrombocytopenia in hGPIbα^FW mice, we analyzed MKs within femoral bone marrow. There was a 1.4-fold increase in the number of MKs in the bone marrow of hGPIbα^FW mice (Figure 2A), and these cells appeared normal by hematoxylin and eosin staining (supplemental Figure 3). Platelet production requires the development of an elaborate internal demarcation membrane structure in MKs. To investigate the potential ultrastructural changes in the DMS of hGPIbα^WT mice, we performed superresolution STED microscopy on bone marrow MKs. Using CD41 to label the external and internal membrane reserves of the MK, we observed abnormal formation of DMS internal membranes in the hGPIbα^FW mice (Figure 2B). In contrast to the dense CD41 staining of platelet-demarcation territories within hGPIbα^WT MKs, more porous and disorganized CD41 staining of the platelet-forming territory was apparent in hGPIbα^FW MKs (Figure 2B-C). To further investigate these observations, we examined the ultrastructure of MKs using TEM, which confirmed a striking disorganization of the DMS ultrastructure and the absence of the MK peripheral zone in mature MKs (Figure 2D). Although we found no clear differences in the frequency of the 3 stages of MK ultrastructural maturation,³³ there was a greater frequency of morphologically abnormal stage III MKs in hGPIbα^FW mice (Figure 2E). These features were confirmed using high magnification of the DMS from STED images of CD41-stained MKs (Figure 2F). Together, these findings demonstrate that the interaction of GPIbα with flnA is critical for normal DMS development.

To examine the impact of disrupting the link between GPIbα and flnA on the subcellular distribution of flnA, we performed both confocal and STED microscopy on bone marrow MKs stained with an anti-flnA antibody. Although flnA was distributed homogenously throughout the cytoplasm of hGPIbα^WT MKs, it was localized to a peripheral ring in the MKs in hGPIbα^FW mice (Figure 2G). Quantitative analysis revealed that this pattern of flnA distribution was present in ∼75% of hGPIbα^FW MKs compared with 14% of hGPIbα^WT MK controls (Figure 2H and supplemental Figure 4). flnA binds to the major platelet adhesion receptors, GPIbα and integrin α_IIbβ₃, and facilitates their linkages with the actin cytoskeleton. Superresolution GSD microscopy was used to analyze the flnA distribution at the single-molecule level. Although discrete foci of flnA staining were evident throughout the cytoplasm in association with F-actin and CD41 in hGPIbα^WT mice, we found a striking accumulation of flnA at the cell periphery, not in direct connection with CD41 or F-actin in hGPIbα^FW mice (Figure 2I). We next evaluated the colocalization of flnA with GPIbα and F-actin in both mouse strains and observed loss of the colocalization of flnA with GPIbα and F-actin in hGPIbα^FW mice (Figure 2J-K). Interestingly, F-actin and GPIbα were still found homogenously distributed throughout the MK, despite disrupted linkage between flnA and GPIbα in the GPIbα^FW cells. Together, these findings indicate that the flnA-GPIbα interaction is pivotal to the normal regulation of flnA distribution within MKs.

To investigate whether changes in the DMS ultrastructure in hGPIbα^FW mice lead to defective MK proplatelets, we initially examined proplatelet formation in vitro using cultured MKs derived from hGPIbα^FW and hGPIbα^WT mice. MKs derived from hGPIbα^FW mice underwent normal maturation in vitro, as evidenced by normal MK ploidy and the expression of cell-surface markers of mature MKs (Figure 3A-B). Moreover, disrupting the link between GPIbα and flnA did not impact on the ability of MKs to differentiate into proplatelets (Figure 3C), however, the development of terminal swellings from proplatelet tips were larger in the hGPIbα^FW MKs from bone marrow explant cultures than the hGPIbα^WT controls (Figure 3D and supplemental Figure 5). MKs with a typical proplatelet morphology have been demonstrated to be plentiful in the lung, whereas proplatelets within the bone marrow are relatively uncommon (<5% of MKs).¹¹ We therefore assessed proplatelet formation within the lungs of mice and found no difference in the number of pulmonary proplatelets in hGPIbα^FW mice compared with hGPIbα^WT controls (Figure 3E-F). Moreover, there was no apparent difference in the morphology of proplatelets in the lungs of both mouse lines. These findings suggest that the flnA-GPIbα interaction and corresponding alterations in the DMS are dispensable for MK cytoplasmic maturation and proplatelet formation in vitro and in vivo.

The description of platelet biogenesis by membrane budding provides an additional mechanism of in vivo platelet production and a potential new avenue for mechanistic understanding of abnormal platelet formation. To examine whether the release of platelet-forming buds was dependent on GPIb-flnA interaction, we evaluated the ability of MKs from hGPIbα^FW mice to form buds in the native marrow. Using both 3D STED microscopy and confocal optical sectioning through bone marrows of CD41-stained MKs, we assessed the frequency and size of membrane buds, using previously published protocols.¹⁰ Buds derived from both hGPIbα^WT and hGPIbα^FW MKs were found as extensions of the MK membrane into bone marrow sinusoids (Figure 4A). Moreover, the number of MKs undergoing budding and the number of visible buds per MK were similar between the WT and GPIbα mutant strain (Figure 4B-C). Consistent with the demonstration of larger platelets in circulation, bud sizes were also increased in hGPIbα^FW mice (Figure 4D), and the bud size approximated that seen of platelets observed in bone marrow sections (Figure 4E). Analysis of granule content confirmed the presence of VWF in MK buds, with no difference in the proportion of buds expressing VWF between hGPIbα^FW and WT mice (Figure 4F). Taken together, these data indicate that the GPIb-flnA interaction and DMS are dispensable for membrane budding from the MK surface, however, the cytoskeletal anchorage of GPIb appears to be important in regulating the size of membrane buds.

Budding MKs typically release platelets in a polarized fashion directly into the sinusoidal space.¹¹ However, we observed prominent bud-like structures in the interstitial compartment of the bone marrow of hGPIbα^FW mice (herein referred to as interstitial buds), with limited numbers of bud-sized membrane structures in the bone marrow sinusoids (Figure 5A). Despite a similar number of bud-like structures visible within an individual region of bone marrow (Figure 5B), we found an approximately twofold increase in the number of buds within the interstitial compartment of hGPIbα^FW mice (Figure 5C). To confirm that these interstitial buds were indeed likely to represent platelet precursors, rather than previously described MK-derived blebs or microvesicles,¹² we assessed interstitial buds for the presence of platelet-specific granule contents. 2D superresolution microscopy demonstrated interstitial buds express membrane CD41 with clusters of P-selectin and VWF distributed heterogeneously throughout the bud cytoplasm (Figure 5D-E). Quantitatively, 66% of buds in hGPIbα^FW bone marrow were found in the interstitial compartment, whereas the majority of buds in hGPIbα^WT mice were located intravascularly. This dysregulated bud release coincided with aberrant localization of megakaryocytes away from the vascular sinusoids, suggesting that the interstitial release of buds from megakaryocytes may contribute to the low circulating platelet count in hGPIbα^FW mice (Figure 5F-G and supplemental Figure 6A).

The insertion of a cIL-4Rα into BSS mice is known to reverse the macrothrombocytopenia seen in GPIbα KO.¹⁴ The cIL-4Rα fusion protein possesses a transmembrane domain and cytoplasmic tail of GPIbα, fused to an extracellular domain of the cIL-4Rα, providing an anchorage site for the membrane skeleton through an independent GPIbα tail. To provide further evidence that the size of buds and location of buds within the bone marrow is regulated by the GPIbα-flnA interaction, we crossed cIL-4Rα mice with hGPIbα^FW mice to produce hGPIbα^FW mice, which also coexpressed the cIL-4Rα. The expression of the cIL-4Rα construct in hGPIbα^FW mice did not alter GPIbα levels on the surface of platelets (Figure 6A). However, it did ameliorate the macrothrombocytopenia (Figure 6B-C), albeit to a slightly lower level than that seen in hGPIbα^WT controls. When the flnA distribution was evaluated in bone marrow megakaryocytes, we found that although FW mice showed a threefold increase in the peripheral flnA intensity, cIL-4Rα mice showed normalization of the flnA fluorescence intensity (Figure 6D) and homogenous flnA distribution through the cytoplasm (Figure 6E). Moreover, the CD41 staining of DMS membranes revealed a normalization of internal membrane structure (Figure 6F-G). This improvement in MK ultrastructure was associated with normal bud size (Figure 6H) and numbers within the bone marrow (Figure 6I and supplemental Figure 6B) and a reduction in the proportion of buds within the bone marrow interstitium (Figure 6J). This redistribution of interstitial buds correlated with changes in MK localization toward bone marrow sinusoids, with a similar proportion of interstitial and sinusoidal MKs as hGPIbα^WT controls (Figure 6K). These studies confirm an important role of the cytoplasmic tail of GPIbα in regulating budding morphogenesis. They also suggest a major role of the cytoskeletal anchorage of GPIbα in regulating the localization of MKs within the bone marrow and the corresponding spatial release of buds.

Platelet production by MKs in vivo has been long thought to be exclusively mediated by proplatelets. Historically, 2 models have been proposed to explain steady-state platelet formation; platelets released from proplatelet tips that extend from bone marrow MKs into blood vessels, or alternatively, via the fragmentation of MKs into platelets in the pulmonary microcirculation.³⁴ A third model of platelet biogenesis has been recently proposed, involving the budding of platelet-like structures from the surface of MKs into bone marrow sinusoids.¹¹ Our studies, which have used a new mouse model of macrothrombocytopenia, involving the defective binding of human GPIbα to flnA, have demonstrated normal MK numbers in the bone marrow and, unexpectedly, no major defects in the proplatelet formation in vivo. The principal defect in the platelet production in hGPIbα^FW mice appeared to be related to the dysregulated MK budding. In particular, the aberrant generation of enlarged buds represents a previously undescribed mechanism of giant platelet formation. Moreover, dysregulated MK localization in the bone marrow, away from vascular sinusoids, led to ectopic release of membrane buds into the bone marrow interstitium, providing a plausible mechanistic explanation for the thrombocytopenia observed in these mice.

We and others have previously demonstrated that deletion of flnA or its interaction with GPIbα result in fragile platelets under high shear conditions, which may undergo physical fragmentation and microvesiculation in the circulation.^15,25 We had anticipated that thrombocytopenia and the reduced levels of GPIb surface expression in hGPIbα^FW mice may reflect unstable platelet membranes and accelerated peripheral platelet clearance. Surprisingly, we found no evidence of increased platelet clearance in hGPIbα^FW mice, suggesting that the primary cause of thrombocytopenia is due to defective platelet production. Our findings in the hGPIbα^FW mice indicate that disrupting flnA binding to GPIbα does not prevent proplatelet formation in vitro, nor did we see reduction in the frequency of pulmonary proplatelets in vivo. These in vivo findings are consistent with previous in vitro studies on an isogenic stem cell model of megakaryopoiesis, in which the expression of mutant forms of flnA that have a selective defect in GPIbα binding, exhibits normal proplatelet formation.¹⁷ 

Our studies have demonstrated that disrupting the linkage between flnA and GPIbα had a major impact on flnA distribution in the MK cytoplasm, resulting in a striking impairment in DMS formation. The DMS is primarily considered the main source of internal membrane reserves necessary for proplatelet formation. Our studies in hGPIbα^FW mice demonstrate that MK membrane budding per se is not dependent on a well-organized DMS. However, the impaired development of the platelet-forming territories in the demarcating membrane system of MKs appears to be linked to the dysregulated bud sizes in hGPIbα^FW mice. Similar to previous reports on flnA-null³⁶ MKs, dysregulated flnA localization was associated with a lack of peripheral zone in hGPIbα^FW bone marrow MKs. Whether flnA redistribution directly or indirectly affects the peripheral zone and whether this is also linked to aberrant budding remains to be established.

FlnA can bind to up to 100 different proteins and is anchored to the membrane cytoskeleton through its interactions with integrins^37,38 and ion channel proteins^39,40 in most mammalian cells.⁴¹ GPIbα is primarily expressed in platelets and MKs, and our studies highlight the critical importance of GPIbα in regulating flnA subcellular distribution in the latter cells. The aberrant flnA redistribution observed in hGPIbα^FW bone marrow MKs was corrected by the coexpression of the GPIbα tail in cIL-4Rα–hGPIbα^FW bone marrow MKs, confirming the critical role of the physical linkage between GPIbα cytoplasmic tail and flnA in regulating flnA subcellular distribution. Redistribution of flnA has been previously noted in platelets and HEK293T cells,^42,43 and it is believed to be influenced by the ratio of flnA and GPIbα expression in a reciprocal manner, with an imbalanced ratio impacting platelet size and production. These studies have demonstrated that reciprocal expression of flnA and GPIbα at an optimal ratio is important to prevent protein sequestration in the endoplasmic reticulum and for the generation of normal sized platelets. Our studies here have revealed that rather than endoplasmic reticulum sequestration, in the absence of its GPIbα binding, flnA becomes enriched in the MK periphery. FlnA binds to and regulates the filamentous actin network; however, in MKs, we could find no evidence that flnA redistribution in hGPIbα^FW mice was associated with gross alterations in the F-actin network. Previous investigations of filaminopathies causing macrothrombocytopenia lead to the identification of a patient with a flnA mutation that incorporated the GPIbα binding site.³⁶ Notably, this individual had preserved interaction between flnA and F-actin, however, it is unclear whether this filamin mutation fully disrupted the link between flnA and GPIb. F-actin at the periphery of MKs is tightly coordinated and assembled into podosome-like structures that facilitate the docking of MKs to sinusoids.⁴⁴ Whether the peripheral enrichment of flnA in hGPIbα^FW mice leads to aberrant F-actin assembly and abnormal podosome formation remains to be explored.

Our findings demonstrate an important role of the GPIb-flnA interaction in regulating the morphogenesis of MK-derived membrane buds. At the molecular level, flnA plays a major role in regulating the mechanical stability of cell membranes in numerous cell types.⁴⁵ FlnA is a prominent actin filament cross-linking protein,⁴⁶ present in the cortex of numerous cells types,^47-50 and facilitates the link between the actin cytoskeleton and plasma membrane. The importance of flnA in regulating the delamination of the plasma membrane from the cortical cytoskeleton was first recognized in flnA-deficient human melanoma cells, leading to blebbing of plasma membrane.⁵¹ Although our findings demonstrate that the physical link between GPIbα and flnA plays a major role in regulating the internal membrane organization of the DMS, it remains to be seen how important this interaction is in regulating the stability of the membranes within the DMS and MK plasma membrane and how important these additional membrane changes may be to altered bud morphogenesis.

Our studies on the hGPIbα^FW mice have revealed the presence of large numbers of bud-like structures within the interstitial regions of the bone marrow. Ectopic release of platelets into the bone marrow interstitial space, presumed by aberrant proplatelets, has previously been proposed as a potential cause of microthrombocytopenia.^52-56 Our studies demonstrate an alternative explanation for ectopic platelet release through the production of buds by MKs within the bone marrow interstitium. Abnormal MK localization has been identified in the context of GPIbα^-/- mice,⁸ and our preliminary studies on GPIbα^–/– mice have confirmed aberrant budding in the interstitial compartment ( M.E., Y.Y., and S.P.J., written communication 22 September 2023), suggesting a mechanistic link between ectopic MKs and interstitial, extravascular bud-like structures. Based on the quantitative estimates of Potts et al,¹¹ 89% of circulating platelets in mice are derived from MK buds. Therefore, the 45% reduction in the number of buds localized to bone marrow sinusoids in hGPIbα^FW mice is likely to contribute substantially to the overall 60% reduction in the circulating platelet counts. The mechanisms leading to impaired sinusoidal engagement of MKs from hGPIbα^FW mice is currently unclear, however, it does not seem related to GPIb expression or ligand binding. This is based on our observations that the coexpression of cIL-4Rα in hGPIbα^FW mice restored the MK sinusoidal attachment and that high GPIbα-expressing hGPIbα^FW mice do not display improved platelet count relative to lower expressors. This impairment is possibly intrinsic to the observed redistribution of flnA and consequent changes to the cytoskeleton required for megakaryocyte docking to sinusoids. It has been observed that a high flnA concentration is associated with tighter F-actin bundles that are stiffer under biomechanical force,⁵⁷ whereas lower flnA concentration is linked to more dynamic and softer F-actin.⁵⁸ Furthermore, a temporal dissolution of the peripheral zone has been observed during proplatelet release from MKs in the bone marrow.⁵⁹ Because hGPIbα^FW MKs also lack peripheral zones, it is plausible that this may also lead to aberrant bud release.

Evaluation of in vivo bone marrow MKs, particularly with respect to the evaluation of cytoskeletal structures at single-molecule resolution, is an important avenue for future investigation. However, this is technically difficult to achieve with respect to membrane buds. Sectioning tissue thinly enough to conduct superresolution microscopy, while simultaneously evaluating MKs buds in 3D, is a major experimental challenge. Although focussed ion beam-scanning electron micrsocopy (FIB-SEM) has been used to evaluate MKs previously,⁶⁰ a multicolor 3D correlative light-electron microscopy (CLEM)-based approach is likely to be required to provide sufficient spatial resolution to identify subtle alterations in the membrane and cytoskeletal structures associated with DMS and bud formation. Combining genetic mouse models of macrothrombocytopenia with advances in single-molecule imaging should provide important new insights into the molecular mechanisms regulating bud morphogenesis and platelet generation.

Our demonstration that the GPIbα cytoskeletal interaction is a critical determinant for regulating the size and localized release of buds into the bone marrow sinusoids provides important new insights into the causes of macrothrombocytopenia. Whether dysregulating MK budding is responsible for other causes of macrothrombocytopenia remains to be seen. Nonetheless, these new insights may open new avenues to modulate platelet biogenesis and ameliorate platelet production defects in vivo.

The authors acknowledge the technical and scientific assistance of Sydney Microscopy & Microanalysis, the University of Sydney node of Microscopy Australia, in particular, the assistance provided by Neftali Flores Rodriguez.

This work was supported by the Kanematsu Research Award (Royal College of Pathologists of Australasia) (M.L.E.) and New South Wales Health (NSW, Australia) Cardiovascular Capacity Building Grants (S.P.J.); and supported by a National Health and Medical Research Council of Australia Investigator Grant (Leadership 3, No.1176016; S.P.J.).

Contribution: M.L.E. performed experiments, analyzed and interpreted data, and wrote and revised the manuscript; A.T., I.A., R.S., J.P., A.E., and J.M. performed experiments and assisted with data interpretation; S.L.C. generated mice, performed experiments, and analyzed data; F.H.P, Z.M.R., and F.L provided intellectual contributions and reviewed the manuscript; S.M.S. provided intellectual contributions, reviewed/revised the manuscript, and assisted with the preparation of figures; S.T. assisted with planning experiments and data interpretation; and Y.Y. and S.P.J. wrote and revised the manuscript, interpreted data, and provided academic oversight of the project.

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

Correspondence: Shaun P. Jackson, Heart Research Institute, and Charles Perkins Centre, The University of Sydney, Level 3, D17, Camperdown, NSW, 2006, Australia; email: shaun.jackson@hri.org.au.