Regulation of actin polymerization by tropomodulin-3 controls megakaryocyte actin organization and platelet biogenesis

Blood platelets arise from megakaryocyte (MK) precursors.¹  During maturation, MKs become polyploid and grow up to 50 to 100 μm in size, forming a well-organized demarcation membrane system (DMS) (also termed invaginated membrane system) in their cytoplasm, defining the cytoplasmic domains of platelet territories, and providing the membrane reservoir for release into proplatelet protrusions.^2,3  Released proplatelets are then processed into mature platelets in the circulation.^4,5  Platelet biogenesis relies on cytoskeletal rearrangements of both microtubules and actin filament (F-actin) networks.^2,3,6,7  Microtubule assembly and dynein-mediated microtubule sliding are important for proplatelet elongation and organelle trafficking into proplatelet extensions,^2,6,8,9  whereas F-actin plays a role in numbers and branching of proplatelet protrusions,¹  and their morphology, size, and release from the proplatelet extensions.^2,3,7  Analyses of human mutations and gene-targeted mice have identified several actin cytoskeletal components involved in this process,⁷  including nonmuscle myosin IIA (NMIIA) heavy chain (MYH9),^10-13  actin depolymerizing factor (ADF)/n-cofilin,¹⁴  filamin A (FLNA),^15,16  Wiskott-Aldrich syndrome protein (WASp),¹⁷  Rho GTPases,^18,19  the formin DIAPH1 (mDia1),²⁰  and α-actinin1.^21,22  Human mutations and mouse knockouts for these actin cytoskeletal factors often result in thrombocytopenias with abnormally sized and poorly functioning platelets, as well as aberrant MKs characterized by perturbations in F-actin organization and impaired formation of proplatelet extensions. However, MK numbers and polyploidy are relatively unaffected by mutations in these cytoskeletal proteins.^11,15,18 

In addition to defects in F-actin organization, a common feature underlying the MK phenotypes of some of these mutants is an inability to elaborate the DMS that defines the cytoplasmic territories destined to become platelets and supplies the extensive membrane reservoir for proplatelet formation.²³  This is particularly evident in Myh9^−/− (NMIIA-null) mouse MKs, which have an uneven organelle distribution and fragmentary DMS with defective F-actin organization.^10,11  Patel-Hett et al have demonstrated that the DMS is supported by a spectrin-actin cytoskeleton,²⁴  similar to platelet membranes.²⁵  Transduction of mouse MKs with a spectrin tetramer-disrupting construct inhibits formation of the DMS, impairs formation of proplatelet extensions, and leads to loss of the characteristic barbell shape of newly formed proplatelets.²⁴  Actomyosin contraction and spectrin-actin networks both rely upon F-actin stability, yet other studies have demonstrated that platelet biogenesis depends upon F-actin disassembly and turnover promoted by ADF/n-cofilin.¹⁴  To reconcile these apparently contradictory roles for F-actin dynamics, we hypothesize that a balance between F-actin stability and turnover may be critical for platelet biogenesis, with F-actin stabilizing proteins in MKs counteracting F-actin disassembly-promoting proteins such as ADF/n-cofilin.

Tropomodulins (Tmods) play key roles in stabilizing F-actin networks by binding tropomyosins (TMs) and capping the pointed ends of TM-associated F-actin.²⁶  TMs bind along the length of F-actin and inhibit disassembly by ADF/cofilin or severing by gelsolin, and TM-mediated F-actin stabilization is enhanced by Tmods.^27-33  Tmods are components of membrane-associated TM–F-actin networks, including the spectrin-actin network of the erythrocyte membrane (Tmod1), ocular lens fiber cell membranes (Tmod1), polarized epithelial cell plasma membranes (Tmod3), and the sarcoplasmic reticulum membranes of skeletal muscle (Tmod3), where Tmod-mediated stabilization of TM–F-actin is critical for membrane morphology, mechanics, and physiology.^34-37  Tmod3 is also involved in endothelial cell migration,³⁸  erythroblast enucleation and erythroblast-macrophage adhesion in erythropoiesis,³⁹  and insulin-stimulated GLUT4 trafficking in adipocytes.⁴⁰ 

In this study, we show that Tmod3 is the only Tmod in platelets and MKs and confirm that TM4 (Tpm4.2, encoded by the Tpm4/δTm gene⁴¹ ) is the predominant TM in MKs and platelets.^23,42  We investigated a role for Tmod3 in platelet formation by studying Tmod3^−/− mice, which are embryonic lethal at E14.5 to E18.5 with anemia due to impaired fetal liver erythropoiesis.³⁹ Tmod3^−/− embryos also display hemorrhages, which may be partly due to reduced platelet numbers and enlarged platelets in the embryonic circulation. Tmod3^−/− fetal liver MKs have reduced proplatelet formation in vitro, but no significant changes in MK ploidy and a moderate increase in MK numbers. However, Tmod3^−/− MK cytoplasmic morphogenesis is impaired, with insufficient DMS formation and aberrant organelle distribution, and many proplatelet buds from cultured MKs are abnormally large with incorrect numbers of von Willebrand factor (VWF) negative and positive granules. Tmod3 associates with TM4 and the actin cytoskeleton in MKs, and absence of Tmod3 leads to defective F-actin polymerization and organization in MKs and proplatelet buds. Deletion of Tmod3 may disrupt F-actin in MKs via perturbation of TM4–F-actin interactions, resulting in F-actin instability and reorganization, leading to defective DMS formation and organelle distribution, thereby impairing proplatelet formation, morphology, and organelle content.

Tmod3^−/− embryos were as described.³⁹  All experiments were performed according to the National Institutes of Health animal care guidelines, as approved by The Scripps Research Institute’s Institutional Animal Care and Use Committee.

E14.5 embryos were subjected to whole-mount X-gal staining as described.³⁹  Embryos were then fixed, paraffin-embedded, and sectioned for H&E staining.

Dissociated cells from E13.5 to E14.5 mouse fetal livers were cultured in Dulbecco’s Modified Eagle’s medium (Gibco Life Technologies) with 5 ng/mL recombinant murine thrombopoietin (mTPO, Life Technologies), 10% fetal bovine serum (HyClone, Thermo Scientific), penicillin, and streptomycin.⁴³  MKs were collected using a 2-step bovine serum albumin (BSA) density gradient on day 3 and incubated in Dulbecco’s Modified Eagle’s medium with mTPO for an additional 24 hours until proplatelet formation.⁴³ 

Whole mouse embryos were processed for cryosectioning and labeled for antibody as described.³⁹  For imaging of fetal liver cytospins, E14.5 fetal livers were dissociated into Dulbecco’s phosphate-buffered saline (PBS), suspended cells were fixed in 4% paraformaldehyde overnight at 4°C, permeabilized, blocked, and labeled with antibodies in suspension, and then spun onto slides.³⁹  For staining of proplatelet-forming MKs from in vitro cultures, cells were fixed in 4% paraformaldehyde for at least 30 minutes at room temperature and then centrifuged onto poly-l-lysine coated slides.⁴³  Cells or tissue preparations were permeabilized in 0.3% Triton X-100 in PBS and blocked overnight at 4°C in 2% BSA with 1% goat serum in PBS + 0.1% Triton X-100. Slides were incubated with primary antibodies (see supplemental Table 1 on the Blood Web site) overnight at 4°C, followed by secondary antibodies (supplemental Table 1) for 2 hours at room temperature, and mounted in ProLong Gold antifade reagent (Life Technologies). Images were acquired using a Zeiss LSM 780 confocal laser scanning fluorescence microscope, and analyzed using Volocity 6.3 (PerkinElmer) or ImageJ software as indicated in figure legends.

Peripheral blood from E14.5 mouse embryos was collected by allowing embryos to bleed out into Dulbecco’s PBS with 5 mM EDTA and added to K₃EDTA containing tubes (Becton Dickinson). Blood or dissociated fetal liver cells were fixed overnight in ice-cold 4% paraformaldehyde and 1.5% glutaraldehyde in 0.1 M sodium cacodylate at pH 7.2. Cells or tissues were processed for embedding and thin sectioning as described.⁴⁴  Sections were examined on a Philips CM100 electron microscope (FEI). Images were collected using a MegaView III CCD camera (Olympus Soft Imaging Solutions GmbH).

Dissociated E14.5 fetal liver cells were cultured in StemSpan Serum-Free Expansion Medium (Stemcell Technologies) supplemented with 30% BSA, insulin, and transferrin (Stemcell Technologies) and 50 ng/mL mTPO (ConnStem) for 4 days at 37°C. Cultured cells were harvested for flow cytometry as described (supplemental Figure 2).⁴⁵ 

Mouse blood was collected via cardiac puncture of 3-month-old male mice directly into K₃EDTA Vacutainer tubes (Becton Dickinson). Human peripheral blood from healthy donors was collected by venipuncture into K₃EDTA according to an Institutional Review Board–approved human subjects blood collection protocol (11-5773). Red blood cells and leukocytes were removed by centrifugation at 200g for 10 minutes, followed by collection of platelets from the plasma by centrifugation at 1000g for 5 minutes. MKs were obtained from in vitro cultured mouse fetal liver cells on day 3 as described above. Western blotting was performed as described.³⁹  Primary and secondary antibodies are shown in supplemental Table 1.

Cultured MKs were lysed at 4°C in modified radioimmunoprecipitation assay buffer (50 mM Tris pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.1% sodium dodecyl sulfate, and 1% NP40) supplemented with protease inhibitor cocktail (1:100, Sigma-Aldrich), centrifuged for 10 minutes at 4°C in an Eppendorf microcentrifuge to remove insoluble material, and affinity-purified rabbit anti-Tmod3 (1 µg), rabbit anti-TM4, or pre-immune control rabbit IgG (1 µg) was added. After incubation for 1 hour, μMACS Protein A MicroBeads (Miltenyi Biotec) was added, incubated overnight at 4°C, run through MACS separation columns (Miltenyi Biotec), and antibody-protein complexes were eluted with sodium dodecyl sulfate-gel sample buffer. Western blotting of co-IP products and input whole-cell lysate were then performed, with images acquired with an Odyssey Imager (LI-COR Biosciences).⁴⁶ 

Data are presented as mean ± SEM for small sample sizes (<10) and mean ± SD for large sample sizes (>30). A 2-tailed unpaired Student t test was used for data analysis. P < .05 was considered statistically significant. Prism Version 5 software (GraphPad) and Microsoft Excel were used for statistical analysis.

Tmod3 is the only Tmod isoform in the human platelet proteome, with an estimated 11 900 copies per platelet, considerably less abundant than cofilin (CFL1 + CFL2) at 337 200 copies, or profilin (PFN1) at 504 000 copies per platelet.⁴²  Western blotting confirms that Tmod3 is present in both human and mouse platelets, whereas no Tmod1 is detected (Figure 1A). Similar results are obtained using in vitro cultured MKs from E13.5 mouse fetal livers (Figure 1B). LacZ staining of fetal livers from Tmod3^+/− mice³⁹  also reveals Tmod3 expression in large multinucleated MKs (Figure 1C), along with fainter blue staining in other cells, as previously shown.³⁹  Immunostaining for CD41 and β-gal confirmed abundant β-gal in CD41⁺, multinucleated MKs (Figure 1D). Indeed, Tmod3 expression appears to be higher in MKs in comparison with other fetal liver cell types. Based on the restricted expression of Tmod2 and Tmod4 in neurons and skeletal muscle, respectively,²⁶  Tmod3 is the only Tmod in platelets and fetal liver MKs.

We previously noticed small hemorrhages in ∼75% of Tmod3^−/− embryos at E14.5 (Figure 2A).³⁹  As Tmod3 is expressed in endothelial cells^38,39  and platelets (Figure 1), hemorrhages may be due to endothelial barrier and/or platelet dysfunction. We focused on the platelet phenotype, because Tmod3^−/− embryos had fewer circulating platelets (∼50% of wild-type [WT]) (Figure 2B,C), with larger and more variable sizes (Figure 2B-E). Tmod3^−/− platelets were also more irregular in shape and often displayed small protrusions, unlike WT platelets (Figure 2D). TEM revealed more vesicular structures with clear lumens in some Tmod3^−/− platelets (Figure 2F), and confocal microscopy of GPIbα revealed internal staining in contrast to typical membrane staining of WT platelets (supplemental Figure 1). This could be due to partial activation of Tmod3^−/− platelets leading to GPIbα internalization⁴⁷  or impaired membrane biogenesis (see below in Figures 4 and 5). However, no changes were evident in the morphology of the circumferential microtubule coils of Tmod3^−/− platelets (Figure 2F, red arrows). (Tmod3^+/− embryos and platelets were similar to WT, due to increased Tmod3 expression in heterozygotes.³⁹ ) Therefore, Tmod3^−/− embryos exhibit decreased numbers of platelets with variable sizes and abnormal morphologies, resembling a macrothrombocytopenia.

To investigate whether reduced platelet numbers in Tmod3^−/− embryos are due to decreased MK number and/or impaired differentiation, we quantified CD41⁺ MKs in fetal liver cryosections. WT fetal livers contained sparsely distributed bright CD41⁺ MKs, as expected, whereas Tmod3^−/− fetal livers contained ∼1.5-fold more CD41⁺ MKs (Figure 3A-B). This may reflect increased MK progenitors, based on increased colony-forming unit (CFU)-MKs in Tmod3^−/− fetal livers (Figure 3C). We then examined the ploidy of isolated fetal liver MKs differentiated in vitro with mTPO (Figure 3D and supplemental Figure 2). As expected, the average ploidy of WT fetal liver MKs was close to ∼8N, which is lower than that of bone marrow MKs from adults (Figure 3D-E).^48-50  However, the relative numbers of Tmod3^−/− MKs with different ploidy were not significantly different, except for a small increase in 8N MKs (from 32% to 40%) (Figure 3D). Furthermore, there was no change in average ploidy (Figure 3E). Thus, alterations in MK numbers and ploidy are unlikely to account for decreased numbers of circulating platelets in Tmod3^−/− embryos. Increased MKs and progenitors in Tmod3^−/− fetal livers may reflect a compensatory response to the deficit in platelet production by Tmod3-deficient MKs (see below).

We next investigated whether the platelet deficits in Tmod3^−/− embryos were due to abnormalities in MK cytoplasmic morphogenesis. TEM of mature MKs in WT fetal liver sections reveals an extensive DMS surrounding electron-dense granules within cytoplasmic platelet territories (Figure 4A, left). In contrast, Tmod3^−/− MKs displayed two types of morphology, with about half resembling immature MKs containing large clumps of vesicles resembling a pre-DMS structure⁵¹  (Figure 4A, middle), and the other half containing relatively normal, elongated DMS tubules surrounding platelet territories (Figure 4A, right). In this latter type of Tmod3^−/− MK, the DMS often enclosed larger cytoplasmic regions containing fewer electron-dense granules (Figure 4A, right).

To further investigate DMS morphology and organelle distribution in MKs, we immunostained the DMS with anti-GPIbα^51,52  and mitochrondria with anti-TOM20, and imaged isolated MKs by confocal fluorescence microscopy (Figure 4B). Although all large-sized Tmod3^+/+ or Tmod3^+/− MKs with multilobular nuclei contained DMS membrane tubules organized into one or more rows of territories around the cell circumference, DMS tubule organization in Tmod3^−/− MKs was variable, forming a narrow band of territories around the circumference of some MKs (Figure 4B), but only a few territories in others (Figure 4C). Strikingly, mitochondria were clearly visible within each platelet territory in Tmod3^+/− MKs, but in Tmod3^−/− MKs only some territories contained mitochondria, and many mitochondria collected in cytoplasmic aggregates near the nucleus (Figure 4B-C). Further, in yet other Tmod3^−/− MKs, consistent with the TEM images (Figure 4A, middle), the DMS appeared as one or two large clumps of vesicles, resembling pre–DMS-like structures.⁵¹  Nonetheless, these Tmod3^−/− MKs were large with similar multilobular nuclear morphology as Tmod3^+/− MKs in which the DMS had formed tubules surrounding territories (supplemental Figure 3). Thus, although Tmod3^−/− MKs achieve high nuclear ploidy similar to Tmod3^+/+ and Tmod3^+/− (Figure 3C-D), cytoplasmic morphogenesis is impaired, with abnormal DMS formation that does not create normal platelet territories with appropriate distributions of organelles.

To test the ability of Tmod3^−/− MKs to produce proplatelets, we cultured fetal liver MKs in mTPO and observed the morphologies and numbers of MKs forming proplatelets.⁴³  Proplatelet buds formed from cultured Tmod3^−/− MKs contained microtubule rings but were more variable in size and often much larger than WT proplatelets (Figure 5A). Furthermore, Tmod3^−/− fetal livers contained about half as many proplatelet-forming MKs as WT fetal livers (Figure 5B), indicating that the proplatelet-forming ability of Tmod3^−/− MKs is impaired. For the subset of Tmod3^−/− MKs that did form proplatelets, proplatelet buds exhibited an approximately threefold increase in average area (Figure 5C), consistent with larger platelet sizes in embryonic blood (Figure 2E). The Tmod3^−/− MKs that do not form proplatelets may correspond to MKs with clumped DMS vesicles (Figure 4A, middle, and supplemental Figure 3).

Proplatelet buds formed from cultured Tmod3^−/− MKs had defective organelle composition, based on reduced staining for VWF, a marker of α-granules (Figure 5D-E). In proplatelet buds formed from Tmod3^+/+ MKs, the number of VWF⁺ puncta is linearly related to proplatelet bud area (Figure 5F), as shown previously.¹⁰  However, in the absence of Tmod3, there are 2 populations of proplatelet buds: one containing VWF⁺ puncta in proportion to bud area, similar to WT, and another with abnormally large proplatelet buds, containing variable numbers of VWF⁺ puncta with no relationship to size (Figure 5G). The Tmod3^−/− proplatelet buds with VWF⁺ content that scales with size may originate from regions of MK cytoplasm with relatively normal DMS tubule arrangements, whereas the abnormally large Tmod3^−/− proplatelet buds with variable VWF⁺ content may originate from DMS-poor regions of MK cytoplasm with uneven organelle distributions (Figure 4).

To investigate whether defective F-actin polymerization underlies abnormalities in Tmod3^−/− MKs and impaired proplatelet formation, we phalloidin-stained isolated fetal liver MKs. In control MKs, F-actin is associated with GPIbα-labeled DMS tubules and plasma membrane, with little cytoplasmic staining (Figure 4B-C). In contrast, in Tmod3^−/− MKs, brightly stained cytoplasmic aggregates of F-actin are often present (Figure 4B, red arrows), in addition to F-actin at the DMS tubules and plasma membrane (Figure 4B-C). In the most abnormal Tmod3^−/− MKs with no DMS-defined platelet territories, F-actin was present in the clumps of pre–DMS-like vesicles (supplemental Figure 3). Additionally, WT MKs extend protrusions and pseudopods with membrane-associated F-actin, whereas Tmod3^−/− MKs form rudimentary protrusions, with cytoplasmic rather than membrane-associated F-actin (supplemental Figure 5).

To directly determine whether F-actin polymerization might be defective in Tmod3^−/− MKs, we exploited the ability of MKs to assemble F-actin bundles upon spreading on collagen-coated coverslips. As expected, WT fetal liver-derived MKs assembled robust F-actin bundles as well as podosomes on their ventral surfaces (Figure 6A).^11,17  Tmod3-stained puncta were often observed associated with the F-actin bundles, suggesting the F-actin pointed ends are capped (Figure 6A, bottom panels, arrowheads). NMIIA also assembles along the F-actin bundles of WT MKs (Figure 6B), appearing in a striated pattern indicative of their contractile function.⁵³  In contrast, in Tmod3^−/− MKs spreading on collagen, F-actin polymerization with NMIIA into bundles is severely impaired (Figure 6C). Instead, F-actin forms amorphous aggregates near the periphery of the spreading MKs, and NMIIA does not assemble (Figure 6C). Thus, Tmod3 regulates F-actin polymerization in fetal liver-derived MKs.

Tmod3 binds TMs to cap and stabilize F-actin pointed ends,³³  with TM4 as the predominant TM in platelets.^42,54-56  Indeed, TM4 is detected in whole-cell lysates of in vitro cultured MKs from mouse fetal livers (supplemental Figure 4A-B), whereas long, muscle-type TMs are not detected (data not shown). Co-IP shows that MK Tmod3 and TM4 are associated with actin and associated cytoskeletal proteins (NMIIA, FLNA, and β2-spectrin) (supplemental Figure 4A-B). Confocal fluorescence microscopy of proplatelets further reveals that Tmod3 and TM4 partially colocalize with F-actin and are enriched at the cortex of proplatelet buds (supplemental Figure 4C-F), similar to cortical enrichment of F-actin in circulating embryonic platelets (supplemental Figure 1).

We hypothesized that absence of Tmod3 leading to disruption of Tmod3–TM4-actin associations might alter F-actin abundance and/or organization in Tmod3^−/− proplatelet buds. Tmod3^−/− proplatelet buds exhibited highly variable F-actin levels, with some similar to WT and others with up to approximately four- to fivefold more F-actin (Figure 7A-B). Moreover, similar to VWF⁺ puncta (Figure 5G), normal-sized Tmod3^−/− proplatelet buds had relatively normal F-actin levels, whereas larger buds had highly variable levels of F-actin (Figure 7C).

Next, we evaluated F-actin and TM4 distributions in proplatelet buds. Tmod3^+/+ proplatelet buds exhibited cortical enrichment of F-actin and TM4 (Figure 7D,G, supplemental Figure 4D,F, and supplemental Figure 6). However, in Tmod3^−/− proplatelet buds, the relative F-actin and TM4 intensity in the cytoplasm was increased as compared with the cortex, indicating TM4 and F-actin redistribution (Figure 7D,G and supplemental Figure 6). Quantification showed an approximate threefold increase for F-actin in the cytoplasmic vs cortical intensity ratio per proplatelet bud (Figure 7E), and an approximate twofold increase for TM4 (Figure 7H). Notably, the increased cytoplasm/cortex ratios of F-actin and TM4 were more pronounced in the abnormally large Tmod3^−/− proplatelet buds (Figure 7F,I). Thus, the deletion of Tmod3 disrupts TM4–F-actin cytoskeletal organization, with more pronounced effects in abnormally large proplatelet buds.

Abnormal F-actin organization was also evident in Tmod3^−/− circulating embryonic platelets, where foci and F-actin aggregates were present in the platelet interior, in contrast to Tmod3^+/+ embryonic platelets where F-actin concentrates at the cortex were present (supplemental Figure 1). We then examined F-actin organization in embryonic platelets spreading on collagen-coated coverslips. WT platelets formed peripheral lamellipodia with F-actin at the leading edge, as well as an internal F-actin ring and central concentration (Figure 7J). In contrast, Tmod3^−/− platelets formed lamellipodia at their periphery, but not the internal F-actin ring, and, in others, F-actin formed irregular bundles or partial rings (Figure 7J). Thus, Tmod3 regulates F-actin organization in proplatelets from cultured MKs and in embryonic circulating platelets.

We identified a new feature of Tmod3^−/− mouse embryos: hemorrhaging with macrothrombocytopenia. Tmod3 is the only Tmod in mouse fetal liver MKs and platelets, and is important for cytoplasmic morphogenesis of MKs. Tmod3^−/− MKs display an immature or incomplete DMS, which forms insufficient tubules for platelet territories and the membrane reservoir required for proplatelet protrusion, leading to impaired ability to produce proplatelets. Tmod3^−/− MKs may contain pre–DMS-like vesicular clumps,⁵¹  due to delay or inability to remodel the DMS, or contain reduced DMS tubules creating fewer platelet territories with irregular organelle distributions. Tmod3^−/− MKs produce 2 varieties of proplatelet buds: ones within a normal size range (∼2 to 20 μm²),¹⁰  and others that are abnormally large (up to ∼60 μm²). In normal-sized WT and Tmod3^−/− proplatelet buds, VWF⁺ granule content scales with proplatelet bud size, suggesting that Tmod3 regulation of F-actin is not required for organelle transport down proplatelet extensions, which depends on microtubule-dependent transport and sliding processes.^8,57  However, F-actin polymerization is aberrant in Tmod3^−/− MKs, and increased F-actin and redistribution of F-actin and TM4 from the cortex to the cytoplasm is only observed in abnormally large Tmod3^−/− proplatelet buds. Therefore, normal-size Tmod3^−/− proplatelet buds may originate from regions of MK cytoplasm containing DMS tubules, whereas abnormally large proplatelet buds may originate from DMS-poor regions of cytoplasm with irregular distributions of organelles and F-actin aggregates.

How might Tmod3 control MK DMS formation, organelle distribution, and proplatelet formation? The DMS is associated with F-actin (Figure 4B-C)²³  in a spectrin-actin network,²⁴  and spectrin tetramer formation and F-actin assembly are known to influence DMS formation and platelet biogenesis.^{14,23,24,58,59}  Spectrin is also implicated in intracellular vesicular transport,^60-64  suggesting Tmod3 may influence DMS and organelle distributions in MKs via stabilization of spectrin-actin networks associated with intracellular membranes. A precedent for direct association of Tmod3 with an intracellular membrane receptor is that of Tmod3 with sAnk1.5, a transmembrane protein in the sarcoplasmic reticulum of skeletal muscle.³⁶  Alternatively, Tmod3 may associate with MK and proplatelet membranes indirectly via F-actin’s links to spectrin networks or to FLNA, which binds GP1bα.^7,25 

Several lines of evidence in our study support the idea that impaired DMS formation and aberrant organelle distributions in Tmod3^−/− MKs are caused by a disrupted F-actin cytoskeleton. Tmod3^−/− MKs display aberrant cytoplasmic accumulations of F-actin (Figure 4B, supplemental Figures 3 and 5), and fail to polymerize F-actin during spreading on collagen (Figure 6C). Tmod3 is present in these F-actin bundles in WT MKs (Figure 6A), consistent with Tmod3 capping and promotion of F-actin assembly in the MK actin cytoskeleton. In vitro, Tmod3’s F-actin–capping activity is enhanced by Tmod3 binding TM, which binds along F-actin and stabilizes filaments.^26,30,33  TM4 (TPM4 gene) is the primary TM in platelets^42,54-56  and MKs (Figure 1 and supplemental Figure 4), and Tmod3 associates with TM4 and the actin cytoskeleton, which includes FLNA, β2-spectrin, and NMIIA (supplemental Figure 4). Since Tmod3 binds directly to TMs,^30,33  including TM4,⁶⁵  it is likely that Tmod3 caps TM4-coated F-actin linked to FLNA, β2-spectrin, and/or NMIIA in MKs and proplatelets.

Loss of Tmod3 leads to F-actin instability,^30,33,34  yet overall, more F-actin is observed in large Tmod3^−/− proplatelet buds (Figure 7B,C), and F-actin aggregates are observed in Tmod3^−/− MK cytoplasm (Figure 4B and supplemental Figure 5). What might explain this? Increased actin monomer levels resulting from disassembly of Tmod3-capped F-actin would be available for polymerization nucleated by other actin assembly factors present in MKs and proplatelets, such as WASp^17,58  or mDia1.²⁰  In other cell types, assembly of diverse F-actin structures are promoted by actin nucleation factors that compete for a limited pool of globular actin; thus when one assembly factor and pathway is eliminated, the others promote excessive F-actin assembly into alternative cytoskeletal structures.^66,67 

The actin cytoskeletal proteins that play important roles in MK platelet biogenesis (FLNA, ADF/n-cofilin, Rac/Cdc42, RhoA, and WASp) are also required for platelet spreading, adhesion, and granule secretion to form clots.^7,68,69  Indeed, Tmod3-null embryonic platelets spreading on collagen display defective F-actin polymerization and reorganization (Figure 7J). Although Tmod3-null platelets form lamellipodia, an actin–polymerization-dependent process, they do not form the internal F-actin rings observed during organelle centralization.^47,68  Moreover, abnormal numbers of VWF⁺ granules in Tmod3^−/− proplatelets (Figure 5), and fewer granules and more empty vesicular structures observed by TEM in circulating platelets in Tmod3^−/− embryos (Figure 2) support the idea that Tmod3^−/− platelet function is likely defective. Future studies with a MK-restricted Tmod3 deletion in adult mice will be required to define the precise role of Tmod3 in platelet function in thrombosis and hemostasis.

The authors thank Malcolm Wood in the The Scripps Research Institute Core Microscopy Facility for TEM of MKs and platelets, Zaverio Ruggeri and Alessandro Zarpellon for anti-GPIbα antibodies and discussions, and David Gokhin for help with editing.

This study was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (R01-HL083464) (V.M.F), and National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK094934 and R01 DK086267) (D.S.K.), and funding from the State of Connecticut Stem Cell Research Fund (D.S.K).

Contribution: Z.S. analyzed protein expression, performed MK cultures, confocal microscopy and quantitative image analyses, prepared figures, and wrote portions of the manuscript; R.B.N. bred mice, performed embryo analysis, histology and confocal microscopy, and prepared the figures; C.S. and S.H. performed fetal liver dissections for flow cytometry, analyzed data, and prepared the figures; D.S.K. provided intellectual input, experimental design, data analysis, and read the manuscript; and V.M.F. conceived the project, designed experiments, interpreted data, and wrote the manuscript.

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

Correspondence: Velia M. Fowler, Department of Cell and Molecular Biology, CB163, The Scripps Research Institute, 10550 N. Torrey Pines Rd, La Jolla, CA 92037; e-mail: velia@scripps.edu.