Nbeal2 interacts with Dock7, Sec16a, and Vac14

Cellular secretion is a key biological event involved in many physiological processes, including hemostasis.^1,2  Secretory cells store granules, which are released according to predetermined temporal and spatial programs.³  The function of secretory cells is to a large extent linked to their types of granules, their content, and the signaling events leading to secretion. Therefore, the ontology and homeostasis of granules must be highly regulated (ie, from the formation of the vesicles themselves and cargo sorting through cytoplasmic maintenance of formed vesicles and eventual exocytosis).^4,5 

Blood platelets are anucleated secretory cells originating from megakaryocytes (MKs).⁶  Their primary role is in hemostasis,⁷  but they are also relevant in other pathophysiological pathways.^8-12  Platelet function is mediated by the release of their 3 types of granules upon vascular injury.¹³  Aberrations in any of these 3 granule types lead to inability of platelets to sustain hemostasis, resulting in bleeding.¹⁴  Defects in lysosomes are observed in Chediak-Higashi syndrome and caused by mutations in LYST encoding a member of the family of Beige and Chediak-Higashi (BEACH)-domain (BEACH hereafter) containing proteins.¹⁵  Aberrations of platelet δ-granules are observed in Hermansky-Pudlak syndrome caused by mutations in 11 genes encoding proteins with roles in the formation and trafficking of granules.¹⁶  Haploinsufficiency of another BEACH family member, Neurobeachin (NBEA), is associated with severe autism and atypical δ-granules in humans¹⁷  and Nbea heterozygous mice.¹⁸ 

The α-granules are the most abundant organelles and granules in platelets. They contain hundreds of proteins with important functions in hemostasis and wound healing.^8-12,19  Loss of function variants of the BEACH-domain gene NBEAL2 in gray platelet syndrome (GPS) cases leads to absence of platelet α-granules.^20-23 Nbeal2^−/− mice present with a faithful copy of human GPS with the typical large platelets of grayish appearance, and interestingly, neutrophils and mast cells are also dysfunctional.^24,25  In vivo studies highlight the role of Nbeal2 beyond hemostasis, particularly in inflammation, innate immunity, and cancer.^24,26-28  However, the molecular function of Nbeal2 protein has not been explored.

The BEACH proteins Lyst, Nbea, and Nbeal2 are deemed essential for molecular processes underlying the formation, retention, and secretion of the 3 types of granules in platelets.²⁹  All 9 BEACH genes encode large proteins between 2000 and 4500 residues containing the Armadillo-type, Concanavalin-A lectin, Pleckstrin Homology (PH), and WD40 repeat domains in addition to the typical BEACH domain.²⁹  This domain of ∼35 kDa is unique to the BEACH family, is highly conserved in eukaryotes, and does not share sequence homology with other proteins.¹⁵  It is characterized by a unique hydrophobic secondary structure that cannot be classified as either β-strands or random coils³⁰  and strongly interacts with its neighboring PH domain, but the significance of this interaction remains elusive.³¹ 

Here, we use expression of Nbeal2 domains in human cells to define its proximal interactome with confirmation of Dock7, Sec16a, and Vac14 as Nbeal2 binding partners by different biochemical and cellular approaches. We show that GPS-causing mutations strongly reduce the interaction of Dock7 with the BEACH domain, and platelets from Nbeal2^−/− mice and GPS patients are nearly devoid of Dock7 with a direct impact on platelet function.

Short descriptions of the methods used are provided in this section, and complete details are given in the supplemental Methods, available on the Blood Web site.

Human embryonic kidney 293T cells (HEKs) and CHRF-288 cells (CHRFs) were cultured following AATC procedures. Stable transfectants ectopically expressing the protein domains under study were generated by infection with lentiviral particles. HEKs were transduced at a multiplicity of infection of 10 transducing units per cell in the presence of 1 μg/mL polybrene. CHRFs were transduced at a multiplicity of infection of 50 transducing units per cell in the presence of 1 μg/mL protamine sulfate.

Cells were collected and washed twice with cold phosphate-buffered saline (PBS), and pellets were snap-frozen on dry ice and kept at −80°C until immunoprecipitation. Pellets were resuspended with lysis buffer containing 0.1% NP-40 and protease inhibitor cocktail. Tandem affinity purification and sample preparation for mass spectrometry analysis were performed as previously described.³² 

Peptides were redissolved in 0.5% formic acid and analyzed with on-line nano liquid chromatography tandem mass spectrometry on an LTQ FT mass spectrometer (Thermo-Fisher Scientific) coupled with an Ultimate 3000 Nano/Capillary LC System (Dionex). Proteins frequently identified from tandem affinity purifications of unrelated targets in embryonic stem cells were discarded.

The mass spectrometry proteomics data can be retrieved from the ProteomeXchange Consortium via the PRIDE³³  partner repository with the dataset identifier PXD006091.

Total cell lysates were generated as above, and proteins of interest were immunoprecipitated using Protein G Dynabeads (Invitrogen) and selected antibodies.

Human MKs were generated from CD34⁺ hematopoietic stem cells as stated in supplemental Methods. Proximity ligation assays (PLAs) were performed following the manufacturer’s protocol.

Adult animals were used, including age-matched controls of the same genetic background. This study has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (Home Office PPL70/8406). A minimum of 5 different mice per experiment was used unless otherwise stated.

Mouse blood was withdrawn from the inferior vena cava in either EDTA (5 mM final) or ACD (acid-citrate-dextrose; 111 mM glucose, 71 mM citric acid, 116 mM sodium citrate), whereas blood from GPS patients and controls was taken as previously reported.²¹  In both cases, platelets were washed using Tyrode’s buffer pH 6.5.

HEKs, CHRFs, and platelets were lysed by addition of FLAG tandem affinity purification (FTAP)-lysis buffer (0.1% NP-40, 150 mM NaCl, 10 mM Tris, pH 8) containing protease inhibitor cocktail and incubated on ice for 10 minutes before centrifugation at 13 000 rpm for 15 minutes at 4°C.

Platelets were washed as above and adjusted to 3 × 10⁵ platelets per microliter in Tyrode’s buffer pH 7.35. Stimulations with vehicle (PBS) and thrombin (T4648; Sigma, UK) were done at room temperature for different time points, and lysis buffer provided by assay kit (BK127 and BK128, Cytoskeleton) was added. Subsequent steps were carried out following the manufacturer’s protocol. Absorbance at 490 nm was read using an MP5 plate reader (Molecular Devices, UK).

Washed platelets were activated for 10 minutes and then fixed, permeabilized, incubated with 488-Phalloidin for 20 minutes, and read in a Gallios flow cytometer (Beckman Coulter, UK).

A solution of 100 μg/mL of fibrinogen (FIB3; Enzyme Research, UK) in PBS was used to coat coverslips (12392128; Thermo-Fisher Scientific) overnight at 4°C. Platelets at 5 × 10³/μL in Tyrode’s buffer pH 7.35 were seeded onto the coverslips, and platelet agonists or vehicle was added.

Cells were harvested, washed in physiological saline solution, and fixed in 2% glutaraldehyde and 2% formaldehyde in 0.05 M cacodylate solution, pH 7.4 for 4 to 6 hours at 4°C.²⁷  Platelets were labeled as previously reported.³⁴ 

Cells grown on coverslips were fixed with 2% paraformaldehyde-PBS for 30 minutes at room temperature and washed twice with PBS.

Results are shown as mean ± standard error of the mean (SEM). Statistical analysis was performed using the unpaired Student t test. P values < .05 were considered statistically significant.

A complete list of primary antibodies used in this study is provided in supplemental Table 1.

To identify the putative interacting proteins of Nbeal2 (the interactome hereafter), a construct encompassing the PH, BEACH, and WD40 (PBW) domains and a FTAP tag³²  (Figure 1A) was used to transform HEKs (Figure 1B), which endogenously express Nbeal2 (supplemental Figure 1A). An FTAP-tagged green fluorescent protein (GFP) construct was used as a control (Figure 1A-B). Cells were then used in a 2-step (FLAG and calmodulin binding peptide) sequential affinity purification, and the resulting eluate was separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) followed by mass spectrometry analysis (Figure 1C). Proteins identified in GFP samples and from an “in-house contaminant” database as stated in the “Materials and methods” were removed from further analysis. The classification of the remaining 132 proteins (129 entries by gene name excluding NBEAL2; supplemental Table 2) according to the most frequent gene ontology terms shows an array of biological functions (Figure 1D).

To better comprehend its functional role, we used the interaction data to embed Nbeal2 in a protein-protein interaction network (PPIN) representing the processes of hemostasis, megakaryopoiesis, and platelet formation.³⁵  The PPIN of 5112 nodes (proteins) and 27 441 edges (biochemical interactions) was obtained by retrieving first-order interactions from Reactome³⁶  and IntAct³⁷  using a set of 637 bait proteins (baits hereafter). The set of genes encoding the baits resulted from an integrative analysis of a genome-wide association study for platelet traits in 173 480 individuals^38,39  and a promoter capture Hi-C assay in human MKs⁴⁰ ; the 79 genes implicated in Mendelian disorders of hemostasis were also included.^14,41  Of the 129 nodes in the Nbeal2 interactome, 91 are present in the PPIN (orange nodes in Figure 1E) and the remaining 38 are first-order interactors with 1 or several proteins of the PPIN (purple nodes in Figure 1E).

Previous studies on proteins from the BEACH family suggest that they may function as scaffolds with a critical role of Nbeal2 in homeostasis of α-granules.^26,27,29  Based on this and on a detailed inspection of the PPIN, 10 candidate proteins were selected for validation studies by reverse immunoprecipitation (supplemental Table 2), and the interactions of 3 proteins were confirmed, namely Dock7 (a guanidine exchange factor [GEF]); Sec16a (an endoplasmic reticulum membrane protein); and Vac14 (a regulator of phosphatidyl inositol 3,5-bisphosphate) (triangular orange nodes in Figure 1E). Their first-order interactors are illustrated in the subnetworks in Figure 1F-H (supplemental Figures 2-4 for fully labeled versions).

To confirm the interactions with Nbeal2, we first performed immunoprecipitation of the selected candidates in PBW-FTAP HEKs, and positive results for Dock7, Sec16a, and Vac14 were obtained (Figure 2A- B). These experiments showed that endogenous Sec16a and Vac14 pulled down PBW-FTAP and Dock7, and endogenous Dock7 pulled down PBW-FTAP and Vac14 (Figure 2B; supplemental Figure 5). Assays in GFP-FTAP HEKs, as expected, showed no interactions with any of the 3 proteins (Figure 2C). Second, we used CHRFs, which have a transcriptome reminiscent of MKs⁴²  and a robust Nbeal2 protein expression (supplemental Figure 1B). Lentiviral transduction with PBW-FTAP followed by the same experimental approach as used for HEKs confirmed the interaction with Dock7, Sec16a, and Vac14 in PBW-FTAP CHRFs (Figure 2D). Furthermore, endogenous Nbeal2 was also detected when the 3 interactors were pulled down in nontransduced CHRFs, although the recovery levels for Dock7 and Sec16a were low (Figure 2E). Finally, after the initial successful validation experiments with cell lines, we used PLA in MKs derived from human hematopoietic stem cells.⁴³  The 2 most abundant platelet membrane receptor complexes for fibrinogen (αIIbβ3 integrin) and von Willebrand factor (GPIbα, GPIbβ, GPV, GPIX), respectively, were used to validate the assay. As expected, targeting αIIb and GPIX simultaneously produced a weak PLA signal similar to that using isotype controls, while targeting GPIbα and GPIX generated a strong signal (Figure 2F). When compared with the negative control, we observed a significant increase in PLA signals in Dock7-Nbeal2 and Vac14-Nbeal2 labeled-MKs, further supporting the interactions between these proteins (Figure 2F).

The BEACH domain intimately interacts with its adjacent aminoterminal PH domain over an interface of 1300 Ų, suggesting they may fold as a single functional unit.^30,31  We therefore overexpressed a construct comprising both the PH and the BEACH domains (PB) in HEKs (Figure 3A). Dock7 and Vac14 were present in PB-FTAP and PBW-FTAP pull downs, whereas Sec16a was recovered only with PBW-FTAP (Figure 3B). Hence, the binding of Sec16a to Nbeal2 may be mediated through the WD40 domain, whereas the other 2 proteins bind to either the PH or the BEACH domain.

To determine the possible effects of GPS-causing mutations, we cataloged all reported missense variants (n = 22)^21,23,44,45  and those identified “in-house” by the ThromboGenomics⁴¹  and National Institute for Health Research (NIHR) BioResource sequencing projects (n = 14; Kate Downes, Kathleen Freson, and Keith Gomez, unpublished data [see acknowledgement]). This showed a significant fivefold enrichment (Fisher test, P value = 3.64 × 10⁻⁵) of variants in the BEACH domain, which harbored 13 of 36 of the observed variants (36%), whereas the domain only accounts for 292 of the 2754 (11%) amino acids of Nbeal2. We used the coordinates of the crystal structure of the BEACH domain of Lrba (pdb 1T77) to estimate the functional consequences of the 13 mutations and selected 6 with different types of structural changes to determine their effects on the binding of Dock7 and Vac14 (Figure 3C-D). HEKs were stably transfected with PBW-FTAP constructs and the reference, and 5 variant-carrying baits were expressed at similar levels, indicating a minimal, if any, detrimental effect of each variant on the folding of the domains (Figure 3E). Our results showed that M2080K, P2100L, and G2290W/E2294G (both on the same haplotype) seemed to diminish the binding of Vac14 (Figure 3E); P2100L and R2172H significantly reduced Dock7 binding, and its binding to the G2290W/E2294G bait seemed to be at a reduced level, although the difference with the reference bait was not significant (Figure 3F).

Currently, the subcellular localization of Nbeal2 remains unknown. We investigated this using PBW-FTAP CHRFs by labeling its FLAG epitope. Transmission electron microscopy revealed the presence of intracellular vesicles, mitochondria, and Golgi in CHRFs (Figure 4A-B). Labeling of FLAG by gold-adsorbed immunoglobulins did not show any staining in CHRFs (not shown). However, clusters of gold were observed across the cytoplasm of PBW-FTAP CHRFs, suggesting a cytosolic localization of PBW (Figure 4CF). This pattern was also confirmed by confocal microscopy by labeling PBW and F-actin with anti-FLAG and Phalloidin, respectively (Figure 4G; supplemental Figure 6). The use of a rabbit antiserum revealed the presence of Nbeal2 in the cytoplasm (arrowheads), in close proximity to enclosing membranes of the α-granules (arrows), the open canalicular system (OCS) (circles), and the cell surface (flat ellipses) of human platelets (Figure 4H).

We further investigated the localization of Nbeal2 and its interactors Dock7, Sec16a, and Vac14 by subcellular fractionation of platelets. We first fractionated human platelets and identified cytoskeleton and dense tubular system in fractions 1 to 4 (enriched for tubulin, actin, and calreticulin) and α-granules in fraction 7 and 8 (enriched for the α-granule protein Thrombospondin-1 [Thbs1]) (Figure 5A). Nbeal2 was mostly found in fractions 2 to 4 containing cytoskeletal and dense tubular system proteins (Figure 5A). Fractionation of murine platelets showed a similar pattern (ie, Nbeal2 was primarily found in fractions 3-4) (Figure 5B). As expected, proteins typically stored in α-granules, Thbs1 and vascular endothelial growth factor, are significantly reduced in GPS platelets (Figure 5B). Dock7 was present in the α-granule-containing fraction (7-8) of control platelets but was mislocalized in fraction 4 in Nbeal2^−/− platelets (Figure 5C). Although at lower levels, Dock7 and Nbeal2 colocalize in fractions 3 to 4 and 7 to 8, respectively, providing further evidence of their interaction (supplemental Figure 7). The localization of Sec16a and Vac14 did not differ significantly between platelets from both groups (Figure 5D-E), despite potential apparent increased levels of Sec16a in Nbeal2^−/− platelets (Figure 5D), an observation that was not replicated in whole platelet lysates (Figure 5F-G).

Having identified the impact of the absence of Nbeal2 on the subcellular localization of Dock7, we then asked whether it had an effect on the overall cellular abundance of the 3 interactors. Similar levels of Sec16a and Vac14 were confirmed in platelets from both groups, but the Dock7 protein level was significantly reduced in Nbeal2^−/− platelets (Figure 5F-G; supplemental Figure 8), and importantly, also in the platelets of 2 GPS cases (Pedigree C of the previous report²¹ ) (Figure 5H). Transcript levels of Dock7 in mouse MKs did not differ significantly between control and Nbeal2^−/− mice (Guerrero et al²⁷  and supplemental Table 3), and IBs with anti-Dock7 did not reveal signs of proteolysis or higher Dock7 levels in plasma (not shown), suggesting that the Nbeal2 protein is likely involved in its posttranslational modification and/or stability.

As a GEF, Dock7 exchanges GDP for GTP on Cdc42 and Rac1, causing their activation, followed by activation of downstream effectors, including the dephosphorylation (activation) of cofilin, a key regulator of actin turnover.^46,47  Actin reorganization then leads to extension of filopodia and generation of lamellipodia,⁴⁶  ultimately resulting in platelet shape change. Given the significant reduction of Dock7 levels and its altered localization in Nbeal2^−/− platelets, we postulated that this canonical signaling pathway may be disrupted and set out to test this using control and Nbeal2^−/− platelets.

First, the phosphorylated (inactive) form of cofilin, P-cofilin, was significantly increased in murine Nbeal2^−/− platelets (Figure 6A-B). Other regulators of this signaling pathway such as Cdc42, Rac1, Iqgap1 and chronophin^46,48  were also measured, and we found that the level of Iqgap1, a protein that stabilizes the active forms of Cdc42 and Rac1, was also significantly higher in Nbeal2^−/− platelets (Figure 6A-B), suggesting that the increase in P-cofilin was likely due to an imbalance not only of Dock7 but also of Iqgap1. In fact, our previous gene expression array of control and Nbeal2^−/− MKs revealed a significantly increased Iqgap1 transcript level in the latter (Guerrero et al ²⁷  and supplemental Table 4), which is validated at the protein level in platelets in this study.

We then asked if the disequilibrium in the Dock7/Iqgap1/P-cofilin pathway had an effect on the function of platelets. First, levels of active Rac1 at 2 minutes after activation by thrombin were significantly higher (P value = .026) in platelets from control vs Nbeal2^−/− mice, and a similar trend was observed for Cdc42, but this did not reach significance (Figure 6C-D). Second, we measured F-actin formation in platelets. Despite higher levels of F-actin in basal Nbeal2^−/− platelets due to their larger size (Figure 6E; P value = .048), induction of actin polymerization by thrombin revealed a lower increase in Nbeal2^−/− platelets (Figure 6E-F; P value = .0005). Last, we checked the endpoint of this pathway (ie, platelet shape change) by measuring the ability of platelets to spread onto fibrinogen after activation with both adenosine 5′-diphosphate and thrombin. Enumeration of activated vs nonactivated platelets assessed by shape change and spreading revealed that nearly all control platelets became activated, whereas the Nbeal2^−/− platelets showed significantly reduced spreading (Figure 6G-H; P value ≤ .0005). Altogether, these results are compatible with the notion that Nbeal2 plays an important role in the regulation of cytoskeletal rearrangements in platelets, most likely through its interaction with Dock7.

Despite the severity of inherited granule pathologies caused by mutations in NBEAL2, NBEA, and LYST, the key roles of the encoded BEACH proteins in the ontogeny of α-, δ-, and lysosomal granules remain poorly defined. Here, we report the results of studies aimed at defining the role of Nbeal2 at the biochemical and cellular level.

We identified the putative interactors of Nbeal2 in human cells and selected 10 candidate interactors for validation. The interactions of Dock7, Sec16a, and Vac14 with Nbeal2 were confirmed, and the physical proximity of the former 2 to Nbeal2 in human MKs was corroborated by PLA. Furthermore, the introduction of GPS-causing mutations in the BEACH domain showed profound effects on the interactions with Dock7. Although Nbeal2 is mostly present in the platelet cytoplasm, Dock7 localizes to the α-granules, and, interestingly, murine and human platelets devoid of functional Nbeal2 showed strongly reduced Dock7 levels, resulting in changes in its canonical signaling pathway.

The fivefold enrichment of GPS-causing variants in the BEACH domain prompted us to perform the pull-downs with a construct containing the BEACH and its PH and WD40 flanking domains. Despite the potential presence of false-positive interactors frequently seen in proteomics assays, 70% of the 129 proteins in Nbeal2’s interactome are present as primary nodes in the PPIN, which represents a lion’s share of the proteome of human MKs. So far, no similar interaction data are available for any of the BEACH proteins, except for the 21 proteins possibly interacting with Lyst.⁴⁹ 

The observation that Nbeal2 localizes in the vicinity of the membrane, OCS, and α-granules of platelets is compatible with its assumed role in membrane dynamics, and especially α-granule trafficking. Our results are supportive of the notion that the GEF Dock7 may be a critical intermediary in this membrane dynamics, which is also essential in the formation of platelets by MKs.⁴⁶  Therefore, the discovery that Dock7 is nearly depleted from platelets of GPS cases and Nbeal2^−/− mice provides a plausible explanation as to why these platelets have an increased volume. In addition, our genome-wide association study identified variants associated with platelet traits in 9 genes encoding proteins in the Nbeal2 interactome (Nbeal2 itself, Dock7, and the 7 circular orange nodes with appended gene names in Figure 1E). The variant rs60286666 in DOCK7 is associated with the distribution width of platelet volume (P value = 8.59 × 10⁻³³), providing further evidence of this Nbeal2 interactor possibly being important in the formation of platelets by MKs.³⁹  The role of Dock7 in platelet function has not been studied, but the misty mouse, with a truncation in Dock7, has a prolonged bleeding time possibly explained by a reduced ability of platelets to support hemostasis.⁵⁰  The relevance of the canonical Rac1/Cdc42/Dock7 signaling pathway in megakaryopoiesis and platelet function has however been extensively studied. The conditional deletion of cofilin⁵¹  and simultaneous deficiency of Rac1 and Cdc42⁵²  result in decreased platelet count, larger platelets, and impairment of their function, including defective shape change. Moreover, dysregulation of cofilin was also observed in a mouse knock-in and a family with type 2B von Willebrand disease, both presenting with macrothrombocytopenia.⁵³  Similar to these reports, we show that several elements of this signaling pathway (ie, P-cofilin and Iqgap1, a stabilizer of the active form of Cdc42 and Rac1) are dysregulated in murine Nbeal2^−/− platelets, probably as a consequence of the reduced levels of Dock7. Hence, GTP-Rac1 levels are diminished in thrombin-stimulated Nbeal2^−/− platelets, which show reduced actin polymerization and are unable to change shape efficiently (Figure 6D-H). Indeed, taking all the above evidence together, it is reasonable to assume that the macrothrombocytopenia in Nbeal2^−/− mice and in GPS cases is in part explained by the specific loss of Dock7.

Dock7 interacts with the PH-BEACH domain, and this interaction seems to be diminished by all but 1 of the 6 GPS-causing mutations surveyed (Figure 3F). It has been suggested that Dock7 or 1 of its binding partners is localized in or on the surface of α-granules, and its PIP2/PIP3 binding domain makes the latter more likely.⁵⁴  Our results are supportive of this assumption, and we reason that the interaction of Dock7 and α-granules is mediated through the BEACH domain.

Sec16a defines the budding sites on the endoplasmic reticulum membrane for trafficking of vesicles through the Golgi,⁵⁵  and disruption of its binding to mutant WD40 domains (Figure 3B) could potentially affect platelet homeostasis. Similarly, changes in the function of Vac14, which in complex with Fig4 and Pikfyve (Figure 1H) regulates the levels of PI(3-5)P2, may alter endosomal trafficking and signaling.⁵⁶ 

In summary, from 129 binding partners of Nbeal2 identified by mass spectrometry, we have confirmed the interaction of 3, Dock7, Sec16a, and Vac14, by different biochemical and cellular approaches and have also studied the functional relevance of 1, Dock7, in platelet biology and its involvement in the abnormal formation of platelets by MKs in GPS.

The authors thank Kate Downes (University of Cambridge, UK), Kathleen Freson (University of Leuven, Belgium), and Keith Gomez (University College London, UK) for reviewing the sequencing results.

Research in the Ouwehand laboratory is supported by grants from the European Commission (BLUEPRINT grant code HEALTH-F5-2011-282510, HORIZON 2020 grant code 692041), NIHR (for funding of the Cambridge Biomedical Research Centre, grant RG64219), Medical Research Council grant WT098503, Bristol Myers-Squibb, BHF grants RP-PG-0310-1002 and RE/13/6/30180 for the BHF Cambridge Centre of Excellence, and National Health Service Blood and Transplant (NHSBT). The sequencing of the cases with GPS was supported by the NIHR BioResource grant RBAG163. T.K.B. is the recipient of a Clinical Research Training Fellowship award from the British Society for Haematology and NHSBT, J. Collins is a Medical Research Council Clinical Training Research fellow, and W.H.O. is an NIHR Senior Investigator.

Contribution: L.M., M.J., S.A.d.H., and J.A.G. designed and performed experiments and analyzed data; M.P., L.Y., and J. Choudhary carried out proteomics analysis; J. Collins, T.K.B., L.G., R.P., S.M., P.A.S., and W.J.A. analyzed data; P.N. performed EM of platelets; R.F. provided GPS samples; W.W.Y. carried out structural analysis; and W.H.O. and J.A.G. wrote the manuscript.

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

The current affiliation for M.P., L.Y., and J.C. is The Institute of Cancer Research, Cancer Biology Division, Chester Beatty Laboratories, London, United Kingdom.

The current affiliation for S.M. is European Molecular Biology Laboratory European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, United Kingdom.

Correspondence: Jose A. Guerrero, Department of Haematology, University of Cambridge, Long Road, CB2 0PT, Cambridge, UK; e-mail: jg652@medschl.cam.ac.uk.