Loss of the Arp2/3 complex component ARPC1B causes platelet abnormalities and predisposes to inflammatory disease

Human actin-related protein 2/3 complex (Arp2/3), required for actin filament branching, has two ARPC1 component isoforms, with ARPC1B prominently expressed in blood cells. Here we show in a child with microthrombocytopenia, eosinophilia and inflammatory disease, a homozygous frameshift mutation in ARPC1B (p.Val91Trpfs*30). Platelet lysates reveal no ARPC1B protein and greatly reduced Arp2/3 complex. Missense ARPC1B mutations are identified in an unrelated patient with similar symptoms and ARPC1B deficiency. ARPC1B-deficient platelets are microthrombocytes similar to those seen in Wiskott–Aldrich syndrome that show aberrant spreading consistent with loss of Arp2/3 function. Knockout of ARPC1B in megakaryocytic cells results in decreased proplatelet formation, and as observed in platelets from patients, increased ARPC1A expression. Thus loss of ARPC1B produces a unique set of platelet abnormalities, and is associated with haematopoietic/immune symptoms affecting cell lineages where this isoform predominates. In agreement with recent experimental studies, our findings suggest that ARPC1 isoforms are not functionally interchangeable.

I n eukaryotic cells, the monomeric ATP-binding protein globular actin (G-actin) is assembled into dynamic filamentous actin (F-actin) in cytoskeletal structures. This process involves a variety of actin-binding proteins 1,2 that sequester/ deliver actin monomers and facilitate the nucleation, elongation, capping, severing, depolymerization and crosslinking of F-actin 3 . Disruption of these processes can result in dysregulation of the actin cytoskeleton, which is associated with metastatic cancer, autoimmune disorders and congenital defects 4 . In Wiskott-Aldrich syndrome (WAS) and X-linked thrombocytopenia (XLT), mutations in WAS (encoding WASP) cause microthrombocytopenia (that is, reduced numbers and size of blood platelets), immunodeficiency, eczema, increased malignancies and autoimmune symptoms including vasculitis and inflammatory bowel disease [5][6][7][8] . WASP is one of several nucleation promoting factors that can promote branching of F-actin via the actin-related protein 2/3 complex (Arp2/3), which plays a central role in cell migration, endocytosis, vesicular trafficking and cytokinesis 3,4,9 . Cytoplasmic WASP has an auto-inhibited conformation that is activated due to phosphorylation of tyrosine 291 (refs 10,11) by the Rho family GTPases CDC42 (cell division cycle 42) and NCK1 (non-catalytic region of tyrosine kinase 1). Activated WASP binds Arp2/3 with 2:1 stoichiometry by interacting with the ARP2, ARP3 and ARPC1 subunits, inducing conformational changes that facilitate binding of actin monomers and daughter filament growth 12,13 .
The mammalian Arp2/3 complex contains five unique components: ARP2, ARP3, ARPC2, ARPC3 and ARPC4, together with one molecule each of the isoform pairs ARPC1A and ARPC1B, and ARPC5A and ARPC5B. ARPC1A and ARPC1B are located in tandem on human Ch.7 (ref. 14). The isoforms they encode have 68% amino acid sequence identity 15 and both have six WD40 domain repeats predicted to form a b-propellerfold 3, 16 . ARPC1A has been implicated as a regulator of cell migration and invasion in pancreatic cancer 14 , while increased ARPC1B has been linked to oral squamous cell carcinoma 17 . Independent of its function in Arp2/3, ARPC1B has also been identified as a centrosomal protein involved in mitosis 18,19 . Total loss of Arp2/3 function is embryonic lethal 20 , while its inhibition within cells blocks lamellipodia formation and migration 21,22 . To our knowledge no inherited human deficiency of ARPC1B or other Arp2/3 components has been previously reported.
Here we describe three individuals from two families with homozygous mutations in ARPC1B. Patient 1 with p.Val91Trpfs*30 ARPC1B mutation that results in complete loss of ARPC1B protein and decreased Arp2/3 complex in platelets, leading to microthrombocytopenia, decreased platelet dense granules, defective platelet spreading with prominent filopodia but limited lammellipodia. The patient also has had repeated invasive infections, inflammatory bowel disease, leukocytoclastic vasculitis and eosinophilia. Patients 2 and 3 with a p.Ala105Val ARPC1B mutation that results in minimal ARPC1B protein in their platelets that are microthrombocytes with dense granule deficiencies and similar spreading abnormalities. Patient 2 has leukocytoclastic vasculitis and Patient 3 had intermittent eczematous-like rashes and both have eosinophilia. Thus ARPC1B deficiency in humans results in defective Arp2/3 actin filament branching that is associated with multisystem disease including platelet abnormalities, cutaneous vasculitis, eosinophilia and predisposition to inflammatory diseases.

Results
Identification of ARPC1B-deficient patients. We investigated two independent families where patients presented early in life with failure to thrive, platelet abnormalities, eosinophilia, small vessel vasculitis, eczema and other indicators of inflammatory/ immune disease (Fig. 1a- Fig. 1e,f). This 2 base pair duplication produces a frame shift in exon 4, resulting in a premature stop codon predicted to truncate ARPC1B at amino acid 119. This would yield a product lacking five of the six WD40 repeat domains (Fig. 1g) that form a b-propeller required for interaction of ARPC1B with mother actin filaments 16 . WES of Patient 2 and his parents (Supplementary Table 1, Supplementary Data, Fig. 1f,g) identified homozygosity for two missense ARPC1B variants (c.314C4T and c.712G4A encoding p.Ala105Val and p.Ala238Thr) in this patient and a sibling (Patient 3). Both variants affect WD40 domains; c.314C4T is novel and predicted to be deleterious while c.712G4A is predicted to be neutral (Supplementary Table 1). Patient 1's parents are consanguineous; thus we focused our initial genetic analysis on Mendelian autosomal recessive mutations. As shown in Supplementary Data, there were no homozygous mutations in any known genes associated with primary immune deficiency, including WASP and WASP interacting protein (WIP), or platelet disorders that could explain the complex phenotype observed in Patient 1. Furthermore, we identified no overlapping compound heterozygous, X-linked, or de novo mutations shared between Patients 1 and 2 ( Supplementary Fig. 1). We then focused on novel genes and examined known biological function, known diseases associated with genes, gene expression profiles and available animal models of the candidates described in the Methods section and Supplementary Fig. 1. The critical role of ARPC1B in Arp2/3 function made ARPC1B a viable candidate for the WAS-like disease phenotype observed in both patients, and ARPC1B was the only mutated gene they had in common (see Methods, Identification of ARPC1B mutations). Immunofluorescence (IF) microscopy of skin and intestinal biopsies ( Supplementary Fig. 2) indicated loss/reduction of ARPC1B expression in patients. Expression of ARPC1B and other Arp2/3 components in these patients was further studied in a readily accessible cell source provided by blood platelets.
ARPC1B expression and Arp2/3 complex in platelets. Expression of Arp2/3 components (Fig. 2a) was examined in normal, unaffected family members and patient platelet lysates via immunoblot analysis (Fig. 2b, Supplementary Fig. 3a-c), which showed the absence of ARPC1B in Patient 1, and greatly reduced levels in Patients 2 and 3. Platelet levels of other Arp2/3 components (ARP2, ARP3, ARPC2, ARPC3 and ARPC5; Fig. 2b) and WASP (Fig. 2c) were normal in all patients, and also in unaffected family members ( Supplementary Fig. 3c). Recent proteomics data indicate that ARPC1A is present at o6% of the level of ARPC1B in normal platelets 23 (Supplementary Table 2), with isoform levels and ratios differing considerably among cell/ tissue types (Supplementary Table 3). We observed low expression of ARPC1A relative to ARPC1B in normal platelets, and increased ARPC1A in ARPC1B-deficient cells (Fig. 2b,c). Since platelets lack nuclei, this indicates upregulation of ARPC1A in ARPC1B-deficient platelet precursor megakaryocytes (MKs). This parallels the reported upregulation of ARPC1B when siRNA was used to suppress expression of ARPC1A in HeLa cells 24 . The same study also reported, as we saw, that there was no effect of loss of ARPC1B on the expression of other Arp2/3 components.
Morphology of ARPC1B-deficient platelets. Abnormally small platelets (microthrombocytes) in humans have previously only been observed in patients with WAS with complete loss of WASP expression, or in patients with XLT where WASP expression is impaired but not abolished 7,8 . While both conditions are accompanied by thrombocytopenia and bleeding, XLT is associated with minimal or no immunodeficiency 5 . Features of WAS including thrombocytopenia, recurrent infections and eczema were also observed in a patient with loss of WIP, where platelet volume was normal 25   (1) epidermis, (2) dermal-epidermal junctions, (3) dermis. At low magnification (c,d, left panels) areas of leukocytoclastic vasculitis (4) were evident, which at higher magnification (c,d, right panels) showed vessel wall destruction (5) and neutrophil infiltration (6). (e) ARPC1B is located on Chromosome 7 (position numbering relative to GRCh37), immediately preceded by ARPC1A. (f) Nucleotide positions of identified mutations (black arrows) relative to ARPC1B coding exons (accession #: NM_005720.3). Patient 1 is homozygous for c.269_270dupCT, Patient 2 is homozygous for two missense variants (c.314C4T and c.712G4A). (g) ARPC1B has 6 WD40 repeat domains forming a b-propeller required for Arp2/3 complex function. The amino acid change caused by the mutation in Patient 1 causes a frame shift predicted to yield a protein lacking the last five WD40 domains; both mutations carried by Patient 2 affect WD40 domains. Adapted from http://smart.embl-heidelberg.de/smart/show_motifs.pl; ARC1B_HUMAN, O15143.
Thin section transmission electron microscopy (TEM) was used to examine platelets from the ARPC1B-null patient, a WASP-null patient and a normal donor. All had platelets containing typical cellular structures including mitochondria and a-granules (Figs 3a and 4). ARPC1B-null and WASP-null platelets showed a propensity towards small size (Fig. 4), and examination via IF microscopy ( Fig. 3b) confirmed these platelets to be small and dysmorphic compared to normal. A comparison of circumferential tubulin ring diameters (Fig. 3c) confirmed that both ARPC1B-null and WASP-null platelets are significantly smaller than normal, and do not differ significantly from each other. ARPC1B-null platelets can thus be classified as microthrombocytes, as can ARPC1B-deficient platelets from Patients 2 and 3 (Fig. 5a). A significant proportion (B20%) of ARPC1Bnull platelets shared dysmorphic features with WASP-null platelets (Fig. 3d,e) that included odd shapes, collapse/loss of circumferential microtubule coils and highly variable P-selectin and thrombospondin-1 content (both indicators of a-granules). As has been reported for WAS platelets 28 , whole-mount TEM 29 revealed a reduction/absence of calcium-rich platelet dense granules in ARPC1B-null and -deficient platelets (Fig. 5b,c). Clinical lumi-aggregometry analysis of platelets from Patients 1 and 2 confirmed decreased dense granule ATP release (0.16 and 0.19 nmol respectively; normal range 0.29-1.93 nmol). Platelet aggregation investigations with collagen, SFLLRN, arachidonic acid, ristocetin and ADP were normal for both patients.
Spreading behaviour of ARPC1B-deficient platelets. Actin rearrangements within cells can produce several types of membrane protrusions 30 . Spindle-like filopodia are driven outwards by parallel actin filaments generated by formin protein family members 30,31 , while broad lamellipodia involve branching of actin networks and elongation of filaments 32 . Platelets spreading on a surface typically extend filopodia before producing lamellipodia 33,34 . This process proceeds normally in WASP-deficient platelets, because the nucleation promoting factor required for activating Arp2/3 (refs 35,36) during lamellipodia formation is WAVE/SCAR 37,38 rather than WASP.
We examined the consequences of ARPC1B deficiency for platelet spreading on fibrinogen-treated surfaces using highresolution fluorescence microscopy and scanning electron microscopy to monitor cell morphology and intracellular localization/distribution of tubulin, F-actin and Arp2/3 components. Comparisons of both washed platelets (Fig. 6) and platelets in plasma (Fig. 7) showed that maximally spread cells from normal donors formed typical near-circular lamellipodia. As expected from experimental observations 36 , normal platelet lamellipodia had peripheral localization of F-actin and ARPC5 (Fig. 7), and also displayed prominent F-actin stress fibres and podosome-like nodules. In contrast, maximally spread ARPC1Bnull and ARPC1B-deficient platelets typically formed spiky structures with tubulin-rich tips (Figs 6 and 7) that contained fewer and often elongated F-actin fibres and showed little evidence of podosome-like nodule formation. The spread Immunoglobulins IgG (g l À 1 ) 11.6 (4.5-14.3) 9.7 (1.1-7.0) 11.9 (5.4-13.6) IgA (g l À 1 )

Specific antibodies
Anti-tetanus (IU ml À 1 ) platelet surface area was significantly reduced in ARPC1B-null and -deficient platelets compared to normal ( Supplementary  Fig. 4). Allowing ARPC1B-null and -deficient platelets more time to spread did not alter their observed behaviour. These observations indicate a profound loss of actin branching required for lamellipodia formation 30,32 in ARPC1B-deficient platelets, despite their increased ARPC1A content. This is consistent with experimental observations that isoforms of Arp2/3 containing ARPC1B are significantly better than complexes containing ARPC1A at promoting the rapid assembly of stable branched actin networks 24 .
Proplatelet formation in ARPC1B knockout megakaryocytic cells. The formation of proplatelets by MKs is a key stage in platelet development, which can only proceed to completion in the presence of shear flow 39,40 . The actin cytoskeleton has been linked to the adhesion of proplatelet-forming MKs to the extracellular matrix 41 and in the bifurcation of proplatelets, which increases the number of tips that give rise to platelets 42 . Reduced platelet size and thrombocytopenia are the most common findings in WAS and XLT patients 5 . The cellular mechanisms of WAS-associated microthrombocytopenia are not fully understood. Bone marrow MK numbers are normal in most WAS patients 5,43 , but proplatelet formation may occur prematurely, hampering platelet release into the bloodstream 44,45 .
Peripheral destruction of platelets in the spleen may also be involved, since splenectomy can correct platelet count and size 46,47 . We observed adequate MK numbers and normal morphology in a bone marrow biopsy sample from Patient 1 (Fig. 8). This indicates that MK depletion is not a cause of thrombocytopenia, although the possibility of abnormal proplatelet formation cannot be ruled out. Platelet counts were always low in Patient 1, low (c) Platelet ARPC1A was increased in all three patients relative to normal, while WASP expression was equivalent (GAPDH used as loading control). (d) Immunoblot analysis of a normal platelet lysate after native gel electrophoresis showed a band corresponding to Arp2/3 complex detected by probing for ARPC1B (shown) or other Arp2/3 components ( Supplementary Fig. 3d). This band was resolved on a second dimension SDS-PAGE gel, and immunoblotting confirmed the presence of Arp2/3 components (ARPC1B, ARPC2, ARP2 and ARP3 shown). (e) Immunoblotting of platelet lysates after native gel electrophoresis for ARPC5 (left) showed a greatly reduced level of Arp2/3 in Patient 1 (P1) platelets relative to normal (N). ARPC1A (right) was detected in the Arp2/3 complex in Patient 1 (P1) but not in normal (N) platelet lysate (native GAPDH tetramer used as loading control). (for example, 79 Â 10 9 l À 1 ) to normal in Patient 2, and normal in Patient 3 ( Table 1). The microthrombocytes we observed in all three patients suggest that decreased ARPC1B in MKs affects the cytoskeletal dynamics of platelet formation 42,48 sufficiently to influence platelet size. The thrombocytopenia observed in Patient 1 indicates that total loss of ARPC1B leads to depression of platelet production and/or increased clearance.
To explore the potential impacts of loss of ARPC1B expression on platelet production by ARPC1B-null MKs, we used imMKCL cells, a stable immortalized MK progenitor. Unlike most megakaryocytic cells, imMKCL cells can be stimulated by thrombopoietin to generate proplatelet-producing cells in culture 49 . imMKCL lines lacking functional ARPC1B were made using CRISPR/Cas9 gene deletion. Loss of ARPC1B expression in an ARPC1B knockout line was confirmed via IB (Fig. 9a), which also detected increased ARPC1A expression relative to wild-type cells, paralleling our observations in ARPC1B-null platelets (Fig. 2c). We examined the abilities of thrombopoietin-induced wild-type and ARPC1B-null imMKCL cells to form proplatelet-like extensions in culture, and observed that ARPC1B knockout cells formed proplatelets much less frequently than wild-type cells (Fig. 9b). In addition, while some wild-type imMKCL cells were observed to form multiple proplatelets containing tubulin and branched actin filaments (Fig. 9c), this was not seen in ARPC1B-null cells.
It is likely that the decreased ability of ARPC1B-null imMKCL cells to form proplatelets in culture reflects a similar phenotype in ARPC1B-null MKs, resulting in the thrombocytopenia seen in Patient 1. Since the generation of normal platelets cannot be studied in culture, as the bloodstream is required for final maturation 40 , it is not possible to draw definitive conclusions from this experiment regarding the presence of small platelets   with structural abnormalities observed in our patients (Fig. 3e).
It is reasonable to propose that MKs with altered actin dynamics produce small and dysmorphic platelets, and/or that platelets with altered actin dynamics have increased susceptibility to deformation as they circulate. Our results indicate that, as we observed with platelet spreading (Figs 6 and 7), upregulation of ARPC1A in ARPC1B-null imMKCL cells had little or no effect on restoring Arp2/3 functions required for proplatelet formation. This is in keeping with what we would expect to see in proplatelet formation in the absence of ARPC1B.

Discussion
The Arp2/3 complex is ubiquitous in eukaryotic cells 3 and essential for the survival of multicellular organisms 20 . It has been reported that ARPC1B is both an activator and substrate of Aurora A kinase which is critical in the maintenance of mitotic integrity in mammalian cells 18 . Our results indicate that near-complete loss of ARPC1B expression results in platelet abnormalities including microthrombocytes and spreading defects, and also eczema, leukocytoclastic vasculitis, eosinophilia and elevated IgA and IgE. Elevated IgE levels are also seen in hyper IgE immune deficiency syndrome patients 50 , where the pathological effects of elevated IgE are poorly understood but likely involve several immune pathways, including increased Th2 cytokine production 51 . ARPC1B-deficient Patients 2 and 3 have normal platelet numbers, while ARPC1B-null Patient 1 has persistent thrombocytopenia. Our experiments with imMKCL cells indicate that this is likely linked to decreased proplatelet production by ARPC1B-null MKs. It may also be that these patients have increased rates of peripheral platelet clearance as observed in WAS 5 , which may be associated with their elevated IgA levels. If so, this would exacerbate the consequences of low platelet production by ARPC1B-null MKs.
Our results indicate that ARPC1B deficiency is associated with severe multisystem disease including recurrent infections, inflammatory changes in the intestine (crypt distortion with severe eosinophilic infiltration) and elevated autoimmunity markers (anti-nuclear and anti-neutrophil cytoplasmic antibodies; Table 1). This multi-system pathology is similar to that produced by complete loss of WASP expression 5 , and it is consistent with the predominant expression of ARPC1B in haematopoietic/immune cells (Supplementary Table 3). These include B-and T-lymphocytes (including T-regulatory cells and natural killer cells), antigen-presenting dendritic cells, monocytes/ macrophages and neutrophils. All of these cells require coordinated actin dynamics for development, migration, recruitment, signalling and activation of innate and adaptive immune responses 52 .
We observed considerable heterogeneity in the manifestation of disease among the patients we studied. All three had failure to thrive, platelet abnormalities, eosinophilia, eczema and other indicators of inflammatory/immune disease including elevated IgA and IgE, and anti-neutrophil autoantibodies. Only Patient 1 had chronic infections and colitis. Patient 3 had a history of eczema-like rash from birth, although not severe enough to be biopsied, and it is unclear if underlying vasculitis is associated with this rash, as was seen in Patients 1 and 2. In addition to ARPC1B mutations, there are likely to be other genetic and/or environmental factors associated with the different disease phenotypes observed among patients. It is nevertheless tempting to propose that as with thrombocytopenia, the milder immune  indicate the presence of generally small and morphologically variable platelets in patient samples (see also Fig. 3). Dense granules were evaluated by whole mount transmission electron microscopy (Fig. 5b,c).
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14816 ARTICLE manifestations observed in Patients 2 and 3 may be attributable to their residual ARPC1B expression compared to Patient 1. This would parallel the milder phenotype observed in XLT that is attributed to residual WASP expression 5 .
As with WAS 11,52 , many questions remain regarding the mechanisms associated with the abnormalities we have observed in patients with absent/reduced ARPC1B expression. Our observations that the loss of ARPC1B has profound consequences for platelet spreading, and most likely for MK proplatelet production, are consistent with cellular phenotypes that would be expected with the loss of Arp2/3 function in these cells. The effects of ARPC1B loss would presumably be most severe in cells/tissues where it is the predominant isoform present in functional Arp2/3, since as we observed in platelets and MKs, compensatory upregulation of ARPC1A has little effect. This is likely due to cell lineage-specific variations in Arp2/3 assembly and/or function 24 .
While it is difficult to connect our experimental observations to the entire spectrum of disease observed in ARPC1B-deficient patients, Arp2/3-driven actin polymerization has recently been reported to be essential for several relevant cellular, physiological and developmental processes. These include cell secretion 53 , phagocytosis 54,55 , autophagy 56 , migration 57,58 , haptotaxis 59 , focal adhesions 60 and intracellular tight junctions required for epidermal barrier formation 61 , vesicle trafficking and transcytosis in the small intestine 62 . With regards to immunity and inflammation, Arp2/3 function has been reported to be critical for the formation of immune cell synapses 63,64 and T-regulatory cell function, which is aberrant in WAS patients leading to a high susceptibility to develop Th2-mediated food allergies 65 . Genetic correction of induced pluripotential stem cells from WAS patients demonstrated the restoration of defective natural killer and T-lymphoid cell development and function, confirming the critical role of WASP 66 . Given the central role that Arp2/3 plays in so many processes, it is reasonable to expect that ARPC1B deficiency may be associated with a broad range of developmental and immune defects. Our results also point to the possibility that gene variants affecting other Arp2/3 components may be associated with human disease.

Methods
Subjects. All experiments were carried out with the approval of the Research Ethics Board at the Hospital for Sick Children, Toronto, Canada. Informed consent to participate in research was obtained from all participants. A copy of the consent is available on the interNational Early Onset Paediatric IBD Cohort Study (NEOPICS) website at http://www.neopics.org/study-documents.html. Patients were consented to the registry and tissue bank of the Canadian Centre for Primary Immunodeficiency. Sequencing of the patient with Wiskott-Aldrich syndrome (WASP null) was done by PreventionGenetics (Marshfield, WI, USA) for clinical diagnosis. A hemizygous missense variant c.256C4T in the WAS gene was predicted to result in the amino acid substitution p.Arg86Cys, previously documented to cause WAS 67 .
Genomic sequencing data analysis and validation. WES was performed at the Centre for Applied Genomics, Hospital for Sick Children, Toronto, Canada. Exome library preparation was performed using the Ion Torrent AmpliSeq RDY Exome Kit following the manufacturer's recommended protocol. In brief, 100 ng of DNA quantified by Qubit DNA HS or BR assay was used in the target amplification under the following conditions: 99°C for 2 min, followed by ten cycles at 95°C for 15 s and 60°C for 16 min, and final hold at 10°C. Incorporated primer sequences were partially digested using a proprietary method. Ion Torrent Proton adapters were ligated to the amplicons at 22°C for 30 min followed by 72°C for 10 min, and the library was purified with Agencourt Ampure XT Beads. Libraries were quantified by qPCR, and 7 pM used for sequencing on an Ion Torrent Proton Sequencer using a PI chip V2 following the manufacturer's protocol. All data were aligned to the hg19/GRCh37 reference genome and quality trimmed via Ion Torrent Suite Version 4.  were then scored using the database for non-synonymous functional predictions (dbNSFP 2.8), filtering out variants found to have no damaging score (Polyphen2, SIFT, MutationTaster, MutationAssessor, FATHMM). As well, dbNSFP scores variants with conservation scores (PhyloP and GERP þ þ ). Sanger sequencing was performed to validate mutations identified by WES. The following primers were used to sequence ARPC1B Exon4 (forward 5 0 -GCAGA TACAGCTTCCACC-3 0 and reverse 5 0 -CCCTAACAGCCCACTC-3 0 ) and ARPC1B Exon7 (forward 5 0 -GCTGAGAGTACAGGTGCG-3 0 and reverse 5 0 -CCTGCTGTGACCACACAC-3 0 ).
Identification of ARPC1B mutations. Patient 1. WES of Patient 1 and parents resulted in the identification of 50,020 variants relative to the reference genome Hg19. In all, 31,997 were considered high quality, passing thresholds for genotype quality and read depth. A total of 2,853 were considered rare variants with a minor allelic frequency less than 1% (MAFo0.01) using three databases: NHLBI ESP6500SI-V2, 1,000 Genomes and ExAC v.0.3. 1,046 variants were predicted to be damaging, as they were classified as missense mutations or predicted to cause loss of function. From this list we identified 14 non-synonymous homozygous variants (Supplementary Data) inherited in an autosomal recessive manner. The ARPC1B variant for which Patient 1 is homozygous is a two base pair duplication (c.269_270dupCT; Fig. 1f) causing a frame shift in exon 4 at position 91 and a premature stop codon, with a predicted truncated 119 amino acid protein lacking five of the six WD40 domains (Fig. 1g) required for formation of the functional p40/ARPC1 b-propeller 16 . c.269_270dupCT is a novel variant and the only one predicted to cause loss of function by RefSeq version 105v2 (Supplementary Table 1); it was validated by Sanger sequencing and independent WES. Patient 2. WES of Patient 2 and parents resulted in the identification of 51,866 variants. In all, 35,114 were considered high quality, passing thresholds for genotype quality and read depth. A total of 1,704 were considered rare variants with a minor allelic frequency less than 1% (MAFo0.01; see above). In all, 676 variants were predicted to be damaging as they were classified as missense mutations or caused loss of function. From this list we identified nine non-synonymous homozygous variants (Supplementary Data) inherited in an autosomal recessive manner. Two homozygous ARPC1B missense variants were found near the same region of the mutation identified in Patient 1 (c.314C4T and c.712G4A encoding p.Ala105Val and p.Ala238Thr; Fig. 1f). Analysis using the ExAC, NHLBI, ESP and 1,000 Genomes databases revealed the c.712G4A variant to be rare and predicted to be benign, while c.314C4T is novel and predicted to be damaging (Supplementary Table 1), disrupting the second ARPC1B WD40 domain and thus the p40/ARPC1 b-propeller. Both variants were validated using Sanger sequencing.
Patient 1's parents are consanguineous, and he has a very severe and complex disease. Therefore we focused our initial genetic analysis on Mendelian autosomal recessive mutations with a homozygous inheritance pattern that could explain his disease. As shown in Supplementary Data no homozygous mutations were detected in predicted pathogenic variants for known genes associated with immune deficiency (including WASP and WIP) or platelet disorders. Also, no overlapping compound heterozygote mutations, X-linked mutations, or de novo mutations were shared by Patient 1 and Patient 2. We then focused on novel genes and examined known biological function, known diseases associated with genes, gene expression profiles and available animal models of the candidates outlined in Supplementary Information. The only gene that fit the disease profile observed in Patient 1 was ARPC1B, since the role of ARPC1B as the WASP-binding component of the Arp2/3 complex pointed to a WAS-like phenotype. In Patient 2, with a similar spectrum of disease, WES also identified ARPC1B as the only viable candidate, and this was the only gene found to be mutated in both Patients 1 and 2 (see Venn diagram Supplementary Fig. 1).
Platelet immunoblot and native gel electrophoresis analysis. Blood from patients and normal controls was collected by venipuncture with 3.2% sodium citrate anticoagulation and centrifuged (150g, 15 min) before collection of platelet-rich plasma (PRP), from which platelets were pelleted via centrifugation Washed platelets were allowed to spread on fibrinogen-treated coverslips for 45 min prior to fixation and imaging by SEM (right column; bars ¼ 2 mm), or by spinning disc laser fluorescence confocal microscopy (left columns; 3D renders of deconvolved z-series, bars ¼ 1 mm) after surface staining with wheat germ agglutinin (WGA; red), staining for F-actin with phalloidin (green) and immunostaining for alpha tubulin (magenta). Maximally spread platelets from a normal donor (top row) typically show fully formed lamellipodia with thin tubulin filaments and multiple F-actin filaments intersecting with podosome-like actin nodules. In contrast, maximally spread platelets from both ARPC1B-null (middle row) and ARPC1B-deficient (bottom row) patients tend to form spiky filopodial-lamellipodial structures lacking podosomes that sometimes show elongated actin filaments (see also Fig. 7). (1,000g, 10 min). Prior to obtaining lysates, platelets were washed twice by resuspension in phosphate-buffered saline (PBS) buffer adjusted to pH 6.1 with ACD (PBS/ACD) and pelleting. For direct IB platelets were resuspended at 10 9 per ml in PBS plus 2 Â protease inhibitor (Roche Complete EDTA-free, Roche Diagnostics) and lysed with Triton-X100 (0.5%). For native gel electophoresis, washed platelets were resuspended as above with added phosphatase inhibitor (Roche PhosSTOP) and lysed by sonication via two 5 s pulses 1 min apart in a Heat Systems Sonicator Ultrasonic Processor XL XL2010 (Farmingdale, NY, USA) set at 20 V and amplitude 2. All lysates were cleared by centrifugation at 21,000g for 2 min after which the supernatants were retained and analysed for protein content by IB 68 or for Arp2/3 complex via blue Native gel electophoresis, where sonicated platelet lysates were applied to a 4-16% Bis-Tris    491, 561, 642 nm), emission filters specific for Alexa Fluor dyes: 405 (447±60 nm), 488 (525 ± 50 nm), 568 (593 ± 40 nm) and 647 (676 ± 29), and an ASI motorized XY stage controlled with an Improvision Piezo Focus Drive. Images were acquired with 250 nm Z-stepping via an Olympus UPLSAPO Â 100/1.40 NA oil objective and a Â 1.5 internal magnification lens (Spectral Applied Research) for a final magnification of Â 150. Laser intensity, camera and exposure settings were established with minimal/undetectable levels of autofluorescence, channel crosstalk and non-specific primary/secondary background fluorescence. Acquisition, image deconvolution, registry correction (maximum one z-pixel required in our system) and cell surface area analysis were done with Volocity 6 software (Perkin-Elmer). Laser fluorescence confocal structured illumination microscopy (SIM) was done using a Zeiss ELYRA PS.1 microscopy system (Axio Observer Z1 core) and a Â 63/1.4 NA oil-immersion objective with Â 1.6 optovar. The system is equipped with an Andor iXon3 885 detector, 405, 488, 561 and 640 nm laser lines, Zeiss motorized XY stage and Z-piezo focus. Acquisition control and SIM image processing (including channel alignment) were done with Zeiss Zen 2012 software using optimized settings and current calibration data sets. Rendered volume images were created from laser fluorescence confocal SIM and spinning disc confocal microscopy data using Imaris 8 software. Images were exported to Adobe Photoshop for labelling and presentation.
Electron microscopy. TEM of resting platelets was done as previously described 68 . Briefly, PRP was fixed with 2.5% glutaraldehyde in PBS and fixed overnight. Subsequently, platelets were post-fixed with 2% osmium tetroxide in H 2 O for 1 h and dehydrated in a graded series of acetone before embedding in Epon-Araldite. Thin sections were cut and stained with uranyl acetate and lead citrate. Grids were examined with a JEOL JEM-1011 electron microscope at 80 kV. Images were captured with a side-mounted Advantage HR CCD camera (Advanced Microscopy Techniques). Platelet whole-mount imaging was done by placing 2-3 drops of PRP on to Formvar-coated nickel grids (Electron Microscopy Sciences) for 5 min; excess liquid was removed with a filter paper followed by a 5-min fixation with 2.5% glutaraldehyde in PBS pH 7.4. After rinsing with distilled water the grids were placed into a JEOL JEM-1011 electron microscope with a 300/20 mm condenser/objective aperture. Dense granules were quantified by counting the number of dark spots in whole mounted platelets per TEM 29 . Granules were scored in a minimum of 50 platelets at Â 15,000-100,000 magnification, and the mean number of dense granules per platelet was calculated.
Scanning electron microscopy was performed with samples of spread platelets fixed with glutaraldehyde (above) that were dehydrated, sputter coated with gold to 20 nm thickness in a Leica EM ACE200 high vacuum sputter coater and dried in a Bal-Tec CPD030 critical point dryer (32°C, 75 bar). Imaging was done with a Philips XL-30 ESEM environmental scanning electron microscope.
Data availability. The whole exome sequencing data that support the findings of this study are available from the corresponding authors W.H.A.K and A.M.M on request. The data are not publicly available because they contain information that could compromise research participant privacy/consent. All other data generated or analysed during this study are included in this published article (and its Supplementary Information files) and available from the corresponding authors on request.