Maintenance of murine platelet homeostasis by the kinase Csk and phosphatase CD148

Platelets are highly reactive fragments of megakaryocytes (MKs) that interrogate the vessel wall and prevent excessive blood loss following injury. They do so by adhering to exposed extracellular matrix proteins and forming a thrombus that transiently occludes the blood vessel, promoting wound repair and vessel regeneration. However, it remains unclear how the reactivity of platelets in the circulation is regulated.

Platelets contain high levels of Src family kinases (SFKs) that have a coordinating role in initiating and propagating primary activation signals.^1,2  The 3 most abundant SFKs in human and mouse platelets are Src, Lyn, and Fyn (supplemental Figure 1, available on the Blood Web site).^3,4  Src and Fyn act primarily as positive regulators of platelet activation,^5-8  whereas Lyn is both a positive and negative regulator of activation.^9,10  SFKs are constitutively associated with the cytoplasmic tails of several important platelet receptors, including the glycoprotein Ib (GPIb)–IX-V complex,^11,12  the immunoreceptor tyrosine-based activation motif (ITAM)–containing GPVI-Fc receptor (FcR) γ-chain receptor complex,^5,13  and the integrin αIIbβ3.^14,15  SFKs are also involved in transmitting secondary activation signals from G protein–coupled receptors.^16-21 

Equally important, but less well understood are the inhibitory functions of SFKs. Concomitant with phosphorylating ITAM-containing receptors activating platelets,^5,22  SFKs also phosphorylate immunoreceptor tyrosine-based inhibition motif (ITIM)–containing receptors that inhibit platelets.²³  The latter recruit SH2 domain-containing phosphatases, including the nontransmembrane protein-tyrosine phosphatases Shp1 and Shp2,^24-26  that inhibit platelet activation. The 2 most well-characterized platelet ITIM-containing receptors are PECAM-1 and G6b-B, which inhibit platelet activation and modulate platelet homeostasis, respectively.^27,28 

Limited knowledge has been gained about how platelet SFKs are regulated. SFKs are primed by dephosphorylation of the C-terminal inhibitory phosphotyrosine residue by the receptor-like tyrosine phosphatase CD45 in most hematopoietic cells²⁹  and by CD148 in platelets.^30-32  Primed SFKs need to be fully activated through trans-autophosphorylation of a tyrosine residue located within the catalytic domain.³³  On the other hand, inactivation of SFKs can occur through the action of C-terminal Src kinase (Csk) and the structurally related Csk homologous kinase (Chk), also referred to as megakaryocyte-associated tyrosine kinase (Matk), both of which target the C-terminal inhibitory tyrosine residue in SFKs.^34,35  However, it remains unknown how the threshold of individual SFK activation is set in platelets

To address these questions, we generated a series of MK-specific conditional knockout (KO) mouse models leading to deletion of Csk, CD148, or Csk/CD148 and studied the consequences of these deletions for platelet physiology. Disruption of the homeostatic balance of SFK activity in MKs had dramatic and unexpected consequences on the number and reactivity of platelets in the circulation due to cell-intrinsic negative feedback mechanisms involving the ITIM-containing receptor G6b-B and Chk, culminating in antithrombotic and hemorrhagic outcomes.

Csk^fl/fl, CD148^fl/fl, Pf4-Cre⁺, and Csk^AS mice were generated as previously described.^36-39  All mice were on a C57BL/6 background. All procedures were undertaken with United Kingdom Home Office approval in accordance with the Animals (Scientific Procedures) Act of 1986.

Thiazole orange (BD Retic-Count) was from BD Biosciences. PP1 analog IV (3-IB-PP1) was from Calbiochem. Antibodies are described in detail in supplemental Methods.

Thrombocytopenia was induced as previously described.²⁸ 

Blood was collected from terminally CO₂-narcosed mice from the abdominal vena cava into 1:10 (v/v) acid-citrate-dextrose anticoagulant. Washed platelets were prepared as previously described.⁴⁰  Adenosine diphosphate (ADP)-sensitive washed platelets were prepared in the presence of ADP scavenger apyrase as previously described.⁴¹  Platelet counts were normalized and used for spreading (2 × 10⁷/mL), aggregation (2 × 10⁸/mL), or biochemical analysis (4-5 × 10⁸/mL).

Platelet aggregation, adenosine triphosphate (ATP) secretion, spreading, clot retraction, platelet adhesion under flow, total thrombus formation analysis system (T-TAS), stimulation for biochemical analysis, immunoblotting, and immunoprecipitation are described in detail in the supplemental Methods.

Reticulated platelets in blood were double stained with αIIb and thiazole orange (BD Retic-Count; BD Biosciences). Surface protein expression was measured in blood or bone marrow cells with indicated FITC- or PE-conjugated antibodies by flow cytometry (BD Accuri C6 for platelets; BD FACSCalibur for bone marrow cells), as previously described.^25,28 

Experiments were performed on 8- to 10-week-old KO and litter-matched wild-type (WT) mice as previously described.³⁰ 

Laser-induced injury of arterioles in the cremaster muscle and ferric chloride (FeCl₃)-induced injury of carotid arteries were performed and analyzed as previously described.⁴² 

Data presented are mean ± standard error of the mean (SEM). One-way or 2-way analysis of variance (ANOVA) followed by post-hoc tests were used to determine statistical significance (P < .05). Further details on statistical analysis are provided in the supplemental Methods.

MK-specific Csk and CD148 conditional KO (Csk and CD148 KO) mice were generated by crossing Csk^fl/fl and CD148^fl/fl mice with Pf4-Cre⁺ transgenic mice, respectively.^36-38 Csk/CD148 conditional double KO (DKO) mice were also generated to determine whether deletion of both enzymes simultaneously would rescue phenotypes of single KO mice and to reveal novel mechanisms of SFK regulation (Figure 1A).

Platelet count in Csk KO mice was reduced by 65% and platelet volume increased by 33% (Figure 1B; supplemental Table 1), suggesting defects in platelet production and/or clearance. This was supported by a twofold increase in the proportion of reticulated, immature platelets in these mice (Figure 1C; supplemental Figure 2A). The same parameters were normal in CD148 KO mice (Figure 1B-C; supplemental Figure 2A). Although the platelet count was normal in DKO mice, the variability in numbers was greater than in WT and Csk KO mice, and the proportion of large, reticulated platelets was similar to that in Csk KO mice (Figure 1B-C; supplemental Figure 2A; supplemental Table 1). Increased numbers of immune and granulocytic cells in Csk and DKO mice (supplemental Table 1) may be an indirect consequence of nonspecific deletion of floxed Csk and CD148 in other hematopoietic lineages by the Pf4-Cre transgene, which is reported to be “leaky.”⁴³  Interestingly, the proportions of P-selectin⁺αIIb⁺ platelets were normal in Csk KO mice and only marginally elevated in DKO mice (supplemental Figure 2B), pointing to only minimal preactivation of these platelets in the circulation or downregulation of activation markers.

To explore whether increased platelet size and proportion of young platelets correlate with changes in other platelet parameters, we quantified levels of receptors that regulate platelet production and activation. Notably, several receptors we focused on, including αIIbβ3, GPIbα, GPVI-FcR γ-chain, CLEC-2 ,and G6b-B, rely on SFKs to transmit signals.²  Surface levels of integrin αIIb were marginally increased in DKO platelets, whereas GPIbα levels were increased by 63% and 38% in Csk KO and DKO platelets, respectively (Figure 1Di). Surface levels of GPVI were reduced by 77%, 44%, and 88% in Csk KO, CD148 KO and DKO platelets, respectively, correlating with total protein levels of GPVI and FcR γ-chain (Figure 1Di-ii). Surface and total protein levels of CLEC-2 were reduced in Csk KO and DKO platelets, which were significantly more pronounced in DKO (Figure 1Di-ii). In contrast, surface expression of the inhibitory receptor G6b-B was increased by 61% and 73% in Csk KO and DKO platelets, respectively, correlating with total protein levels (Figure 1Di-ii). Surface levels of the integrin α2 subunit and metalloproteinase ADAM10 that mediates shedding of GPVI were normal (supplemental Figure 3A). Intriguingly, expression of surface receptors in αIIb⁺ bone marrow cells was normal in all 3 genotypes (supplemental Figure 3B), suggesting autoregulation of these receptors in platelets. Collectively, these findings demonstrate downregulation of the (hemi-)ITAM–containing activation receptors GPVI-FcR γ-chain and CLEC-2, and concomitant upregulation of the ITIM-containing inhibitory receptor G6b-B in Csk-deficient platelets.

To investigate the cause of low platelet counts, we measured the rate of platelet recovery following anti-GPIbα antibody–mediated platelet depletion. This was significantly reduced in Csk KO mice, marginally reduced in CD148 KO mice, and normal in DKO mice (Figure 1Ei-ii). The clearance of biotinylated platelets was normal in mice of all 3 genotypes (supplemental Figure 4A-B). Both Csk KO and DKO mice exhibited splenomegaly (supplemental Figure 5A-B). Elevated MK counts were observed in spleens in all 3 genotypes, which was associated with myelofibrosis (supplemental Figure 6A). MK counts were moderately elevated in the bone marrow of Csk KO mice, but with no associated myelofibrosis (supplemental Figure 6B). We observed more P-selectin⁺αIIb⁺ hematopoietic cells in Csk KO and DKO mice (supplemental Figure 6C), which may explain the presence of myelofibrosis in the spleen of these mice.⁴⁴  Together, these findings suggest defective MK activation in Csk KO and DKO mice.

We hypothesized that Csk KO mice would be predisposed to thrombosis because of increased SFK activity, rendering platelets hyperactive. To test this, we analyzed mice of all 3 genotypes in established in vivo models of hemostasis and thrombosis. Paradoxically, Csk KO mice exhibited increased bleeding in a tail injury model (Figure 2A). CD148 KO mice exhibited normal hemostasis, whereas DKO mice had increased bleeding compared with Csk KO mice (Figure 2A). Increased bleeding in Csk KO and DKO mice could not be explained solely by reduced platelet count, which was normal in DKO mice.

To explore the kinetics of arterial thrombus formation, we employed the laser injury–induced thrombosis model in small arterioles in mice. Thrombus formation and fibrin deposition were moderately and severely reduced in Csk KO and DKO mice, respectively (Figure 2Bi-iv; supplemental Videos 1 and 2). Thrombus formation was also severely compromised in CD148 KO mice, whereas fibrin deposition was only moderately reduced (Figure 2Bi-iv). Similar results were observed following FeCl₃-induced injury of the carotid artery, albeit thrombus formation was marginally better in CD148 KO than in Csk KO mice (Figure 2Ci-iii; supplemental Video 3). These results demonstrate that the ability of Csk KO and DKO mice to form thrombi in vivo is markedly reduced, resulting in increased bleeding.

To determine the cause of reduced thrombus formation observed in the different in vivo models, we employed the T-TAS microfluidic device, which can differentiate between platelet- and coagulation-driven defects. We found that platelets of all 3 genotypes displayed severe defects in forming stable platelet thrombi on collagen surface, as we were unable to detect an increase in the flow pressure (Figure 3Ai). However, under conditions where platelets were flowed over collagen in the presence of tissue factor to maximize coagulation, platelets of all 3 genotypes formed equally stable thrombi (Figure 3Aii), suggesting coagulation is unaltered in these mice and that defects in thrombus formation are intrinsic to platelets.

To identify specific platelet defects, we measured multiple parameters of platelet function and thrombus formation ex vivo at arterial shear rates on collagen I; von Willebrand factor–binding peptide (vWF-BP) and laminin; and vWF-BP, laminin, and the snake venom rhodocytin, which induces CLEC-2 signaling. Platelet deposition and thrombus formation on collagen I were markedly decreased Csk KO and CD148 KO samples and further reduced in DKO samples (Figure 3Bi-iii,vi). α-Granule secretion, integrin αIIbβ3 activation, and phosphatidylserine exposure (determined as P-selectin, JON/A, and Annexin V staining, respectively) followed the same trends (Figure 3Biv-vi). This likely is a consequence of markedly reduced surface levels of GPVI-FcR γ-chain, defective SFK signaling, and increased inhibitory mechanisms. We found no difference between genotypes in platelet deposition on the vWF-BP and laminin surface on which adhesion is mediated by platelet GPIbα via immobilized plasma vWF and platelet integrin α6β1, which binds laminin,⁴⁵  suggesting that Csk and CD148 do not play major roles regulating GPIbα or α6β1 signaling pathways (Figure 3Bi-ii,vi). Upon complementing the vWF-BP and laminin surface with the CLEC-2 ligand rhodocytin, which drives a more robust platelet deposition,⁴⁶  adhesion was significantly reduced in the 3 KOs (Figure 3Bi-ii,vi). Notably, CLEC-2 expression was marginally and markedly reduced in Csk KO and DKO platelets, respectively, demonstrating that Csk and CD148 participate in SFK regulation during CLEC-2-evoked hemi-ITAM signaling, as previously described.³² 

We next analyzed platelets of all 3 genotypes in a range of in vitro functional assays. Consistent with previous findings, CD148 KO platelets exhibited reduced aggregation and ATP secretion to collagen and CRP stimulation, which was overcome at high concentrations of CRP (Figure 3Ci-ii). Csk KO and DKO platelets also exhibited reduced reactivity to collagen and failed to respond to collagen-related peptide (CRP) (Figure 3Ci-ii). We also investigated the aggregation and ATP secretion response of platelets to anti-CLEC-2 antibody–mediated activation. CD148 KO platelets exhibited an increase in the lag time at low concentration of anti-CLEC-2 antibody and responded normally to a high concentration (Figure 3Ciii). In contrast, Csk KO and DKO platelets exhibited shortened lag times in response to anti-CLEC-2 antibody, despite expressing lower levels of CLEC-2 receptor (Figures 1Di-ii and 3Ciii). Platelets from all 3 mouse models responded normally to thrombin and the TxA₂ analog U46619 (Figure 3Civ-v). As expected, WT platelets prepared in the absence of apyrase did not respond to 10 μM ADP because of P2Y₁ and P2Y₁₂ receptor desensitization (Figure 3Cvi); however, Csk KO and DKO platelets were able to aggregate and secrete ATP in response to the same concentration of ADP (Figure 4Cvi), suggesting reduced P2Y₁ and P2Y₁₂ internalization or increased signaling via these receptors, both of which use SFKs to transmit signals.²  Indeed, washed platelets prepared in the presence of apyrase, preventing receptor desensitization, responded normally to 3 and 10 μM ADP, but DKO platelets hyperresponded (Figure 3Cvii; supplemental Figure 7), supporting the hypothesis of increased P2Y₁₂ signaling.

We investigated platelet adhesion and spreading on fibrinogen, which is dependent on SFK-mediated αIIbβ3 outside-in signaling and cytoskeletal remodeling. As expected, Csk-deficient platelets spread to a greater extent than WT platelets, whereas CD148-deficient platelets exhibited reduced spreading (Figure 3Di-ii). However, DKO platelets and WT platelets spread comparably, suggesting that negative feedback mechanisms were activated in DKO platelets. When platelets were preactivated with thrombin, spreading of Csk KO and DKO platelets was significantly increased compared with WT platelets (Figure 3Di-ii).

To further test the role of Csk and CD148 in αIIbβ3 signaling, we assessed clot retraction in these mice in vitro, revealing a significant reduction in clot retraction of CD148 KO platelets (Figure 3E). These results strongly suggest an important role for functional Src kinase activity in clot retraction.

To determine the mechanism underlying the hypothrombotic phenotypes, we assessed SFK activity in unstimulated platelets of all three genotypes. We measured trans-autophosphorylation of the activation loop tyrosine residue of SFKs (Src p-Tyr418) as an indirect indicator of SFK activity in resting platelets by quantitative capillary electrophoresis-based immunoassays (ProteinSimple Wes). As expected, deletion of Csk resulted in a significant increase in Src p-Tyr418, whereas deletion of CD148 resulted in a marked decrease in Src p-Tyr418 (Figure 4Ai-ii). Deletion of both Csk and CD148 resulted in an unexpected overall increase in Src p-Tyr418, demonstrating high activity SFKs in these platelets (Figure 4Ai-ii,B).

Phosphorylation of the C-terminal inhibitory tyrosine residues of Lyn (Tyr507), Src (Tyr529), and Fyn (Tyr530) is typically inversely related to SFK activity and Src p-Tyr418. Indeed, phosphorylated Lyn Tyr507 (Lyn p-Tyr507) was not detected in resting Csk-deficient platelets (Figure 4Ai-ii), and phosphorylation of Src Tyr529 (Src p-Tyr529) and Fyn Tyr530 (Fyn p-Tyr530) were markedly decreased, suggesting that Csk is the main kinase that phosphorylates these residues and attenuates the activity of these SFKs. Conversely, phosphorylation of all three inhibitory tyrosine residues was increased in resting CD148-deficient platelets (Figure 4Ai-ii), confirming previous findings that CD148-induced dephosphorylation activates SFKs in platelets.³¹  A partial rescue of Lyn p-Tyr507, Src p-Tyr529, and Fyn p-Tyr530 was observed in DKO platelets (Figure 4Ai-ii), suggesting that in the absence of Csk and CD148, another kinase phosphorylates these residues, the obvious candidate being Chk, which was upregulated in Csk KO and DKO platelets (Figure 4C), explaining the rescue of inhibitory site phosphorylation of SFKs in DKO platelets. Expression of the tyrosine phosphatases PTP-1B, Shp1, and Shp2, all of which have been implicated in regulating SFK activity, were normal in all 3 genotypes (Figure 4C). Interestingly, SFK activity (Src p-Tyr418) was highest in DKO platelets, despite increased phosphorylation of the C-terminal inhibitory tyrosine residues of Lyn, Src, and Fyn, compared with Csk KO platelets. It remains to be determined whether phosphorylation of the activation loop and C-terminal inhibitory tyrosine residues coexist in the same molecules or in distinct pools of SFKs. Dually phosphorylated SFKs have been reported to be active in T cells.⁴⁷ 

We next assessed whether SFK activity in MKs is comparable to platelets. Indeed, SFKs were similarly phosphorylated in starved bone marrow–derived MKs (supplemental Figure 8A-B) as they were in resting platelets of the same genotype (Figure 4Ai-ii), compatible with increased SFK activity in Csk KO and DKO MKs and decreased SFK activity in CD148 KO MKs, which may underlie aberrant MK function and myelofibrosis in Csk KO and DKO mice.

Collectively, these findings demonstrate that Csk is an inhibitor of SFK activity in platelets and MKs. The finding that DKO platelets display markedly higher SFK activity than Csk KO platelets suggests that CD148 can act as a dual activator and inhibitor of SFK activity in platelets and MKs (Figure 4B).

To determine why platelets with high SFK activity were hyporeactive, we investigated tyrosine phosphorylation downstream of GPVI-FcR γ-chain, CLEC-2, and αIIbβ3, all of which rely on SFKs and Syk to initiate and propagate signaling. Because of the dramatic reduction of GPVI-FcR γ-chain expression in Csk KO and DKO platelets (Figure 1Di-ii), platelets were stimulated with a high concentration of collagen (30 μg/mL), which signals primarily through GPVI-FcR γ-chain. Basal whole-cell tyrosine phosphorylation (p-Tyr) and Src p-Tyr418 were significantly higher in resting DKO platelets and collagen-stimulated Csk KO and DKO platelets than WT platelets (Figure 5Ai-iii; supplemental Figure 9A). In contrast, basal and collagen-mediated Src p-Tyr418 was significantly lower in CD148 KO platelets than WT platelets (Figure 5Ai-iii; supplemental Figure 9A). Induced phosphorylation of the FcR γ-chain, which is directly mediated by SFKs and acts as a docking site for Syk, was less pronounced in Csk KO and DKO platelets (Figure 5Ai) due to reduced expression of the FcR γ-chain in these platelets (Figure 1Dii). Syk activation was indirectly measured as phosphorylation of Syk tyrosine residues 525 and 526 (Syk p-Tyr525/6) and correlated directly with FcR γ-chain levels and phosphorylation. Syk p-Tyr525/6 was highest in collagen-stimulated WT platelets and increased only marginally in collagen-stimulated platelets of all 3 genotypes (Figure 5Aii-iii; supplemental Figure 9A), despite Csk KO and DKO platelets having high SFK activity. These findings support a model in which receptor-mediated membrane localization of Syk is essential for activation.

Platelets were also stimulated with a high concentration of anti-CLEC-2 antibody (10 μg/mL), mimicking podoplanin-mediated cross-linking of the receptor. The pattern and intensity of whole-cell p-Tyr and SFK phosphorylation generally mirrored that of collagen-stimulation in the various genotypes (Figure 5Bi-iii; supplemental Figure 9B). However, Syk p-Tyr525/6 was much higher in Csk KO platelets than any of the other genotypes (Figure 5Bii-iii; supplemental Figure 9B), despite reduced CLEC-2 expression. Despite the reduced CLEC-2 and GPVI-FcR γ-chain levels in DKO platelets, stimulation by anti-CLEC-2 antibody resulted in normal Syk p-Tyr525/6, whereas stimulation by collagen led to significantly reduced Syk phosphorylation, correlating with normal and reduced aggregation, respectively.

We also investigated αIIbβ3 signaling to determine the cause of increased spreading of Csk KO platelets on fibrinogen (Figure 3Di-ii). To initiate signaling, αIIbβ3 relies mainly on Src and, to a lesser extent, Fyn and is localized exclusively in nonlipid rafts.^8,15,48,49  Whole-cell p-Tyr was increased in fibrinogen-adhered Csk KO and DKO platelets (Figure 5Ci), demonstrating a general increase in outside-in integrin signaling in these platelets. In agreement with increased spreading of Csk KO platelets but normal spreading of DKO platelets on fibrinogen (Figure 3Di-ii, basal), we found that SFK and Syk activity was increased in Csk KO platelets, but not in DKO platelets (Figure 5Cii-iv; supplemental Figure 9C).

In addition to the elevated expression of the ITIM-containing receptor G6b-B in Csk KO and DKO platelets (Figure 1Di-ii), we found a marked increase in G6b-B tyrosine phosphorylation and binding of the tyrosine phosphatases Shp1 and Shp2 under resting and collagen-stimulated conditions (Figure 5Di), suggesting increased inhibitory signaling via the G6b-B-Shp1-Shp2 complex in Csk KO and DKO platelets (Figure 5Dii). Thus, increased SFK activity in Csk KO and DKO platelets leads to compensatory upregulation of inhibitory ITIM signaling and parallel downregulation of the (hemi-)ITAM–containing GPVI-FcR γ-chain and CLEC-2 receptors, explaining the reduced activity of these platelets.

To circumvent developmental and compensatory mechanisms arising from deletion of Csk, we used a transgenic mouse model expressing a PP1-analog (3-IB-PP1)–sensitive form of Csk (Csk^AS), enabling rapid and specific inhibition of Csk^AS in these mice.³⁹  Platelet count, volume, and receptor expression were normal in Csk^AS mice (data not shown). Platelets from Csk^AS mice aggregated and secreted normally to all agonists tested in the presence of vehicle alone (dimethyl sulfoxide) (Figure 6Ai-iv) and marginally more robustly to a subthreshold concentration of anti-CLEC-2 antibody (3 μg/mL) in the presence of 10 μM 3-IB-PP1 (Figure 6Aiii). Inhibitor-treated Csk^AS platelets also spread marginally better than control on fibrinogen (Figure 6Bi-ii). Collectively, these findings validate that Csk is an inhibitor of both ITAM- and integrin receptor–mediated platelet functions. SFK activity (Src p-Tyr418) was significantly increased in resting, CRP (30 μg/mL)–, and anti-CLEC-2 antibody (10 μg/mL)–stimulated Csk^AS platelets treated with 10 μM 3-IB-PP1 (Figure 6Ci-iii,Di-iii; supplemental Figure 10A-B). However, Lyn p-Tyr507, Src p-Tyr529, and Fyn p-Tyr530 were only minimally reduced in the presence of 3-IB-PP1 (Figure 6Cii-iii,Dii-iii; supplemental Figure 10A-B), suggesting that tyrosine phosphatases do not automatically dephosphorylate these residues in the absence of Csk. Syk activity was significantly increased at later time points after CRP stimulation (Figure 6Cii-iii; supplemental Figure 10A). Concomitantly, G6b-B was hyperphosphorylated and had more associated Shp1 and Shp2 in collagen-stimulated Csk^AS platelets treated with 3-IB-PP1 (Figure 6E). Because G6b-B KO platelets display markedly elevated Syk activity,²⁸  we hypothesized that the increased inhibitory signaling provided by G6b-B impeded further enhancement of CRP- and CLEC-2 antibody–mediated aggregation and overall tyrosine phosphorylation in Csk^AS platelets. These findings suggest that SFKs activate inhibitory pathways in parallel with activation pathways to prevent platelet hyperactivation.

Here, we describe a fundamental mechanism controlling SFK activity in the MK lineage involving the kinase Csk and the phosphatase CD148. We show that Csk is a major inhibitor of SFKs in platelets, whereas CD148 primarily activates SFKs but also attenuates SFK activity under undefined conditions (Figures 4B and 7A). These functions of Csk and CD148 correlate with what has previously been described in immune cells.^31,50-55  We show that lack of Csk in the MK lineage in mice results in increased SFK activity in platelets, paradoxical bleeding, and reduced thrombosis due to negative feedback mechanisms, including downregulation of (hemi-)ITAM–containing receptors and concomitant upregulation of the inhibitory ITIM-containing receptor G6b-B, rendering platelets less responsive to vascular injury (Figure 7B). In contrast, deletion of CD148 results in reduced SFK activity and thrombus formation as a consequence of reduced (hemi-)ITAM and integrin receptor signaling. However, bleeding was normal in CD148 KO mice because of residual SFK activity and intact positive feedback mechanisms. Intriguingly, deletion of both Csk and CD148 markedly increased SFK activity, with more pronounced bleeding and thrombotic defects than in Csk KO mice, because of enhanced negative feedback effects (Figure 7B). A summary of the phenotypes of Csk KO, CD148 KO, and DKO mice is provided in Table 1. Thus, sustained high SFK activity in platelets does not culminate in a general increase in platelet reactivity but results in overcompensation of negative feedback mechanisms that attenuate the platelet response to various thrombogenic substrates. Moreover, the reduced platelet count in Csk KO mice was not due to increased platelet clearance; rather, it was a result of a reduction in platelet production, agreeing with a lack of activation and phagocytosis markers on platelets in these mice. Although mean platelet count was rescued in DKO mice, the variability in count, platelet volume, and the proportion of reticulated platelets in the circulation suggested a less robust system controlling platelet homeostasis in the absence of Csk and CD148.

Negative feedback mechanisms activated in the Csk KO and DKO mice, including downregulation of the (hemi-)ITAM–containing receptors GPVI-FcR γ-chain and CLEC-2 and upregulation of the ITIM-containing receptor G6b-B and tyrosine kinase Chk, likely represent cell-intrinsic adaptation of platelets to high SFK activity. These findings are in agreement with the reduction of the ITAM-containing T-cell receptor in Csk-deficient T cells⁵⁶  and increased phosphorylation of the ITIM-containing receptor Sirpα, the lipid phosphatase SHIP1, and Shp1 following inhibition of Csk^AS in immune cells.⁵⁷  A plausible explanation for the downregulation of (hemi-)ITAM–containing receptors is that these receptors are direct substrates of CD148; similarly, the ITAM-containing T-cell receptor subunit CD3 ζ-chain is suggested to be a substrate of the tyrosine phosphatase CD45.^58,59  A clear difference was observed in the downregulation of GPVI-FcR γ-chain and CLEC-2, suggesting differential regulation of these receptors. Whereas the GPVI-FcR γ-chain was severely reduced in both Csk KO and DKO platelets, CLEC-2 was markedly downregulated only in DKO platelets. This may be explained by the different affinities of Syk to different phospho-(hemi-)ITAMs,^60,61  and thus the ability of the SH2 domains of Syk to protect tyrosine residues within these motifs from dephosphorylation by CD148, as proposed for the related kinase ZAP-70.^62-64  We hypothesize that the higher affinity of the tandem SH2 domains of Syk for the dual phosphotyrosine residues of the FcR γ-chain ITAM in Csk KO platelets prevents CD148-mediated dephosphorylation. Sustained phosphorylation would in turn lead to internalization and degradation of the GPVI-FcR γ-chain complex. In contrast, because of the lower binding affinity of individual SH2 domains of Syk for single phosphotyrosine residues with the hemi-ITAM of CLEC-2, Syk is less capable of protecting CLEC-2 from dephosphorylation by CD148. As a consequence, less CLEC-2 is destined for degradation than FcR γ-chain in Csk KO platelets. However, this is not the case for DKO platelets, where no CD148 is present to dephosphorylate CLEC-2; hence, it is markedly downregulated. Further work is needed to validate this hypothesis.

Analysis of tyrosine phosphorylation downstream of GPVI, CLEC-2, and αIIbβ3 revealed that activation of Syk is dependent not only on SFK activity but also on the presence of appropriate docking sites at the plasma membrane, providing evidence that increased SFK activity alone is insufficient to initiate downstream signaling. A growing body of evidence has established G6b-B as a major inhibitor of (hemi-)ITAM–containing receptor signaling in platelets, acting primarily at the level of Syk.^28,65  It is therefore likely that increased G6b-B expression, phosphorylation, and Shp1 and Shp2 association contributes to the attenuation of (hemi-)ITAM signaling in Csk KO and DKO platelets. Upregulation of Chk in the absence of Csk likely also contributes to the attenuation of SFK-mediated signaling in these platelets, albeit less efficiently than its structurally related counterpart, Csk. Although, Chk was previously reported in platelets,⁶⁶  it was not detected in either human or mouse platelets by mass spectrometry,^3,4  in agreement with our findings from WT platelets. Thus, the combination of reduced (hemi-)ITAM–containing receptor expression combined with increased ITIM-containing receptor and Chk expression result in a reduction in platelet reactivity to specific substrates. This is better tolerated than preactivated platelets that can trigger disseminated intravascular coagulation and death.

To circumvent masking effects of negative feedback mechanisms in Csk KO or DKO mice, we used Csk^AS mice. Inhibition of Csk^AS in platelets from these mice resulted in significantly reduced phosphorylation of the C-terminal inhibitory tyrosine residues of SFKs and increased SFK and Syk activity, similar to that reported in immune cells.^39,57,67  However, inhibition of Csk^AS in platelets failed to have an effect on platelet aggregation. This can be partially explained by increased formation of the G6b-B-Shp1-Shp2 complex, counteracting the effect of increased SFK activity on Syk activation.^25,28  Our findings also suggest that Csk plays little role in attenuating GPVI and CLEC-2 signaling once initiated, mainly providing a break prior to ligand engagement and receptor clustering to prevent unwanted signaling by the receptors. However, this was not the case for αIIbβ3-mediated platelet spreading on fibrinogen, which was enhanced either in the absence of Csk or following inhibition of Csk^AS, suggesting Csk differentially regulates integrin and ITAM-containing receptor signaling. Previous work from our group suggests that G6b-B facilitates rather than inhibits integrin-mediated responses in platelets and MKs²⁸ ; thus, upregulation of the G6b-B-Shp1-Shp2 complex is predicted to enhance rather than attenuate platelet spreading, as observed.

In addition to filling a major gap in our knowledge of how SFKs are regulated in the MK lineage, findings from this study are also clinically relevant, demonstrating for the first time that either increased or decreased platelet SFK activity can lead to reduced thrombus formation and bleeding. Our study highlights, that loss of an inhibitor of platelet activation, such as Csk, can lead to paradoxical bleeding due to the robust negative feedback mechanisms that get activated. This may provide important mechanistic insights into the phenomenon of early platelet dysfunction observed in severe trauma patients displaying “exhausted” platelets,⁶⁸  showing reduced thrombus formation on collagen.⁶⁹  Our findings suggest that platelets in the circulation can undergo cell-intrinsic changes, leading to downregulation of activation pathways and upregulation of inhibitory mechanisms, which may contribute to platelet dysfunction in trauma patients. Moreover, certain tyrosine kinase inhibitors used to treat cancer patients inhibit SFKs or Csk and are accompanied by bleeding side effects. Accordingly, the increased bleeding risk associated with use of the Bcr-Abl inhibitor dasatinib is ascribed to its off-target effects on platelet SFKs.^70,71  Similarly, bleeding side effects of the Btk inhibitor ibrutinib may be partially explained by its potent off-target effect on Csk.^72,73  Based on our findings, CD148 is an attractive antithrombotic drug target, and its inhibition should attenuate, but not completely block, the main platelet activation pathways and have minimal bleeding side effects. Tyrosine phosphatases are increasingly considered to be druggable,⁷⁴  which is supported by the recent development of the highly specific Shp2 inhibitor SHP099⁷⁵ ; thus, targeting CD148 is not inconceivable.

The authors thank Timo Vögtle for critical reading of the manuscript and invaluable discussions, Quadratech Diagnostics for providing the T-TAS device, and all members of the Birmingham Biomedical Sciences Unit for maintenance of mouse colonies.

J.M. and Z.N. are British Heart Foundation (BHF) postdoctoral research associates. A.M. is a BHF Intermediate Basic Science Research Fellow (FS/15/58/31784), and Y.A.S. is a BHF Senior Basic Science Research Fellow (FS/13/1/29894). R.A.G. was supported by the government of the Sultanate of Oman. This work was supported by BHF project grant PG/11/108/29237 and program grant RG/15/13/31673.

Contribution: J.M., Z.N., A.M., and Y.A.S. contributed to conceptualization; J.M., Z.N., J.W.M.H., A.M., and Y.A.S. contributed to methodology; J.M., Z.N., C.W.S., R.A.G., J.P.v.G., B.M.E.T., G.E.J., A.M., and Y.A.S. contributed to formal analysis; J.M., Z.N., G.D.N., C.W.S., M.J.G., R.A.G., J.P.v.G., S.H., L.B., J.N.C., L.T., M.J.E.K., P.H., J.W.M.H., A.M., and Y.A.S. contributed to investigation; Y.A.S. contributed resources; A.T. and A.W. contributed mouse models and reagents; P.H. contributed equipment and expertise; J.M., Z.N., A.M., and Y.A.S. wrote the manuscript; J.M., Z.N., J.W.M.H., G.E.J., A.W., A.M., and Y.A.S. revised and edited the manuscript; and A.M. and Y.A.S. supervised the study.

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

Correspondence: Yotis A. Senis, Institute of Cardiovascular Sciences, College of Medical and Dental Sciences, University of Birmingham, Birmingham B15 2TT, United Kingdom; e-mail: y.senis@bham.ac.uk.