LRRC8 complexes are ATP release channels that regulate platelet activation and arterial thrombosis Open Access

Platelets are essential for hemostasis and thrombosis.^1-3 Upon activation, platelets change shape through cytoskeletal rearrangements,⁴ followed by the secretion of adenosine triphosphate (ATP), adenosine diphosphate (ADP), serotonin, and thromboxane A2. Secreted platelet content activate their respective G protein–coupled receptors in autocrine/paracrine manners to promote thrombus formation.^5-7 Although platelet shape change occurs early in the activation process and is considered a requirement for aggregation,⁴ the molecular mechanisms responsible for sensing and transducing these changes are unknown.

Cytosolic calcium levels contribute to platelet activation and are increased by calcium release from intracellular stores, as well as influx across the plasma membrane.^8,9 Intracellular calcium release occurs downstream of phospholipase C, which can be activated by G protein–coupled receptors, such as the ADP receptor P2Y₁, resulting in calcium mobilization. In addition, ADP activates P2Y₁₂ receptors, which subsequently results in phosphoinositide 3-kinase (PI3K)–protein kinase B (AKT) signaling.^10,11 Both P2Y₁ and P2Y₁₂ activation results in activation of αIIbb3 integrin.¹² A major pathway of calcium entry across the plasma membrane is mediated by both STIM1-Orai1^13,14 and P2X1.¹⁵ P2X1 is an ATP-gated, calcium-permeant ion channel which, when activated by ATP released from damaged endothelium and/or activated platelets, increases intracellular calcium to amplify platelet activation.^16,17 

Ion channels are ubiquitous proteins that mediate ionic currents in almost all cells, including platelets. Various types are implicated in platelet function, including ligand-gated cation channels, voltage- and calcium-gated potassium channels, transient receptor potential channels,^18-20 and transmembrane protein 16F (TMEM16F) channels. TMEM16F channels are lipid scramblases, which, when deficient, are responsible for Scott syndrome, a rare bleeding disorder.^21,22 In addition, mechanosensitive and mechanoresponsive ion channels, which translate mechanical forces into electrical signaling across the lipid bilayer, such as Piezo1, have been recently described to regulate platelet function.²³ Leucine-rich repeat–containing protein 8A (LRRC8A; also known as SWELL1) is the essential subunit of the heterohexameric LRRC8 volume-regulated anion channel (VRAC),^24,25 which contain at least 1 LRRC8A subunit in combination with LRRC8B, LRRC8C, LRRC8D, and/or LRRC8E subunits. LRRC8 channels are mechanoresponsive anion channels expressed and functional in many cell types, including adipocytes,²⁶ endothelial cells,²⁷ pancreatic β-cells,^28-30 skeletal muscle cells,^31,32 immune cells,^33-35 astrocytes,^36-39 microglia,^40-42 and neurons.^43-46 However, it remains unknown whether LRRC8 channels are expressed and functional in platelets and contribute to arterial thrombosis.

In this study, we first identify a genetic signal for LRRC8 proteins as important regulators of platelet volume and function in humans. We show that LRRC8A-D are expressed in murine and human platelets, form functional ATP-permeant VRAC currents in megakaryocytes (MKs), and contribute to platelet activation, adhesion, and aggregation, as well as arterial thrombosis in vivo. Mechanistically, the LRRC8 complex mediates cytosolic ATP release to amplify agonist-stimulated calcium influx and PI3K-AKT signaling. In addition, we highlight the opportunity for targeting LRRC8 channels by small-molecule LRRC8 inhibitors as a novel anti-thrombotic strategy.

All studies involving animal experimentation were performed in accordance with the recommendations in the guide for the care and use of laboratory animals of the National Institutes of Health. All animals were handled according to the approved institutional animal care and use committee protocols of Washington University in St. Louis. Experiments using CD34⁺ cell–derived MKs were approved by the institutional review boards of both The University of Utah and Washington University in St. Louis. Experiments using human platelets isolated from whole blood were performed in accordance with the institutional review board of Washington University in St. Louis from consenting healthy adults.

Detailed methods for all experimental procedures can be found under supplemental Methods (available on the Blood website).

Human LRRC8A-E are encoded by 5 genes on 3 chromosomes: LRRC8A on human chromosome 9q34.11; LRRC8B, LRRC8C, and LRRC8D in tandem on chromosome 1p22.2; and LRRC8E on chromosome 19p13.2.⁴⁷ We searched the National Health Genome Research Institute, European Bioinformatics Institute Catalog of genome-wide association studies (ebi.ac.uk/gwas/home) for single-nucleotide polymorphisms (SNPs) and identified several significant associations between mean platelet volume (MPV) and SNP (both intronic and exonic) variants assigned to LRRC8 genes, with the 3 exonic SNPs listed in Table 1. SNP variants assigned to LRRC8A and LRRC8C are associated with increased MPV, and the variant assigned to LRRC8D is associated with decreased MPV. To determine the relative strength of the association of the MPV phenotype with each variant, we conducted a Phenome-Wide Association Study search for phenotypes associated with each SNP variant and found that MPV was the phenotype most closely associated with all 3 variants, with P values from 3.7 × 10⁻¹² to 2.7 × 10⁻³⁴ (Table 1).

Because LRRC8 channels are heterohexamers comprising different combinations of 5 LRRC8 subunits, and MPV is the most closely associated phenotype with exonic SNPs in genes assigned to 3 of these subunits located on 2 different chromosomes, we interpret these human genetic data as evidence that LRRC8 channels are involved in regulating platelet volume and function.

To determine whether LRRC8A-E transcripts are expressed in human platelets, we searched a publicly available plateletomics database (plateletomics.com, using data from Rowley et al⁴⁸ and Simon et al⁴⁹), which details both human and murine platelet transcriptomics. LRRC8A, LRRC8B, LRRC8C, and LRRC8D transcripts are all expressed in platelets (top 75th, 13th, 42nd, and 5th percentile of 5911 transcripts, respectively; supplemental Figure 1A). LRRC8E was not detected. Absolute LRRC8 transcript expression levels in human platelets revealed LRRC8D > LRRC8B > LRRC8C > LRRC8A (Figure 1A), whereas in murine platelets Lrrc8c > Lrrc8b > Lrrc8d > Lrrc8a (Figure 1D). In human and mouse platelets, LRRC8 messenger RNA (mRNA) expression levels were substantially higher than mechanosensitive PIEZO1, PIEZO2, and other mechanoresponsive ion channels (TRPV4), and comparable with highly expressed TMEM16F (Ano6), important for platelet lipid scrambling and causal in Scott syndrome (Figure 1B,E).²² Consistent with mRNA expression data, LRRC8A-D proteins were expressed in both human (Figure 1C) and murine (Figure 1F) platelets, whereas LRRC8E was not.

Multiple regression analyses of human platelet aggregation in response to stimulation with ADP, protease-activated receptor 1 agonist peptide (PAR1-AP), PAR4-AP, or arachidonic acid, adjusting for mRNA level, age, sex, race, body mass index, and platelet count identified a complex relationship between LRRC8 subunit mRNA levels and agonist-induced platelet aggregation. LRRC8A mRNA levels are significantly associated with enhanced ADP- and PAR4-AP–stimulated aggregation; LRRC8B mRNA levels with reduced ADP- and PAR1-AP–stimulated aggregation; and LRRC8D mRNA levels with reduced PAR1-AP–induced aggregation (supplemental Figure 1B). Overall, these data provide correlative evidence to suggest a modulatory role for VRAC and LRRC8 proteins in regulating platelet function in humans.

To explore the biological mechanisms implied by the human genetic studies, we generated MK-specific LRRC8A conditional knockout (cKO) mice by crossing Lrrc8a^fl/fl mice (Swell1^fl/fl)^26-28,31,50 with platelet factor 4-Cre (Pf4-Cre) mice to generate Lrrc8a^{fl/fl;Pf4-cre+} (cKO) and littermate control Lrrc8a^fl/fl mice (wild type [WT]). LRRC8A protein is completely deleted in cKO mouse platelets while maintaining expression of integrin b3 and P-selectin (Figure 1G-H). Consistent with increased MPV associated with the SNP assigned to LRRC8A (Table 1), platelet-targeted LRRC8A deletion also increases MPV compared with both WT and platelet-targeted heterozygous controls (cHet; Lrrc8a^{fl/+;Pf4-cre+}; Figure 1I,K), without affecting platelet counts (Figure 1J), or other hematological parameters (supplemental Figure 2). Thus, LRRC8A deletion in mouse platelets increases MPV, which phenocopies human genetic data (Table 1) implicating LRRC8 proteins as regulators of platelet volume.

To confirm that LRRC8 proteins form functional VRAC channels in platelets, we measured VRAC in freshly isolated murine MKs under isotonic conditions and with hypotonic swelling (Figure 1L). The cell capacitance of LRRC8A cKO MKs (176 ± 17.9 picofarad [pF]; n = 10) are notably larger than WT MKs (121 ± 16.6 pF; n = 14; P < .05; Figure 1M). WT MKs develop robust VRAC in response to hypotonic swelling, which is completely inhibited by 10 μM SN-401 (4-(2-butyl-6,7-dichloro-2-cyclopentylindan-1-on-5-yl)oxybutyric acid [DCPIB]) and abolished upon LRRC8A deletion (Figure 1N-Q), confirming LRRC8A as mediating this volume-sensitive/mechanoresponsive current. Similarly, robust VRAC currents are elicited in response to hypotonic swelling in the human megakaryoblast MEG-01 cell line, fully inhibited by 10 μM SN-401 (supplemental Figure 3), and significantly diminished by short hairpin RNA–mediated LRRC8A knockdown (KD; Ad-shLRRC8A-mCherry) as compared with control (Ad-shSCR-mCherry; supplemental Figure 3).

LRRC8A depletion markedly impairs platelet adhesion to fibrillar collagen-coated surfaces under both low and high shear conditions (650 s⁻¹, Figure 2A; 2600 s⁻¹, Figure 2B; supplemental Videos 1 and 2), and reduces thrombin-stimulated platelet activation as assessed by P-selectin exposure, a measure of α-granule release⁵¹ (Figure 2C) and binding of JonA to activated αIIbb3 integrin (Figure 2D), whereas collagen-related peptide (CRP)–stimulated platelet activation shows impairments in P-selectin exposure but not activated αIIbb3 integrin (Figure 2C-D). Interestingly, application of Ca²⁺ ionophore A23187 preserves P-selectin exposure (Figure 2C) and largely rescues activated αIIbb3 integrin (Figure 2D) in LRRC8A-null platelets. To investigate the function of LRRC8A in human cells, we used CD34⁺ cell–derived human MKs isolated from cord blood⁵² (supplemental Figure 4A). Dual-guide RNAs targeting LRRC8A in these MKs significantly depleted LRRC8A expression as compared with cells treated with nontargeting control guide RNA (supplemental Figure 4B). In LRRC8A-deficient MKs, thrombin-stimulated P-selectin exposure is significantly reduced (supplemental Figure 4C). Similarly, binding of PAC-1, a pentameric immunoglobulin M, which binds to activated αIIbb3 integrin,⁵³ is also significantly decreased (supplemental Figure 4D). Platelet-specific LRRC8A depletion impairs platelet aggregation in response to multiple agonists, including thrombin (Figure 2E), ADP (Figure 2F), CRP (Figure 2G), and a thromboxane A2 analog (U46619; Figure 2H). As in the platelet activation results (as measured by P-selectin exposure), aggregation in response to A23187 is preserved (Figure 2I), suggesting a signaling defect upstream of cytosolic Ca²⁺. ATP secretion in LRRC8A-null platelets is markedly impaired compared with WT in response to multiple agonists (Figure 2J-M), despite only a mild reduction in total ATP content under resting conditions (Figure 2N).

Because the functional defects observed in LRRC8A-null platelets in response to various agonists suggest impairments in common signaling pathways, we measured cytosolic Ca²⁺ levels in response to thrombin (Figure 2O-Q), and PI3K-AKT signaling after PAR4 stimulation (Figure 2R-S) in WT and LRRC8A-null platelets. LRRC8A deletion lowers peak cytosolic Ca²⁺ levels (Figure 2O-P) and rate of rise of intracellular Ca²⁺ (Figure 2Q) in thrombin-stimulated platelets, indicating impaired Ca²⁺ signaling. This is also associated with marked reductions in PI3K-AKT signaling in LRRC8A-null platelets stimulated with PAR4-AP (Figure 2R-S), and in MEG-01 cells after short hairpin RNA–mediated LRRC8A KD followed by thrombin stimulation (supplemental Figure 6). Collectively, these data reveal that LRRC8A promotes platelet adhesion, activation, and aggregation, potentially by facilitating Ca²⁺ and PI3K-AKT signaling pathways.

Because LRRC8 complexes are anion channels, we asked how loss of anion conductance might impair Ca²⁺ signaling (Figure 2O-Q). Based on our observations that agonist-induced ATP release is abrogated in LRRC8A-null platelets (Figure 2J-M), and the finding that LRRC8A can form an ATP release channel,^40,54,55 we hypothesized that platelet LRRC8 channels contribute to ATP efflux to amplify platelet function via P2X1 receptor signaling.^16,17,56 To test this hypothesis, we directly measured swell-activated ATP currents in WT MKs in which all intracellular anions were replaced with either 1 or 50 mM ATP⁴⁻ (Figure 3A-G). MKs dialyzed with 50 mM ATP⁴⁻ as the only permeant anion yield robust SN-401 (DCPIB)–sensitive swell-activated inward ATP currents, indicative of ATP efflux, 4.4-fold larger than cells dialyzed with 1 mM ATP⁴⁻, and these ATP currents are abolished in LRRC8A-null MKs (Figure 3C,F,G), demonstrating LRRC8A-dependent ATP currents in response to MK volume changes. Similar results were obtained using MEG-01 cells (Figure 3H-J), including Ad-shSCR– and Ad-shLRRC8A–transduced MEG-01 cells (Figure 3K-N). These data demonstrate that platelet LRRC8 channel complexes permeate cytosolic ATP and thus may contribute to ATP efflux alongside ATP secreted from dense granules.^57,58 Therefore, markedly reduced ATP release observed in agonist-stimulated LRRC8A-null platelets (Figure 2J-M) may reflect loss of LRRC8 channel–mediated ATP efflux.

To determine the relative contribution of LRRC8 channel–mediated ATP efflux vs dense granule ATP secretion from platelets, we compared platelet aggregation, platelet activation (P-selectin exposure), ATP release, and CD63 exposure (a dense granule and lysosomal marker^57,59-63) in response to various concentrations of thrombin stimulation (0.02-0.08 U/mL). At 0.02 U/mL thrombin, LRRC8 channel ablation reduces platelet aggregation by 53% (Figure 4A), α-granule secretion by 30% (Figure 4B), and ATP release by 86% (Figure 4C), with only a 51% reduction in dense granule secretion based on CD63 exposure (Figure 4D). Increasing thrombin stimulation to 0.06 U/mL overcomes platelet aggregation, α-granule, and dense granule release defects in LRRC8A-null platelets (Figure 4E,F,H) but preserves a marked 64% reduction in ATP release (Figure 4G). Further stimulating LRRC8A-null platelets with 0.08 U/mL thrombin, similarly, saturates platelet aggregation (Figure 4I), α-granule secretion (Figure 4J,O), and CD63-derived dense granule release (Figure 4L,P), normalizing these parameters relative to WT platelets, while preserving a robust 58% reduction in peak ATP efflux (Figure 4K,N), suggesting that almost 60% of platelet ATP release in response to 0.08 U/mL thrombin arises from ATP efflux through LRRC8 channels.

Although CD63 is a marker of dense granule secretion, it is also expressed on lysosomes.^57,59,60 As a complimentary approach to quantify dense granule release under these stimulatory conditions, we measured serotonin release in thrombin-stimulated WT and LRRC8A-null platelets, as serotonin is uniquely present in dense granules.^57,59 Using serotonin as a measure of dense granule secretion, the 60% reduction in ATP release observed in LRRC8A-null platelets stimulated with 0.08 U/mL thrombin (Figure 4K) is associated with a mild 16% reduction in dense granule secretion (Figure 4M,Q), which is consistent with the CD63 exposure results (Figure 4H,L for 0.06 and 0.08 U/mL). In addition, we calculated the ratio of ATP to serotonin in thrombin-stimulated WT and LRRC8A KO platelets and observed that ATP release in LRRC8A KO platelets is significantly lower than WT when normalized to dense granule release (Figure 4R), which is consistent with reduced LRRC8A-dependent ATP efflux. Finally, we compared the percentage reduction in ATP efflux with the percentage reduction in serotonin release at 0.08 U/mL thrombin in LRRC8A KO vs WT platelets and found these differences to be robustly statistically significant (Figure 4S).

Given that the marked ATP release defect in LRRC8-null platelets is not attributed to a dense granule secretion defect, these data support the conclusion that 44% to 60% of ATP released from activated platelets arises from efflux of cytosolic ATP through LRRC8 channels. When assessing trends in ATP release (Figure 4N), α-granule secretion (Figure 4O), and dense granule secretion (Figure 4P-Q) over a range of thrombin concentrations in WT platelets, there is a graded increase in ATP release with increasing thrombin concentrations whereas both α-granule and dense granule secretion begin to saturate at high thrombin concentrations. This is consistent with the fact that platelets contain only 3 to 8 dense granules per platelet,⁵⁷ and, thus, significant amounts of ATP are released from the cytoplasm once dense granule secretion saturates, especially at higher agonist doses.

Having demonstrated that LRRC8 releases ATP release during platelet activation, we next asked whether adding exogenous ATP can rescue defective agonist-induced aggregation observed in LRRC8A-null platelets. Addition of 40 μM ATP fully rescues the aggregation defect in LRRC8A-null platelets (Figure 4T). Overall, these results suggest that most agonist-induced ATP release from platelets is through LRRC8 complexes (rather than through dense granule release), which then activates the calcium-permeable ATP-gated P2X1 receptor¹⁶ to drive platelet activation and aggregation.

To determine the pathophysiological function of platelet LRRC8A in thrombosis, we used WT and LRRC8A cKO mice in mouse models of laser-induced cremaster arteriolar thrombosis and FeCl₃-induced carotid arterial thrombosis. Platelet-specific LRRC8A deletion significantly reduces platelet thrombus formation compared with WT mice after laser-induced cremaster arteriolar injury (Figure 5A-C; supplemental Videos 3 and 4). Consistent with this finding, time to occlusion in a FeCl₃-injured carotid artery is markedly prolonged in cKO mice (Figure 5D-E). Notably, 5 of 7 cKO mice reached the maximum time to occlusion (30 minutes). Importantly, the more prominent thrombosis impairments observed in cKO mice in the FeCl₃-induced carotid arterial thrombosis model may be because of different shear rates, differences in the caliber of the cremaster arterioles vs carotid arteries, or both. Despite reduced platelet thrombus formation in cKO mice, no differences were observed in tail bleeding times and blood loss observed at the site of tail tip amputation (Figure 5F-G). These results indicate that platelet LRRC8A deletion impairs arterial thrombosis, without prolonging bleeding times in mice.

To investigate the therapeutic potential of targeting the LRRC8 channel, we examined whether established small-molecule LRRC8 complex inhibitors inhibit platelet function. SN-401 (DCPIB) and congener SN-406 (Figure 6A) are VRAC inhibitors⁵⁰ with 50% inhibitory concentration (IC₅₀) of 1.5 μM and 1 μM, respectively, in MEG-01 cells (supplemental Figure 9). As with SN-401 (Figure 3H-J), SN-406 robustly inhibits swell-activated ATP efflux in MEG-01 cells (Figure 6B-D). Consistent with these effects on LRRC8-mediated ATP currents in patch-clamp studies, both significantly inhibit thrombin-stimulated mouse platelet aggregation (Figure 6E) and reduce total ATP release (Figure 6F-G) and ATP efflux rate (Figure 6H), as compared with vehicle. In addition, SN-406 reduces agonist-stimulated Ca²⁺ flux (Figure 6I), consistent with a mechanism of impaired ATP-stimulated Ca²⁺ signaling in mouse platelets and suppresses platelet aggregation (Figure 6J) and Ca²⁺ signaling (Figure 6K) in human platelets. SN-406 exhibits a clear dose-response relationship in PAR4-AP–induced mouse platelet aggregation (Figure 6L), yielding an IC₅₀ of 1.1 μM (Figure 6M), which is comparable with the 1 μM IC₅₀ for VRAC inhibition in MEG-01 cells (supplemental Figure 9). Applying 1.1 μM SN-406 to WT and LRRC8A-null platelets shows significant suppression of platelet aggregation in WT platelets with no effect on LRRC8A-null platelets, consistent with SN-406-LRRC8 complex on-target activity (Figure 6N). In addition, we investigated SN-406 effects on aggregation, granule release, and ATP secretion when stimulated with 0.04 U/mL thrombin (supplemental Figure 11). Comparable with results obtained for WT and LRRC8A-null platelets treated with 0.06 and 0.08 U/mL thrombin (Figure 4), SN-406 inhibited ATP release by 36% (supplemental Figure 11C), without reducing aggregation (supplemental Figure 11A), α-granule secretion (supplemental Figure 11B), or dense granule secretion (supplemental Figure 11D).

Next, we tested the VRAC inhibitor dicoumarol (Figure 6O), which is structurally distinct from the SN-40X compounds optimized off the SN-401 (DCPIB; Figure 6A) backbone, and has been recently shown to inhibit microglial LRRC8A-mediated ATP release.⁴⁰ We find dicoumarol also inhibits swell-induced ATP efflux in MEG-01 cells (Figure 6O-Q), and markedly suppresses platelet aggregation, which is rescued by exogenous ATP (Figure 6R).

Through further compound optimization, we developed SN-89B, a potent LRRC8 inhibitor, with lower predicted plasma membrane protein binding than those previously described. SN-89B strongly inhibits ATP efflux (Figure 7A-B; supplemental Figure 12) elicited in response to hypotonic swelling in MEG-01 cells. In agonist-stimulated human (Figure 7C-D) and mouse (supplemental Figure 13A-B) platelets, activation is fully suppressed by SN-89B as assessed by P-selectin exposure, and PAC-1/JonA binding to activated αIIbb3 integrin. SN-89B also markedly impairs adhesion of human platelets to von Willebrand factor–coated surfaces under shear (1800 s⁻¹) conditions (Figure 7E). In human platelets, SN-89B markedly inhibits agonist-induced aggregation (Figure 7F) in a dose-dependent manner (supplemental Figure 13C), with an IC₅₀ of 0.84 μM (supplemental Figure 13D). Finally, in human platelets, SN-89B inhibits thrombin-induced increases in cytosolic Ca²⁺ levels (Figure 7G-H), and PI3K-AKT signaling after PAR1 + PAR4 stimulation (Figure 7I-J). To study SN-89B in vivo, we treated WT mice with vehicle or 50 mg/kg SN-89B intraperitoneally for up to 10 days, followed by laser-induced cremaster arteriolar thrombosis and FeCl₃-induced carotid arterial thrombosis. We also measured bleeding times and blood loss in response to tail tip amputation. Ticagrelor, a clinically used P2Y₁₂ receptor antagonist, was used as a positive control. Treatment with SN-89B reduces platelet thrombus formation at 30, 180, and 240 seconds after laser-induced cremaster arteriolar injury (Figure 7K-M; supplemental Videos 5 and 6), and prolongs FeCl₃-induced carotid arterial thrombosis (Figure 7N), without increasing tail bleeding parameters tested, in contrast to the ticagrelor positive control (Figure 7O-Q).

Taken together, these data demonstrate that LRRC8 channels are ATP release channels that promote Ca²⁺ mobilization and PI3K-AKT signaling, facilitating platelet activation, adhesion, and aggregation, and in vivo thrombosis, in addition to demonstrating the pharmacological tractability of targeting LRRC8 channels as a novel antithrombotic strategy.

Arterial thrombotic diseases are the leading cause of death in the United States. Platelets are the first blood cells recruited to the site of vascular injury. Adherent platelets undergo shape and volume changes early in platelet activation, which subsequently facilitates their aggregation.⁴ However, the molecular mechanisms responsible for sensing and transducing these changes in platelet volume are unknown. Because our initial findings indicated LRRC8 proteins are highly expressed in human and mouse platelets, and changes in LRRC8 proteins are associated with altered function, we hypothesized that LRRC8 channel complexes function as mechanoresponsive sensors linking platelet shape change to platelet activation. Collectively, our data from selective LRRC8A deletion in platelets, and CRISPR-edited human MKs, reveal significant impairments in cell activation, adhesion, aggregation, and ATP release in response to multiple agonists, suggesting an LRRC8A-dependent defect in common signaling pathways.

These observations led us to investigate how activation of the LRRC8 anion channel might promote both cytosolic Ca²⁺ mobilization and PI3K-AKT signaling. Because the LRRC8 complex is an anion channel that also permeates larger anions such as ATP^54,55,64,65 and has recently been shown to form an ATP release channel in microglia,⁴⁰ we hypothesized that platelet shape change activates LRRC8-mediated ATP efflux, which subsequently activates ATP-gated P2X1 receptor^16,17,56 Ca²⁺ influx in a paracrine/autocrine manner to amplify platelet activation and aggregation. In line with this mechanism, Tomasiak et al showed that thrombin-induced ATP secretion from porcine platelets increases under hypoosmotic stimulation conditions expected to activate LRRC8 channels.⁶⁶ We find LRRC8A-dependent swell-activated ATP efflux can be directly measured in MKs in patch-clamp studies, under conditions in which intracellular anions are replaced with 50 mM ATP as the only permeant anion. Because the primary source of ATP release during platelet activation is thought to be secretion from dense granules,^67-69 as opposed to channel-mediated efflux from the platelet cytosol, we measured ATP release in response to various concentrations of thrombin and compared this to 2 readouts of dense granule secretion: CD63 exposure and serotonin secretion. High concentrations of thrombin rapidly saturated platelet aggregation, activation, and dense granule secretion in both WT and LRRC8A-null platelets based on both CD63 and serotonin release, but, despite this, a 60% reduction in peak ATP release persisted in the LRRC8A-null platelets. Importantly, this large reduction in ATP secretion is not fully explained by the mild reduction observed in basal intracellular ATP levels in LRRC8A-null platelets. These data indicate that at least half of the ATP release occurring during platelet activation arises from cytosolic ATP efflux via LRRC8 channels. This model is consistent with the burst in cytosolic ATP that occurs from aerobic glycolysis induced during platelet activation and aggregation, which was previously thought to occur to satisfy platelet energetic demands.⁷⁰ Thus, the graded increase in ATP release in response to increasing thrombin concentrations most likely reflects increases in de novo ATP synthesis from aerobic glycolysis and subsequent efflux through LRRC8 channels. We postulate that the spike of ATP synthesized by aerobic glycolysis provides the source of the ATP-concentration gradient required for LRRC8-mediated ATP efflux and provides a link between platelet metabolism and activation, via paracrine and autocrine stimulation of platelet P2X1. Indeed, impaired aggregation in both LRRC8A KO platelets and dicoumarol-inhibited platelets can be effectively rescued by application of exogenous ATP, further supporting this proposed mechanism. In addition to providing an ATP efflux pathway, it is also possible that LRRC8 proteins may also regulate ATP generation by aerobic glycolysis, and this is the focus of future studies.

Also consistent with the proposed mechanism, P2X1-receptor KO mice phenocopy platelet-targeted LRRC8A cKO mice with respect to platelet Ca²⁺ influx, adhesion, aggregation and thrombus formation,⁷¹ with relatively preserved bleeding times in both P2X1 mice⁷¹ and LRRC8A cKO mice. P2Y₁ KO mice also exhibit impaired ADP-induced aggregation and impaired thrombosis, but with 70% to 240% increases in mean bleeding times compared to respective controls,^72,73 whereas P2Y₁₂ KO mice exhibit severely prolonged mean bleeding times, 700% to 800% greater than controls.^74,75 These comparatively smaller increases in bleeding time in P2X1 and P2Y₁ KO mice are potentially due to decreased cytosolic calcium signaling, with relatively preserved PI3K-AKT signaling, whereas P2Y₁₂ ablation fully abrogates PI3K-AKT signaling required for stable aggregation. Thus, pharmacological inhibition of LRRC8-mediated ATP efflux may provide a therapeutic strategy to tune P2X1 receptor activation to dampen platelet aggregation and thrombosis, while minimally prolonging bleeding times.

In line with this concept, VRAC inhibitors SN-401 (DCPIB) and SN-406⁵⁰ strongly inhibit agonist-stimulated platelet aggregation, ATP release, and Ca²⁺ flux in murine and human platelets. Furthermore, SN-89B, a more potent SN-401 derivative with reduced predicted protein binding, strongly inhibits platelet activation and aggregation, Ca²⁺ signaling, and PI3K-AKT signaling in vitro, and exerts antithrombotic properties in vivo with no significant effects on bleeding, providing a proof of concept for developing small-molecule LRRC8 inhibitors as novel antiplatelet agents. Importantly, as SN-401 and SN-406 have been previously demonstrated to improve glycemic control and insulin sensitivity in type 2 diabetes mouse models,⁵⁰ this class of molecules possess antiplatelet and antidiabetic properties, which are a desirable combination when treating patients with both diabetes and cardiovascular disease.

As further evidence of LRRC8 channel inhibition as a mechanism of action for antiplatelet activity, we tested dicoumarol, a newly identified VRAC inhibitor⁴⁰ that is structurally distinct from SN-40X compounds. As a 4-hydroxycoumarin anticoagulant, dicoumarol inhibits vitamin K–dependent clotting factor synthesis resulting in impaired coagulation due to reduced clotting cascade efficacy.⁷⁶ As Chu et al recently demonstrated dicoumarol-mediated blockade of LRRC8-mediated ATP release in HeLa cells and microglia,⁴⁰ and our findings linking ATP efflux through LRRC8 channels during platelet activation, we speculated that dicoumarol would also exhibit antiplatelet activity. Our results suggest that, alongside dicoumarol’s known effects on vitamin K–dependent clotting factors, it also exhibits direct antiplatelet activity via inhibition of LRRC8 channel–mediated ATP efflux.

In summary, we identify a genetic signal for a modulatory role for LRRC8 proteins in platelet function in humans and validated this using platelet-targeted LRRC8A KO mice. We demonstrate that the platelet LRRC8 complex forms a novel mechanoresponsive ATP release channel that contributes to arterial thrombosis by providing an ATP efflux pathway for the burst of cytosolic ATP generated upon agonist-stimulation to drive autocrine/paracrine P2X1-mediated Ca²⁺ and PI3K-AKT signaling. In this way, LRRC8 channels function to amplify agonist-stimulated platelet activation and aggregation. Pharmacological modulation of this LRRC8-ATP efflux amplifier may effectively dampen thrombosis without significantly affecting bleeding, highlighting the opportunity to develop LRRC8 targeted antithrombotic agents.

The authors thank the High-Throughput Screening Center at Washington University in St. Louis for use of their FlexStation 3 apparatus, and Thomas Jenstch for kindly sharing LRRC8B, LRRC8D, and LRRC8E rabbit polyclonal antibodies.

This work was supported by the National Institutes of Health (NIH)/National Institute of Diabetes and Digestive Kidney Diseases grants (R01 DK106009, R01 DK126068, R01 DK127080 [R.S.]; and R44 DK121598 and R44 DK126600 [D.J.L.]); the NIH/National Heart, Lung, and Blood Institute grants (R01 HL168600 and I01 BX005072 [R.S.]; R44 HL169181-01A1 [D.J.L.]; R01 HL146559, R01 HL148280, and R01 HL28731 [J.C.]; R01 HL139825 and R61 HL141794 [J.D.P.]; and R01 HL160808 and R01 HL163019 [R.A.C.]); and the American Society of Hematology Graduate Student Award (A.A.).

Contribution: R.S., D.J.L., J.D.T., R.T.M., and A.K. conceptualized the research; J.D.T., R.T.M., A.K., G.B., T.M.A.E.-A., Y.Z., P.A., K.A., V.J., N.A., J.C., D.J.L. N.O.S., and R.S. performed formal analysis; J.D.T., R.T.M., A.K., G.B., D.J.L., N.O.S., T.M.A.E.-A., Y.Z., P.A., K.A., V.J., N.A., C.M., A.A., L.X., J.H., H.Z., T.K., A.L., A.B., D.B., V.K.N., N.M.L., and Y.F. performed investigation; R.S., J.D.T., and R.T.M. wrote the original manuscript draft; R.S., J.D.T., R.T.M., G.B., J.C., and D.J.L. reviewed and edited the manuscript; R.S., J.C., R.A.C., J.D.P., and D.J.L. acquired funding; J.D.T., A.K., G.B., and D.J.L. were responsible for visualization; R.S., J.C., R.A.C., J.D.P., and D.J.L. provided resources; and R.S., J.C., R.A.C., and J.D.P. supervised the study.

Conflict-of-interest disclosure: R.S. is cofounder of Senseion Therapeutics, Inc, a start-up company developing LRRC8A (SWELL1) modulators for human disease. D.J.L. is cofounder and chief executive officer of Senseion Therapeutics, Inc. The remaining authors declare no competing financial interests.

Correspondence: Rajan Sah, Washington University School of Medicine, 425 S Euclid Ave, St. Louis, MO 63110; email: rajan.sah@wustl.edu; and Jaehyung Cho, Division of Hematology, Department of Medicine, Washington University School of Medicine, 425 S Euclid Ave, St. Louis, MO 63110; email: jaehyung.cho@wustl.edu.