CD39 polymorphism enables lung thrombosis in sickle cell disease

Abstract

Sickle cell disease (SCD) is the most common monogenic-hemolytic disorder affecting people of African ancestry. Adenosine diphosphate (ADP) released following intravascular hemolysis activates platelets by stimulating purinergic receptors to promote thrombosis. Despite brisk intravascular hemolysis, which releases high levels of ADP into plasma, and evidence of platelet and hemostatic activation, it remains elusive why only a subset of SCD patients develop lung thrombosis. Using real-time in vivo lung microscopy, we report a surprising finding that humanized SCD mice are protected from ADP-induced lung thrombosis, which is secondary to the degradation of ADP by CD39 present in circulating extracellular vesicles released by the lung endothelium. ADP-induced platelet aggregation is also impaired in the blood of SCD patients with elevated levels of CD39⁺ extracellular vesicles. CD39 polymorphism rs3176891A→G is associated with the incidence of lung thrombosis in SCD patients but not healthy humans of African ancestry. Remarkably, CD39⁺ extracellular vesicles are fewer and ADP-induced platelet aggregation is higher in the blood of SCD patients with rs3176891G allele. This study identifies a novel extracellular vesicle-dependent mechanism preventing lung thrombosis in SCD and reveals how CD39 polymorphisms may impair this protection to increase the risk for lung thrombosis in a subset of SCD patients.

It remains unknown why only some sickle cell disease (SCD) patients develop lung thrombosis. Here, the authors show that an extracellular vesicle-dependent mechanism prevents lung thrombosis in SCD and how a CD39 polymorphism impairs this protection to promote lung thrombosis in subset of patients.

Introduction

Sickle cell disease (SCD) is an inherited hemolytic disorder that affects over 6 million people of African ancestry^1–3. The most prevalent form, sickle cell anemia results from a homozygous mutation (SS) in the β-globin gene⁴. Intra-erythrocytic polymerization of the mutant hemoglobin leads to erythrocyte sickling and premature hemolysis, which promotes a thrombo-inflammatory state in SCD defined by platelet activation and hypercoagulation^5–9. De novo (in situ) pulmonary microvascular thrombosis in the absence of deep-vein-thrombosis^10,11 is the underlying pathology in ~20% of SCD patients hospitalized for acute respiratory failure¹². Autopsy¹³ and computed tomography¹² studies have confirmed that in situ pulmonary thrombosis occurring in these SCD patients is triggered by the occlusion of pulmonary arterioles by platelet-rich thrombi and is associated with the development of thrombocytopenia^12,14,15, suggesting that pulmonary thrombosis in these SCD patients is platelet-dependent. However, the molecular and genetic mechanisms that promote pulmonary thrombosis in only a subset but not most SCD patients remain elusive¹⁶.

Adenosine diphosphate (ADP), a potent platelet agonist is released in abundance from lysed erythrocytes⁴. Recent evidence suggests that ADP released during acute intravascular hemolysis can promote in situ pulmonary microvascular thrombosis by stimulating platelet-purinergic P₂Y₁ and P₂Y₁₂ receptors, leading to αIIbβ3-dependent platelet aggregation^17–19. Although intravascular hemolysis is a predominant pathophysiology in SCD²⁰, the cellular and molecular mechanism that regulates ADP-induced platelet activation in SCD remains unknown^9,17,21,22. Here, we have used quantitative fluorescence intravital lung microscopy (qFILM)^19,23–25 to show for the first time that intravascular (IV) administration of ADP does not trigger pulmonary thrombosis in SCD mice. Nucleoside triphosphate diphosphohydrolase-1 (NTPDase-1) commonly known as CD39 is an ectoenzyme, which enzymatically degrades ADP to adenosine-monophosphate (AMP)^18,26. We show that CD39 present in extracellular vesicles (CD39⁺ EVs) released by the lung endothelium, degrades circulating excess ADP to prevent platelet-dependent pulmonary thrombosis in SCD mice and patients. However, a single nucleotide polymorphism (SNP) rs3176891A→G in the CD39 encoding gene (ENTPD1) was found to be associated with impaired generation of CD39⁺ EVs, leading to a higher incidence of pulmonary thrombosis in a subset of SCD patients. These findings introduce a novel paradigm that circulating CD39⁺ EVs provide protection from pulmonary thrombosis in SCD, and identify how ENTPD1 rs3176891G allele may increase the risk for pulmonary thrombosis in SCD patients.

Results

ADP promotes in situ pulmonary thrombosis in control but not SCD mice

SCD and control mice were IV challenged with a physiological dose (2.5 mg/kg) of ADP^19,26,27 and intravital lung microscopy was used to assess the lung microcirculation in live mice as described in Methods and elsewhere^19,27 (Fig. 1a). As shown in Fig. 1b and Supplementary Movie 1, IV ADP (at t = 0 s) led to the development of in situ pulmonary thrombosis (at t = 15 s) in control mice, which was triggered by the occlusion of pre-capillary pulmonary arterioles (purple) by platelet-rich thrombi (green). The platelet-rich thrombi gradually moved downstream to occlude the pulmonary arteriolar bottle-neck (junction between pulmonary arteriole and capillaries) at t = 30 s and eventually resolved after ~2 min, resulting in a localized and transient pulmonary ischemia (absence of purple fluorescence) in control mice. Surprisingly, unlike the control mice, IV ADP failed to trigger pulmonary thrombosis in SCD mice. The platelet-rich thrombi (green) were absent, and erythrocytes (dark cells) were observed to flow unobstructed through the lung microcirculation (purple) in SCD mice (Fig. 1c and Supplementary Movie 2). Time-series of intravital images were analyzed in each mouse to plot the total pulmonary thrombus area over time (Fig. 1d), and then the area under this curve (AUC; Fig. 1e) was plotted as a combined measure of both the size and lifetime of a thrombus as described in Methods and elsewhere^19,27. Indeed, the AUC was significantly less in SCD than in control mice (Fig. 1e). Unlike IV ADP, IV challenge with 300 μg/kg collagen²⁸ triggered pulmonary thrombosis in SCD mice (Supplementary Fig. 1). The peripheral blood circulating platelet count was comparable between SCD and control mice at baseline (Fig. 1f, g). However, a 3-fold greater drop in platelet count was observed in control than SCD mice within 30 s following IV challenge with 2.5 mg/kg ADP, suggestive of the development of thrombocytopenia secondary to pulmonary thrombosis in control but not SCD mice (Fig. 1h). Unlike IV ADP, IV challenge with 200 μg/kg collagen²⁸ led to thrombocytopenia in both SCD and control mice (Fig. 1i). Next, platelet-rich-plasma (PRP) from control or SCD mice were used in in vitro turbidimetric aggregation assay²⁹, which reports percent increase in light transmission as a surrogate for platelet aggregation and the area under the aggregation response curve (AUC) as a combined measure of the lifetime and size of aggregates. In contrast to control mice PRP, ADP (5 µM)²⁶ treatment led to impaired platelet aggregation (Fig. 1j) and significantly lower AUC in SCD mice PRP (Fig. 1k), however, the collagen (3 µg/ml)²⁶ induced platelet aggregation in SCD mice PRP was neither impaired nor different from control mice PRP (Fig. 1l, m), suggestive of normal platelet function.

Fig. 1. ADP promotes pulmonary thrombosis in control but not in SCD mice.

a Experimental scheme for (b–e): control and SCD mice intravascularly (IV) administered fluorescein isothiocyanate (FITC) dextran and V450-anti-mouse CD49b Ab to visualize lung microcirculation (pseudo-colored purple) and platelets (pseudo-colored green), respectively. Pulmonary circulation was assessed using quantitative fluorescence intravital lung microscopy (qFILM) before (t = 0 s) and immediately following IV challenge with 2.5 mg/kg ADP. Created in BioRender. Brzoska, T https://BioRender.com/48xiunk. QFILM images of lung microcirculation in b control and c SCD mice at different time points. * denotes alveoli. Ellipses denote arterial bottlenecks and arrows denote blood flow direction. Scale bars 50 μm. d Areas of platelet-rich thrombi in b, c are shown as a function of time. e Area under the curve (AUC; shaded areas in d) compared between control (n = 9) and SCD (n = 8) mice post IV ADP. f Experimental scheme for (g–i). Created in BioRender. Brzoska, T https://BioRender.com/x59rs9y. g Baseline circulating platelet count in control (n = 10) and SCD (n = 11) mice. Drop in circulating platelet count (relative to baseline) in (h) control (n = 5) vs SCD (n = 6) mice 30 s post 2.5 mg/kg IV ADP or (i) control (n = 5) vs SCD (n = 5) mice 1 min post IV 200 µg/kg collagen. Representative in vitro platelet aggregation kinetics shown as a percent increase in light transmission over time in platelet-rich plasma (PRP) of control (red) and SCD (blue) mice following the addition of j 5 µM ADP or l 3 µg/ml collagen. Area under the curve (AUC; shaded areas in j) compared between control (n = 6) and SCD (n = 7) mice PRP samples following the addition of k 5 µM ADP or m 3 µg/ml collagen. Means compared using two-tailed Student’s t test. Data represent mean ± SEM. Exact P values shown in the graphs.

Protection from pulmonary thrombosis in SCD mice is CD39 dependent

Impaired development of pulmonary thrombosis in IV ADP challenged SCD mice was associated with rapid depletion of circulating ADP (Supplementary Fig. 2). The intravascular phosphohydrolysis of ADP to AMP primarily depends on the ADPase activity of CD39, a transmembrane ectoenzyme constitutively expressed on the endothelium in the lung^18,26. SCD mice were IV challenged with 2.5 mg/kg of ADP with or without pretreatment with a CD39 inhibitor, sodium metatungstate (POM-1)^30,31 and intravital lung microscopy was used to assess the lung microcirculation in live mice (Fig. 2a). Absence of pulmonary thrombosis at t = 0 s (Fig. 2b) suggested that POM-1 pretreatment alone (without IV ADP challenge) was not sufficient to restore pulmonary thrombosis in SCD mice. Remarkably, CD39 inhibition by POM-1 restored IV ADP-induced in situ pulmonary thrombosis in SCD mice (Fig. 2b and Supplementary Movie 3). Unlike SCD mice administered IV ADP (Fig. 1c and Supplementary Movie 2), but similar to control mice administered IV ADP (Fig. 1b and Supplementary Movie 1), IV administration of ADP to POM-1 pretreated SCD mice (at t = 0 s) triggered occlusion of pulmonary arterioles (purple) by platelet-rich thrombi (green) at t = 30 s followed by resolution of thrombi after ~2 min (Fig. 2b and c and Supplementary Movie 3). The AUC was significantly higher in POM-1-treated than untreated SCD mice following IV ADP challenge (Fig. 2d), suggesting that CD39 inhibition led to an increase in both the size and lifetime of platelet-rich thrombi in the lung of SCD mice. Within 30 s, IV ADP also led to a significantly higher drop in circulating platelet counts in POM-1 pretreated than untreated SCD mice (Supplementary Fig. 3), indicating the development of thrombocytopenia secondary to pulmonary thrombosis. Therefore, we next assessed the effect of CD39 inhibition on ADP-induced platelet aggregation in vitro (Fig. 2e). Pretreatment with 100 µM POM-1³² rescued ADP (5 µM) induced platelet aggregation in SCD mice PRP (Fig. 2f), resulting in significantly higher AUC in POM-1 treated than untreated SCD mice PRP (Fig. 2g). In contrast, pretreatment with POM-1 did not affect the ADP-induced platelet aggregation in control mice PRP (Fig. 2h, i).

Fig. 2. Protection from pulmonary thrombosis in SCD mice is CD39-dependent.

a Experimental scheme for b–d: SCD mice IV administered with fluorescein isothiocyanate (FITC) dextran and V450-anti-mouse CD49b Ab to visualize lung microcirculation (pseudo-colored purple) and platelets (pseudo-colored green), respectively. Next, mice were treated IV twice with 20 mg/kg POM-1. First alone, and 5 min later again with 2.5 mg/kg IV ADP. Pulmonary circulation was assessed using qFILM before (t = 0 s) and after IV ADP + POM-1. Created in BioRender. Brzoska, T https://BioRender.com/fyqulcv. b QFILM images of the same field of view (FOV) at four different time points in the lung of an SCD mouse. * denote alveoli. Ellipses denote arterial bottle-necks and arrows denote blood flow direction. Scale bar 50 μm. c Representative plot showing area of platelet-rich pulmonary thrombi as a function of time in a SCD mouse challenged IV with ADP (blue) vs ADP + POM-1 (green). d Area under the curve (AUC; shaded areas in panel c) compared between SCD mice IV treated with ADP + POM-1 (n = 4) vs ADP alone (n = 8). e Experimental scheme for (f–i). Created in BioRender. Brzoska, T https://BioRender.com/ycnc5z4. Representative in vitro platelet aggregation kinetics shown as a percent increase in light transmission over time in platelet-rich plasma (PRP) of f SCD and h control mice following the addition of 5 µM ADP without vs with 100 µM POM-1. g The area under the platelet aggregation curves (AUC) was compared between SCD mice PRP samples treated with 5 µM ADP (n = 8) and 5 µM ADP + 100 µM POM-1 (n = 4). i AUC compared between control mice PRP samples treated with 5 µM ADP (n = 5) and 5 µM ADP + 100 µM POM-1 (n = 5). Means compared using two-tailed Student’s t test. Data represent mean ± SEM. Exact P values shown in the graphs.

CD39⁺ EVs in the plasma prevent ADP-dependent platelet aggregation in SCD

The PRP lacks endothelial cells, which are the primary source of CD39 in the vasculature^18,26. Therefore, the impairment of ADP-induced platelet aggregation in SCD mice PRP (Fig. 1j, k) and the subsequent restoration following CD39 inhibition (Fig. 2f, g) was an indicative of CD39-dependent ADPase activity in the plasma. However, CD39 activity is strictly dependent on its transmembrane domain^33–35, suggesting that membrane-associated CD39 is the likely source of ADPase activity in the SCD mice plasma. EVs are membrane vesicles released by diverse vascular cells³⁶. Circulating EVs are known to be abundant in the plasma of SCD patients^24,37. Circulating EVs were isolated from equal volumes of control and SCD mice platelet-free plasma (PFP) using size exclusion chromatography³⁸ and characterized using nanoparticle tracking analysis (Fig. 3a and “Methods”)^24,39,40. Consistent with prior reports^24,41, circulating EVs were 50–400 nm in size (Fig. 3b) and significantly (2-folds) more abundant in SCD than control mice plasma (Fig. 3c). Remarkably, western blot analysis confirmed the presence of CD39 in EVs and the expression was two-fold (significantly) higher in SCD than control mice EVs (Fig. 3d and e and Supplementary Fig. 4). The size distribution of CD39⁺ EVs was similar to that of total EVs (Supplementary Fig. 5). ADPase activity was also two-fold (significantly) higher in SCD than control mice EVs, and significantly reduced following CD39 inhibition with ARL67156⁴² (Fig. 3f). SCD mice EVs also led to 3-folds higher (significantly) degradation of ADP than control mice EVs (Supplementary Fig. 6a), which was significantly reduced following pretreatment of EVs with CD39-inhibitor POM-1 (Supplementary Fig. 6b). Importantly, ADP-induced platelet aggregation in control mice PRP was abolished following the addition of SCD mice EVs but not SCD mice EVs pretreated with CD39 inhibitor POM-1 (Fig. 3g–i), suggesting that CD39 present in circulating EVs prevent ADP-dependent platelet aggregation in SCD mice.

Fig. 3. CD39⁺ endothelial extracellular vesicles prevent ADP-dependent platelet aggregation in SCD.

a Experimental scheme for (b–f, j). EVs Extracellular vesicles, SEC Size exclusion chromatography, NTA Nanoparticle tracking analysis. Created in BioRender. Brzoska, T https://BioRender.com/t6bmaj7. b NTA plot showing concentration vs size distribution of EVs isolated from a control and an SCD mouse plasma. c EV concentration in plasma of SCD (n = 7) and control (n = 7) mice. d Western blot micrograph and e the densitometry analysis (arbitrary units) of CD39 protein expression in control (n = 5) and SCD (n = 5) mice EVs. Ponceau-S, loading control. f ADPase activity in control (n = 5) and SCD (n = 5) mice EVs ± incubation with CD39 inhibitor (500 µM ARL67156). g Experimental scheme for (h, i). Platelet-rich plasma, PRP. Created in BioRender. Brzoska, T. (2025) https://BioRender.com/vq6jifk. h In vitro platelet aggregation kinetics in a control mouse PRP sample following the addition of ADP (black), ADP + control mouse EVs (red), ADP + SCD mouse EVs (blue), and ADP + SCD mouse EVs + POM-1 (green). i Area under the curve (AUC) in four groups shown in (h). N = 4 per group. j Imaging flow cytometry images of CD39⁺/CD31⁺/CD144⁺ (row #1) or CD39⁺/CD31⁺/CD106⁺ (row #2) EVs isolated from SCD mice plasma. Bottom row- isotype control Ab stained EVs. Scale bar, 5 µm. Data representative of 3 independent experiments. k Experimental scheme for (l–n). In vitro cultured human lung microvascular endothelial cells (HMVECs-L) ± incubation with 20 µM hemin and EVs isolated from cell culture supernatant. Created in BioRender. Brzoska, T https://BioRender.com/yfnjlra. l NTA plot showing concentration vs size distribution of HMVECs-L EVs. m EV concentration in the supernatant of HMVECs-L incubated with (n = 6 independent experiments) or without (n = 6 independent experiments) hemin. n ADPase activity in EVs isolated from the supernatant of HMVECs-L incubated with (n = 4 independent experiments) or without (n = 4 independent experiments) hemin ±500 µM ARL67156. Means were compared using unpaired two-tailed Student’s t test. Data represent mean ± SEM. Exact P values shown in the graphs.

Lung endothelium generates circulating CD39⁺ EVs in SCD

Next, we used imaging flow cytometry⁴³ to determine the cellular source of CD39⁺ EVs. CD31 is a marker for both platelets and endothelial cells, however, CD39 is minimally expressed in platelets^26,44,45. As shown in the representative imaging flow cytometry images (Fig. 3j), CD39⁺ EVs were also positive for CD31 and CD106 (VCAM-1) or CD144 (VE-cadherin), suggesting that these EVs are derived from vascular endothelium. In contrast, EVs stained with isotype-matched control Abs were undetectable (bottom row in Fig. 3j). Imaging flow cytometry analysis revealed that up to ~70% of CD39⁺ EVs were CD31⁺/CD39⁺. CD39 is known to be abundantly expressed by the endothelial cells in the lung^18,26 and hemin (ferric-protoporphyrin IX) released following intravascular hemolysis promotes lung endothelial activation in SCD^46,47. Therefore, in vitro cultured human lung microvascular endothelial cells (HMVECs-L) were incubated with vehicle or 20 µM hemin and EVs were isolated from the cell culture supernatant (Fig. 3k). EVs derived from HMVECs-L were comparable in size (50 to 400 nm; Fig. 3l) to SCD patient EVs (Supplementary Fig. 9a), and significantly (two-fold) more abundant in the cell culture supernatant of hemin than vehicle-treated HMVECs-L (Fig. 3m). Remarkably, although the ADPase activity was absent in EVs derived from vehicle-treated HMVECs-L, it was significantly elevated in EVs derived from hemin treated HMVECs-L and significantly attenuated following CD39 inhibition by ARL67156 (Fig. 3n). Western blot analysis revealed the presence of CD39 in EVs derived from hemin but not vehicle treated HMVECs-L (Supplementary Fig. 7).

ENTPD1 rs3176891G allele is a risk for pulmonary thrombosis in SCD patients

The G allele of the ENTPD1 SNP rs3176891 was recently found to be associated with a modest risk of thrombosis in people of European ancestry⁴⁴. However, the effect of rs3176891G in SCD patients is unknown. The analysis of the Walk-PHASST SCD patient registry^48,49 identified 437 SCD (including 320 with SS and 78 with SC genotype) patients of African ancestry, including 22 patients with a medical history of pulmonary embolism/thrombosis (Fig. 4a and Table 1). The whole genome analysis⁵⁰ of Walk-PHASST database revealed that frequency of rs3176891 GG (homozygous) genotype was approximately two-fold higher in SCD patients with (SCD + PT; 45%) than without (SCD-PT; 27%) a medical history of pulmonary embolism/thrombosis (Fig. 4b). The rs3176891G allele frequency was also higher in SCD + PT than SCD-PT (66% vs 52%; Fig. 4c). Despite relatively small number of SCD + PT patients (22 out of 437) in the Walk-PHASST database, rs3176891G allele was strongly associated with the history of pulmonary embolism/thrombosis in these patients based on both additive (AA vs AG vs GG; Odds ratio 1.8, P = 0.082) and dominant (AA vs AG + GG; Odds ratio 2.2, P = 0.074) model of genetic association (Fig. 4d). Next, we assessed the association of rs3176891G allele with the risk of pulmonary thrombosis (deep vein thrombosis and pulmonary embolism) in 1891 people (416 cases and 1475 controls) representative of general African ancestry population in the US (largely non-SCD; <0.25% SCD patients) by performing SNP-association analysis using eligible studies in the Trans-Omics for Precision Medicine (TOPMed) database⁵¹ (Fig. 4e). The clinical characterization of human subjects in the TOPMed cohorts has been reported elsewhere⁵¹. Although rs3176891G had a frequency of ~50% in the TOPMed cohort (Fig. 4f), it was not associated with the risk of pulmonary thrombosis (Odds ratio 0.9; P = 0.36; Fig. 4d), suggesting that rs3176891G allele increases the risk of pulmonary thrombosis in humans of African ancestry only if they have SCD.

Fig. 4. ENTPD1 rs3176891G associates with fewer CD39⁺-EVs and greater risk for pulmonary thrombosis in SCD patients.

a Number of SCD patients without (-PT) vs with (+PT) medical history of pulmonary thrombosis in Walk-PHASST registry. Created in BioRender. Brzoska, T https://BioRender.com/aovqt1d. Frequency of b rs3176891 GG genotype and (c) rs3176891G allele among -PT vs + PT SCD patients in (a). Data compared using four-fold table analysis with two-tailed χ2 test. d Odds Ratio for association of rs3176891G allele with risk of pulmonary thrombosis in SCD (n = 437) and non-SCD (n = 1891) humans. Data represent point estimate of Odds Ratio ±95% CI. e Number of non-SCD humans of African ancestry without (-PT) vs with ( + PT) medical history of pulmonary thrombosis in TOPMed database. Created in BioRender. Brzoska, T. (2025) https://BioRender.com/aovqt1d. f Frequency of rs3176891G allele among non-SCD humans in (e). g Experimental scheme for (h–n): blood from SCD patients with AA or AG or GG genotype of rs3176891 processed to generate platelet-rich plasma (PRP) and platelet-free plasma (PFP). Created in BioRender. Brzoska, T https://BioRender.com/d8skzoa. h ADPase activity in EVs of SCD patients with AA (n = 4) vs AG/GG (n = 10) genotype. ADPase activity in EVs of SCD patients with i AA (n = 4) and j AG/GG (n = 10) genotype ± incubation with 500 µM ARL67156. k Imaging flow cytometry images showing CD39⁺/CD31⁺ EVs in the plasma of an SCD patient with AA genotype. Scale bar 5 µm. l Concentration of CD39⁺/CD31⁺ EVs in PFP of SCD patients with AA (n = 3) vs AG/GG (n = 4) genotype. In vitro platelet aggregation kinetics shown as the percent increase in light transmission in PRP of SCD patients with AA (black), AG (blue) and GG (red) genotypes, following the addition of m 1 µM ADP or n 3 µg/ml collagen. Data from more patients are shown as Supplementary Fig. 14. Data compared using unpaired two-tailed Student’s t test in (h, l) (mean ± SEM), paired two-tailed Student’s t test in (i), and two-tailed Wilcoxon matched-pairs signed rank test in (j). Exact P values shown in the graphs.

Table 1.

Clinical characterization of SCD patients from Walk-PHASST registry

	Controls (-PT) N = 415	Cases (+PT) N = 22
Female/Male	218/197	13/9
Age	35 (25–47)	38 (31–52)
Hemoglobin (g/dL)	9.1 (7.9–10.8)	9.7 (8.2–11.6)
White Blood Cells (K/µL)	9.5 (7.3–12.2)	9.8 (7.5–12.9)
Neutrophils (K/µL)	5.0 (3.7–6.9)	4.6 (3.6–7.1)
Platelets (K/µL)	352 (264–447)	337 (281–371)
% HbS	71 (49–85)	49 (44–74)
% HbF	5 (2–10)	3 (2–9)
rs3176891genotype
- AA	96 (23%)	3 (14%)
- AG	205 (49%)	9 (41%)
- GG	114 (27%)	10 (45%)
Hydroxyurea (Y/N)	166/249	8/14

Data shows median (interquartile range) except for the gender, genotype and hydroxyurea status. -PT and +PT, without and with history of pulmonary embolism/thrombosis, respectively.

HbS hemoglobin S, HbF fetal hemoglobin.

rs3176891G associates with fewer CD39⁺ EVs and higher platelet aggregation in SCD

PRP or PFP (Supplementary Fig. 8) was prepared from fresh blood samples collected from SCD patients with rs3176891 AA, AG and GG genotypes, and used for in vitro platelet aggregation or EVs isolation and analysis^52,53 (Fig. 4g, Supplementary Figs. 9, 10). The absence of free-protein and lipoprotein contamination in EVs was verified (Supplementary Figs. 11, 12)^24,39,40. Clinical characterization of these SCD patients is shown in Table 2. Blood hemoglobin concentration (10.04 ± 0.84 vs 9.39 ± 0.43 g/dL; Mean ± SEM; p = 0.47) and plasma cell-free heme levels (Supplementary Fig. 13) were not different between SCD patients with AA vs AG/GG genotype of rs3176891. Remarkably, the ADPase activity was two-fold lower in EVs of SCD patients with AG and GG (combined due to a small sample size) than AA genotype (Fig. 4h), and completely abolished in AA (Fig. 4i) and AG/GG (Fig. 4j) genotypes following treatment with CD39 inhibitor ARL67156. Unlike the SCD mice, the disease severity is known to be heterogeneous among SCD patients⁴, therefore, the EV characterization and platelet aggregation were analyzed separately for each patient. Identical to SCD mice (Fig. 3j), imaging flow cytometry revealed that the CD39⁺-EVs were also present in the plasma of SCD patients with AA genotype, and these EVs were indeed endothelium-derived (CD31⁺CD39⁺; Fig. 4k). Remarkably, the CD31⁺CD39⁺ EVs were significantly less in the plasma of SCD patients with AG or GG than AA genotype of rs3176891 (Fig. 4l). Next, we assessed platelet aggregation in the PRP of these patients. Similar to SCD mice (Fig. 1j), the ADP-dependent platelet aggregation was also impaired in the PRP of SCD patients with AA genotype (Fig. 4m; Supplementary Fig. 14). In contrast, ADP-dependent platelet aggregation was normal in the PRP of SCD patients with AG or GG genotype (Fig. 4m; Supplementary Fig. 14). Identical to SCD mice (Fig. 1l), collagen (3 µg/ml) induced platelet aggregation was normal in the PRP of SCD patients of all three genotypes (Fig. 4n; Supplementary Fig. 14), suggesting that the difference in ADP-dependent aggregation was not due to a defect in platelet function.

Table 2.

Clinical characterization of SCD patients used for blood draws

rs3176891 genotype	AA	AG/GG
Female/Male	1/4	9/6
Age	35.8 (31; 22; 56)	37.73 (34; 20; 60)
Hemoglobin (g/dL)	10.04 (10; 7.9; 12.5)	9.39 (9.1; 6.4; 13.7)
Hematocrit (%)	29.2 (30.4; 22.3; 35.5)	26.91 (26.3; 18.6; 37.7)
White Blood Cells (K/µL)	10.96 (10; 9; 13.9)	10.4 (10; 3.4; 29.9)
Neutrophils (K/µL)	6.3 (6.6; 5.2; 7.18)	6.03 (5.54; 0.52; 21.53)
Lymphocytes (K/µL)	2.69 (3.3; 1.4; 3.75)	2.94 (2.48; 1.1; 6.57)
Monocytes (K/µL)	1.1 (1; 0.84; 1.39)	1.05 (0.96; 0.22; 2.08)
Platelets (K/µL)	322.8 (311; 216; 501)	318.4 (297; 119; 534)
% HbS	66.58 (72.4; 51.2; 80)	71.93 (71.75; 49.6; 80)
% HbF	10.9 (5.9; 2; 31.4)	12.52 (8.4; 1.6; 35.4)
SCD genotype
- SS	4	13
- S/β0	1	2
Hydroxyurea (Y/N)	3/2	12/3

Data shows mean (median; minimum; maximum) except for the gender, genotype and hydroxyurea status.

SS sickle cell anemia, S/β⁰ sickle β⁰ thalassemia, HbS hemoglobin S, HbF fetal hemoglobin.

Discussion

ADP released during intravascular hemolysis in SCD is a prothrombotic agonist¹⁹, then why only a subset of SCD patients develop in situ pulmonary thrombosis? To address this, SCD or control mice were systemically challenged with a physiological dose of ADP. Intravital lung microscopy revealed occlusion of pulmonary arterioles by platelet-rich thrombi in control but not in SCD mice. Identical to most SCD patients, SCD mice were also protected from pulmonary thrombosis, which was secondary to impaired ADP-induced platelet aggregation. Remarkably, CD39 inhibition rescued both ADP-induced platelet aggregation and pulmonary thrombosis in SCD mice in vivo.

Circulating CD39⁺ EVs released by the endothelial cells in the lung and other organs were the major source of CD39-dependent ADPase activity in the blood of SCD mice. This EV-associated CD39 activity was most likely a contribution of both more CD39⁺ EVs and more CD39 per EV in the blood of SCD mice. Indeed, treating human lung endothelial cells with heme led to the generation of CD39⁺ EVs. Circulating CD39⁺ EVs were abundant in the blood of SCD mice and prevented ADP-induced platelet aggregation and pulmonary thrombosis in a CD39-dependent manner. Interestingly, ENTPD1 polymorphism rs3176891A→G was a risk for pulmonary thrombosis in SCD patients but not healthy humans of African ancestry, suggesting that the rs3176891G allele increases the risk primarily in the setting of intravascular hemolysis leading to release of ADP. Identical to SCD mice, circulating CD39⁺ EVs were abundant and the ADP-induced platelet aggregation was impaired in the blood of SCD patients carrying only the wild-type rs3176891A allele. In contrast, the circulating levels of CD39⁺ EVs were low and therefore, the ADP-induced platelet aggregation was normal in the blood of SCD patients carrying the rs3176891G allele. Taken together, our current findings (Fig. 5) suggest that the inflammatory milieu in SCD promotes the generation of CD39⁺ EVs by the vascular endothelium. CD39 in EVs serves to prevent platelet aggregation and pulmonary thrombosis in SCD, by degrading excess plasma ADP. However, the ENTPD1 rs3176891G allele is associated with attenuated levels of CD39⁺ EVs, leading to the increased risk of ADP-dependent platelet activation and pulmonary thrombosis in a subset of SCD patients. These findings are also supported by prior studies showing impaired vs normal ADP-induced platelet aggregation in the blood of different subsets of SCD patients^9,22.

Fig. 5. Schematic showing the main findings.

Although ADP released during acute intravascular hemolysis can trigger in situ pulmonary thrombosis by stimulating platelet-purinergic P2Y1 and P2Y12 receptors, the sterile inflammatory milieu in SCD also promotes the generation of endothelium-derived CD39⁺ EVs that phosphohydrolize ADP to prevent pulmonary thrombosis. However, ENTPD1 rs3176891G allele is associated with impaired generation of CD39⁺ EVs, thus increasing the risk of pulmonary thrombosis in some SCD patients. Created in BioRender. Brzoska, T. (2025) https://BioRender.com/x4rvhil.

The interpretation of our findings is associated with a few limitations that may inspire further investigation in future studies. First, heme-dependent toll-like receptor-4 activation promotes endothelial activation in SCD^47,54, but whether this pathway also promotes the generation of CD39⁺ EVs remains to be determined. Second, although CD39 is most abundantly expressed by the vascular endothelium, a modest expression exists on subsets of leukocytes^32,55, which might also contribute to ADP metabolism. Third, the mechanism underlying the attenuation of CD39⁺ EVs in SCD patients carrying rs3176891G allele remains unknown. Notably, CD39 associates with cholesterol-rich plasma membrane domains^34,56,57, but whether rs3176891G affects these membrane associations, and how this may regulate CD39 expression vs activity in EVs remains to be investigated. Fourth, rs3176891G is an intron-1 SNP that non-randomly associates with other SNPs⁴⁴. The role of such ENTPD1 haplotypes in regulating CD39⁺ EVs needs to be investigated in future studies. Fifth, rs3176891G allele was previously found to be associated with a modest risk of thrombosis in people of European ancestry⁴⁴, however, the TOPMed database used in the current study did not reveal any association with risk of thrombosis in healthy general African ancestry population in the US. Most likely, the differences in cohort propensity and design may have contributed to these different results. Finally, in situ pulmonary thrombosis involves occlusion of pulmonary arterioles by platelet-rich thrombi across the whole lung^10,11, however, the intravital imaging approach used in the current study is limited to the assessment of pulmonary thrombosis in only a small portion of the lung vascular bed¹⁹. Also, EVs isolated from the human plasma may contain platelets and lipoproteins as contaminants, which can affect the interpretation of the findings. Based on the MIBlood-EV guidelines⁴⁰, the current study provides necessary evidence to support that any such contamination does not impact the overall findings of the study.

Nevertheless, our study identifies a novel EV-mediated CD39-dependent mechanism that prevents pulmonary thrombosis in SCD. Despite recent advances in new therapies, in situ pulmonary thrombosis remains a major cause of clinical morbidity in SCD, with no preventive therapy^3,58. We show that ENTPD1 rs3176891G allele identifies a subset of SCD patients, who are at a greater risk of developing pulmonary thrombosis and may benefit from inhibition of purinergic signaling. Ironically, P₂Y₁₂ receptor antagonists did not prevent hospitalization in recent clinical trials, when administered to SCD patients regardless of ENTPD1 genotype^21,59. We hypothesize that such inhibitors may prevent pulmonary thrombosis if given to SCD patients with ENTPD1 polymorphisms associated with lower levels of CD39⁺-EVs such as rs3176891G.

Methods

Mice

Male and female (age ~12–16 weeks old) Townes SCD mice (SS, homozygous for Hba^tm1(HBA)Tow, homozygous for Hbb^{tm2(HBG1,HBB*)Tow}) and non-sickle control mice (AS, homozygous for Hba^tm1(HBA)Tow, compound heterozygous for Hbb^{tm2(HBG1,HBB*)Tow}/Hbb^{tm3(HBG1,HBB)Tow}) were used in this study^{23,24,47,54,60}. Townes-AS mice are Sickle Cell Trait mice that are heterozygous for β^S and wild-type human β^A. AS mice do not develop SCD^{23,24,47,54,61}. For this study, Townes SS and AS mice were used as SCD and control mice, respectively^{23–25,62–66}. All animal experiments were approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh and Versiti Blood Research Institute/Medical College of Wisconsin.

Intravital lung microscopy in mice

Quantitative fluorescence intravital lung microscopy (qFILM) was used to assess platelet-dependent in situ pulmonary thrombosis within the intact lung microcirculation of live SCD and control mice (Fig. 1a). The qFILM experimental setup and approach have been described elsewhere in detail^19,25,27,67. Briefly, intravital lung microscopy was performed using a Nikon A1R Multi-Photon-Excitation (MPE) Ni-E upright motorized microscope (Nikon Instruments; Tokyo, Japan). Two-dimensional (2D) time series of intravital images were acquired with NIS-Elements software using a chirped Chameleon Laser Vision (Coherent; Santa Clara, CA) emitting an excitation wavelength of 850 nm, an APO LWD 25x water immersion objective with 1.1 NA, a high-speed resonant scanning mode capable of acquisition at 512 × 512 resolution with 2x line averaging and bi-directional scanning (~15 frames per second), and four GaAsP NDD detectors. Prior to imaging, mice were anesthetized with an intraperitoneal (IP) injection of 100 mg/kg ketamine HCl (Covetrus, Portland, ME) and 20 mg/kg Xylazine (Akorn Pharmaceuticals, cat# 59399-111-50). A cannula was inserted into the right carotid artery, and a tracheotomy was performed to facilitate mechanical ventilation with 95% O₂ and maintenance anesthesia (1% isoflurane). During intravital imaging^27,68, mice were mechanically ventilated (volume-controlled) at ~120 breaths/min with a tidal volume of 10 μl/g body weight using a MiniVent Type 845 ventilator (Harvard Apparatus; Holliston, MA). The ventilator was also used to deliver 1% maintenance isoflurane anesthesia (Henry Shein Animal Health) with a FiO2 of 0.95. Positive end-expiratory pressure was not applied in this study. The left lung was surgically exposed, and a small portion of the lung was immobilized against a coverslip using a vacuum-enabled micromachined device, as described elsewhere^25,27,67. Fluorescein Isothiocyanate (FITC) dextran (MW 70 KDa, 75 μg/mouse, Thermo Fischer Scientific, Waltham, MA) and Violet 450 (V450) rat-anti mouse CD49b mAb (clone DX5, 7 μg/mouse, BioLegend, San Diego, CA) were intravascularly (IV) administered through the carotid artery catheter to enable in vivo visualization of the lung microcirculation and in vivo staining of platelets, respectively. To trigger in situ pulmonary thrombosis in mice, 2.5 mg/kg adenosine diphosphate (ADP, Millipore Sigma, cat# 01905) or 300 µg/kg collagen (Chrono-Log Corp., cat# 385) was administered through the carotid artery catheter^19,27. In some studies, 20 mg/kg sodium metatungstate³¹ (POM-1, TOCRIS, cat# 12141-67-2) was IV administered twice, 5 min before and then simultaneously with ADP. In each mouse, a single 2D plane (FOV ~ 118,000 μm²) containing a pulmonary arteriole (30 to 50 µm diameter) and the downstream pulmonary capillary bed was randomly selected, and the presence or absence of platelet-rich thrombi was assessed. These observations were carried out in 4 to 9 mice per study group. Time series of qFILM images were recorded prior (t = 0 s) and up to 10 min post IV administration of ADP ± POM-1 or 20 min post IV collagen. The experiment concluded with the euthanization of the mouse using an overdose of anesthesia¹⁹.

QFILM image processing and data analysis

Time series of qFILM images were processed and analyzed using Nikon’s NIS-Elements software (NIKON, USA) as described previously^19,27,64,69. First, an image subtraction algorithm was used to remove autofluorescence and bleed through between fluorescent channels. Second, the signal-to-noise ratio was enhanced by using a median filter algorithm followed by smoothing and denoising algorithms. Third, each channel was pseudo-colored as follows: microcirculation as purple and platelets as green to enhance contrast and facilitate visualization. Next, pulmonary arterioles and downstream capillaries were analyzed for the quantitative assessment of in situ pulmonary thrombosis as described previously^19,27. Briefly, platelet-rich thrombi were defined as platelet aggregates (area ≥ 10 µm²) sequestered within the pre-capillary pulmonary arterioles and extending down into the pulmonary capillaries. Two-dimensional sizes (areas in μm²) of platelet-rich thrombi were estimated in NIS-Elements by converting qFILM images into binary images and adjusting the threshold range of the intensity histograms uniformly over the entire FOV in each frame of the time-series. The sizes of all the platelet-rich thrombi in a single frame were added to estimate the total pulmonary thrombi area, which was plotted as a function of time in GraphPad Prism 7 (GraphPad Software; La Jolla, CA). This approach allowed us to assess the kinetics of the initiation and progression of pulmonary thrombosis in an observed FOV. Changes in total pulmonary thrombi area in an FOV over time was used to estimate the area under the curve (AUC), which is a combined measure of both the size and lifetime of platelet-rich thrombi.

In vivo assessment of thrombocytopenia in mice

SCD and control mice were anesthetized with an IP injection of 100 mg/kg ketamine HCl and 20 mg/kg xylazine. A cannula was inserted into the right carotid artery. Mice were IV administered 2.5 mg/kg ADP or 200 µg/kg collagen (Chrono-Log Corp., cat# 385) through the femoral vein. In some mice, 20 mg/kg POM-1 was IV administered twice, 5 min before and then simultaneously with ADP. Peripheral whole blood samples (30 µl) were collected through the carotid artery cannula into a capillary blood collection tube (Microvette,100 EDTA K3E, Sarstedt, cat# 20.1278.100). Samples were collected immediately prior to ADP or collagen challenge and at 0.5, 1, and 2 min post-ADP challenge, or 1, 2, and 5 min post-collagen challenge. Platelet concentration in each sample was determined using HemaVet 950FS (DREW Scientific Inc.). Maximum platelet concentration changes (K/µl) in peripheral blood were estimated for each mouse by subtracting platelet concentration at 0.5 min post ADP or 1 min post collagen challenge from the baseline platelet concentration (t = 0 min, before ADP or collagen challenge). Additionally, the relative changes in platelet concentration (% of the baseline concentration) after IV challenge with ADP or collagen were also plotted against time.

In vitro mouse platelet aggregation

SCD and control mice were anesthetized with an IP injection of 100 mg/kg ketamine HCl and 20 mg/kg xylazine. Blood was drawn through the inferior vena cava using a 1 ml syringe containing 3.8% sodium citrate (1:10). Blood sample was diluted with an equal volume of saline (maintained at room temperature) and centrifuged (100 g for 8 min at 22 °C) to prepare platelet-rich plasma (PRP). A fraction of PRP was again centrifuged (1800 g for 10 min at 22 °C) to obtain platelet-poor plasma (PPP). Platelet concentration in the PRP was determined using the HemaVet 950FS and adjusted to 150 × 10³ platelets/µl using autologous PPP. Platelet aggregation in the PRP was assessed by optical turbidimetry method^29,70 using a Chrono-log Model 700 Lumi-Aggregometer (Chrono-log Corp.). Percent changes in light transmission, a surrogate for platelet aggregation were recorded for 10 min following the addition of saline, 5 µM ADP or 3 µg/ml collagen to the PRP. In some cases, PRP was pre-incubated with 100 µM POM-1 for 5 min prior to the addition of 5 µM ADP. Aggregation response curves were generated to illustrate changes in % light transmission over time. Higher % light transmission indicated greater aggregation. The area under the curve (AUC as shown in Fig. 1j) was used as a combined measure of the aggregates’ size and lifetime.

Isolation of extracellular vesicles from mouse plasma

SCD and control mice were anesthetized with an IP injection of 100 mg/kg ketamine HCl and 20 mg/kg xylazine. Blood was drawn through the inferior vena cava using a 1 ml syringe containing 3.8% sodium citrate (1:10). Whole blood samples were first centrifuged at 1800 g for 10 min at 22 °C to obtain platelet-poor plasma (PPP). The PPP was then transferred into a new tube and centrifuged at 3000 g for 15 min at 4 °C to remove any remaining cells and cellular debris, yielding platelet-free plasma (PFP)^71,72. Extracellular vesicles (EVs) were isolated from the mouse PFP samples utilizing qEV size-exclusion chromatography (SEC) columns (qEVoriginal/70 nm Gen 2 Column, IZON Science LTD) and the Automatic Fraction Collector V2 (AFC; IZON Science LTD). The PFP sample (500 µl) was loaded on the qEV column and then eluted with 0.22-µm pre-filtered elution buffer. The composition of the elution buffer was chosen based on future EV analysis method and discussed below. The AFC was programmed to discard the initial 1 ml of eluent (column void volume) and subsequently collect twenty 0.5 ml fractions. As shown in the Supplementary Fig. 11, EVs were primarily present in fractions 4–8 and free protein in fractions 9–20. To avoid any contamination by free protein, only SEC elute fractions 4–7, which contained the vast majority of EVs, were pooled to a total volume of 2 ml and concentrated using an Amicon Ultra-2 30 kDa centrifugal filter unit (Millipore Sigma, cat# UFC203024) through centrifugation at 4000 × g for 30 min at 4 °C. The concentration and size distribution of EVs in each sample were measured by nanoparticle tracking analysis (NTA) using NanoSight-300 or fluorescence detection enabled NanoSight-Pro system (Malvern Instruments Limited, Malvern, UK)²⁴. The size distribution of isolated EVs was found to be in the range of 50–400 nm.

Western blot analysis of mouse EVs

SCD and control mouse plasma EVs were isolated using SEC with phosphate-buffered saline (PBS, pH = 7.5) as the elution buffer. To prepare EV lysates, freshly concentrated EV samples were supplemented with 1 µl of HALT® Protease & Phosphatase Inhibitor Cocktail and 1 µl of 0.5 M EDTA (Thermo Scientific, cat# 78440), along with 10 µl of 10x RIPA buffer (Millipore Sigma, cat# 20-188) and PBS to achieve a final volume of 100 µl. The samples were then stored on ice for 30 min and vortexed every 10 min. Finally, the samples were snap-frozen in liquid nitrogen and stored at −80 °C until use. EV lysate samples (30 µl) were reduced using 10X Bolt™ Sample Reducing Agent (Invitrogen, cat# B0009) for 10 min at 70 °C, and subsequently separated on Bolt™ Bis-Tris Plus Mini Protein Gel, 4–12%, 1.0 mm (Invitrogen, cat# NW04122BOX) using 5% MES SDS Running Buffer (Invitrogen, cat# B0002) for 40 min using 200 V and 500 mA power settings. Afterward, proteins were electro-transferred using Bolt™ Transfer Buffer (Invitrogen, cat# BT00061) onto 0.2 µm pore-size nitrocellulose membrane (Invitrogen, cat# LC2000) using 10 V and 500 mA power settings for 60 min. Membranes were subjected to Ponceau S solution (Thermo Scientific, cat# A40000279) staining, washed with tris-buffered saline containing Tween-20 (TBS-T, Thermo Scientific Chemicals, cat# J77500.K8), and then blocked using 5% non-fat milk (Lab Scientific, cat# M0841) in TBS-T for 1 h at room temperature. Next, membranes were incubated overnight with anti-CD39 antibody (EPR3678(2); Abcam, Cat#: ab108248, 1:2000)^73,74 in TBS-T containing 5% non-fat milk at 4 °C. After the overnight incubation, membranes were washed three times for 10 min with TBS-T at room temperature and incubated for 1 h at room temperature with the secondary antibody (Anti-Rabbit IgG; HRP-linked, Cell Signaling, Cat#: 7074S) diluted in TBS-T containing 5% non-fat milk (1:10,000). Membranes were rinsed 3x for 10 min with TBS-T and treated with SuperSignal™ West Pico PLUS Chemiluminescent Substrate (Thermo Scientific, cat# 34580) for 2 min in the dark. Films were developed using Konica SRX-101A and analyzed using NIH-ImageJ software^23,24.

Assessment of ADPase activity in mouse EVs

SCD and control mouse plasma EVs were isolated using SEC with HEPES-buffered saline (NaCl 140 mM, HEPES 50 mM, pH = 7.5; HBS) as the elution buffer. Freshly isolated EVs were concentrated as described above, and each sample volume was adjusted to 100 µl using HBS. In a 96-well plate, duplicate sets of reaction and control samples (total 40 µl per well) were prepared. The reaction sample consisted of 10 µl EV sample, 20 µl ADPase assay buffer (NaCl 140 mM, HEPES 50 mM, CaCl₂ 10 mM, Tetramisole hydrochloride 10 mM), and 2 mM ADP (in HBS), with or without 500 µM ARL 67156 (in HBS, Millipore Sigma, Cat#: A265). Tetramisole hydrochloride (Sigma-Aldrich, cat# T1512) was used to inhibit alkaline phosphatase activity⁷⁵. For each reaction sample, a corresponding control sample without ADP was prepared. Additionally, a control sample containing ADP but lacking EVs was also prepared. Subsequently, all samples were incubated for 2 h at 37 °C. Following incubation, ADPase activity was assessed with QuantiChrom ATPase Assay Kit (BioAssay Systems, Cat# DATG-200) according to the manufacturer’s instructions^76,77. This assay allows measurement of ADPase enzymatic activity by quantifying the concentration of nonorganic free phosphate ions (Pi) produced during ADP phosphohydrolysis. The Pi concentrations of the control samples were subtracted from the respective Pi concentration of the reaction sample. The resulting final Pi concentration for each sample was used to calculate ADPase activity according to the manufacturer’s instructions. In another set of experiments, EVs were isolated from SCD mouse PFP using SEC with HEPES-buffered saline (HBS; 140 mM NaCl, 50 mM HEPES, pH 7.5) as the elution buffer. Freshly isolated EVs were analyzed by NTA. Reaction mixtures were then prepared containing 5 × 10¹⁰ EVs/ml, 5 mM CaCl₂, and 10 mM tetramisole hydrochloride, with or without 25 µM POM-1. Each sample was brought to a final volume of 100 µl with HBS and supplemented with 10 µM N⁶-Etheno-ADP. Samples were incubated at 37 °C for 1 h, and reactions were stopped by heating samples at 95 °C for 90 s. Samples were then snap-frozen and stored at –80 °C until high-performance liquid chromatography (HPLC) analysis of N⁶-Etheno-ADP as previously described⁷⁸. We have recently applied this method to assess extracellular metabolism of adenine nucleotides/nucleosides in EVs^79,80.

Assessing the effect of EVs on ADP-dependent platelet aggregation

SCD and control mice plasma EVs were isolated using SEC with HBS as the elution buffer and then concentrated as described in the previous paragraph. Next, the concentration of EVs in each sample was determined using NTA and subsequently adjusted to 1.2 × 10¹⁰ EVs/ml using HBS. EV samples were supplemented with 5 mM CaCl₂ and 5 mM Tetramisole hydrochloride and incubated with 125 µM ADP ± 50 µM POM-1 for 1 h at 37 °C. Additionally, a control sample lacking EVs was also prepared using the same incubation process. 10 µl of these EV samples were added to 240 µl of control mouse PRP (prepared as described earlier) and platelet aggregation was assessed as described in the previous paragraph.

Imaging flow cytometry of mouse EVs

SCD and control mouse plasma EVs were isolated and concentrated as described above, and each sample volume was adjusted to 100 µl using PBS. For each treatment group, 20 µl of EVs sample was diluted with 80 µl of sterile PBS and incubated in the dark for 15 min at 4 °C with Alexa Fluor 647 anti-mouse CD39 antibody (clone: Duha59; BioLegend cat#143808), Pacific Blue anti-mouse CD106 antibody (clone: 429; BioLegend cat#105722), Brilliant Violet 421 anti-mouse CD144 antibody (clone: BV421; BioLegend cat#138013), and Phycoerythrin anti-mouse CD31 antibody (clone: 390; BioLegend cat# 102408) for staining of CD39, CD106, CD144 and CD31, respectively. Details of all the fluorescent Abs and the respective isotype matched control Abs are listed in Supplementary Table 1. Imaging flow cytometry was conducted using Image Stream equipment (Amnis, Seattle, WA)²³. The laser power was adjusted for proper data acquisition and remained constant between samples. Each sample was run twice, and at least 20,000 unique events were captured. Data was analyzed using IDEAS application software version 6.2.187.0 (Amnis, Seattle, WA) with integrated colocalization of fluorescence and brightfield detection tool²³. Endothelial-derived CD39-positive (CD39⁺) EVs were identified as particles triple positive for CD39, CD31 (PECAM-1) and CD106 (VCAM-1) or CD144 (VE-Cadherin). The MIFlowCyt-EV framework⁸¹ spreadsheet and checklist are included as Supplementary Data 1 and 2, respectively.

EV generation by cultured human lung endothelial cells

Human lung microvascular endothelial cells (HMVECs-L; Lonza, Cat# CC-2527) were cultured (37 °C in a 5% CO₂ atmosphere) up to passage 7 in EGMTM-2 MV medium supplemented with SingleQuots (Lonza, Cat# CC-3202). Hemin (Sigma-Aldrich, cat# H9039) solution was prepared freshly each time and filtered through a 0.22 µm filter unit as described elsewhere^23,47,54. On the day of the experiment, the complete growth media were replaced with 1 ml of reduced media (growth media containing 1% FBS and no hydrocortisone) 1 h prior to any treatment. HMVEC-L were then incubated with 20 µM hemin for 12 h^82,83. After a 12-h incubation period, the cell culture supernatant was collected, centrifuged at 3000 g for 15 min (4 °C), and used for EV isolation. Endothelial cell-derived EVs were isolated using SEC and HBS as the elution buffer. Freshly isolated EVs were concentrated as described above, and each sample’s volume was adjusted to 100 µl using HBS. The concentration and size distribution of EVs in each sample were determined utilizing NTA. The ADPase enzymatic activity of the isolated endothelial cell EVs was analyzed as described above.

Analyzing the association of rs3176891A → G with the risk of pulmonary thrombosis

Walk-PHASST SCD patient database

We used two logistic regression models to calculate the association between each SNP and the risk of pulmonary embolism/thrombosis in SCD patients of African ancestry as described elsewhere⁸⁴. In one model, we used additive genetic (effect of each additional copy) and in another model we used dominant model. Both models were adjusted for age at recruitment and sex. The clinical characterization of these patients is shown in Table 1 and also reported elsewhere^48–50.

TOPMed general African ancestry database

We used data from human genetic sequencing in the National Heart Lung, and Blood Institute’s TransOmics for Precision Medicine (TOPMed) Program and all participants provided informed written consent to use their genetic data for research. The current study was approved by the TOPMed VTE working group. TOPMed cohort is representative of the general population of African ancestry in the US, which might include a small number (<0.25%) of SCD patients based on the prevalence of this disease in the US (1 in 500 humans of African ancestry). Based on this, TOPMed cohort was used in this study as a cohort comprised of largely non-SCD (without SCD) humans of African ancestry. Effect of each additional copy of the alternate allele (additive model) and thrombotic complications was calculated in subjects of African ancestry. Analysis was adjusted for sex, age, study cohort, population structure, and relatedness. The clinical characterization of these human subjects has been reported elsewhere in detail⁵¹.

Blood collection from SCD patients with different genotypes of rs3176891

Steady-state (not in crisis) SCD (SS or S/β°) patient blood was collected via venipuncture in a BD Vacutainer containing sodium citrate (BD Biosciences, Franklin Lakes, NJ) at the University of Pittsburgh Medical Center (UPMC) or Froedtert Hospital under the protocols approved by the Institutional Review Boards at the University of Pittsburgh and Medical College of Wisconsin, respectively. The procedure for blood draw has been described elsewhere in detail^23–25. Only non-smokers who were not on chronic transfusion therapy (no transfusion within the last 1 month) were included in the study. Informed written consent was obtained from all the participants in accordance with the Declaration of Helsinki. All blood samples were processed within 1 h of blood draw. First, genomic DNA was isolated from whole blood using either the Puregene or QiaAmp blood DNA isolation kit (Qiagen) as described elsewhere⁸⁵. DNA (~50 ng) was used to genotype ENTPD1 SNP rs3176891 using a predesigned Taqman assay (C-368543-10) and a QuantStudio 5 qPCR System (ThermoFisher) according to the manufacturer’s instruction and methodology described previously^85,86 to identify SCD patients carrying AA or AG, or GG genotype of rs3176891. Next, fresh blood samples were collected via venipuncture in BD Vacutainers containing sodium citrate from these selected SCD patients with different genotypes of rs3176891 and used in experiments. Clinical characterization of these selected patients and their rs3176891 genotypes are shown in Table 2.

Human platelet-free plasma preparation

Fresh blood samples from SCD patients with AA, AG, or GG genotypes of ENTPD1 SNP rs3176891 were processed to prepare platelet-free plasma (PFP) using the same two-step centrifugation protocol, which was used for isolation of mice EVs and described elsewhere^71,72.

Estimation of cell-free heme in SCD patient blood

Concentrations of cell-free heme in PFP samples from SCD patients with AA, AG, and GG genotypes of ENTPD1 SNP rs3176891 were analyzed using heme assay kit (Abcam cat#AB272534), following the manufacturer’s instructions.

In vitro platelet aggregation in SCD patient PRP

Fresh blood samples from SCD patients with AA, AG, and GG genotypes of ENTPD1 SNP rs3176891 were centrifuged at 250 g (10 min at 22 °C) to prepare PRP. Platelet aggregation in the PRP was assessed following the addition of 1 µM ADP or 3 µg/ml collagen using the methodology described under mice PRP studies.

Isolation of EVs from SCD patient blood

Fresh blood samples from SCD patients with AA, AG, or GG genotypes of ENTPD1 SNP rs3176891 were processed to prepare platelet-free plasma (PFP). Based on MIBlood-EV-2023 guidelines⁴⁰, the depletion of platelets in PFP was confirmed by assessing platelet counts using complete blood-cell count (CBC) analyzer (Supplementary Fig. 8a) and flow cytometry (Supplementary Fig. 8b). EVs were isolated from PFP using the methodology described under mice EV studies and absence of free-protein contamination in EV samples verified (Supplementary Figs. 11, 12).

Assessment of ADPase activity in SCD patient EVs

ADPase activity was assessed in the EVs isolated from the plasma of SCD patients with AA, AG, and GG genotypes of ENTPD1 SNP rs3176891 using the QuantiChrom Assay Kit as described under mice EV studies.

Imaging flow cytometry of SCD patient EVs

Freshly isolated SCD patient EVs were processed for imaging flow cytometry using the methodology described under mice EV studies. For each treatment group, 20 µl of EVs suspension was diluted with 80 µl of sterile PBS and incubated in the dark for 15 min (4 °C) with Alexa Fluor 647 anti-human CD39 antibody (clone: 498403; R&D Systems cat# FAB4397R) and Phycoerythrin (PE) anti-human CD31 antibody (clone: WM59, BD Pharmingen cat# 560983) for in situ staining of CD39 and CD31, respectively. Imaging flow cytometry was conducted using Image Stream equipment as described in the previous paragraphs.

Statistical analysis

GraphPad 7 Prism (GraphPad, Boston, MA) software was used to perform the statistical analysis. Data distribution was analyzed using the Shapiro-Wilk normality test. Data were compared using the two-tailed Wilcoxon matched-pairs signed rank test (when the data was not normally distributed) or the two-tailed paired or unpaired Student’s t-test (when the data was normally distributed). Means of more than two groups were compared using one-way ANOVA with Bonferroni correction for multiple comparisons. Error bars represent standard error of measurement (SEM). A p < 0.05 was used to determine significance.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

T.B. was responsible for experimental design, performance and analysis, manuscript writing, and project supervision. T.W.K. contributed to EV isolation, EV characterization and imaging flow cytometry experiments. O.K., S.E.T., T.P.S., and A.E.A. were involved in the experiments involving animals. C.S.C. and S.C.W. contributed to the QFILM study. A.A.S., C.L., N.D.P., N.L.S., E.L.G., J.S.P., and C.K. analyzed the association of rs3176891G with risk of VTE in TOPMed cohort. MTG and GJK contributed to Walk-PHASST SCD patient registry analysis. Y.Z. performed rs3176891 genotyping in SCD patients. E.M.N. and J.C.J. were involved in blood collection from SCD patients with different rs3176891genotypes. JJF and SH were involved in studies with SCD patient blood. E.V.M., S.P.T., and E.K.J. were involved in CD39 ADPase activity studies. S.M.N. analyzed the association of rs3176891G with risk of pulmonary thrombosis in Walk-PHASST registry. P.S. was responsible for experimental design, manuscript writing, project funding and project supervision. TB and PS wrote the manuscript with consultation and contribution from all coauthors.

Peer review

Peer review information

Nature Communications thanks Nicola Conran, Rienk Nieuwland, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All the data generated or analyzed during this study are included in this published article (and its supplementary information files). Figure 4a-f were generated by the analysis of publicly available Whole Genome Sequencing (WGS) data in NHLBI TOPMed as following: (1) Walk-PHaSST Sickle Cell Disease study with database of Genotypes and Phenotypes (dbGaP) accession number phs001514; (2) Cardiovascular Heart Study (CHS) with dbGaP accession number phs001368; (3) Heart and Vascular Health Study (HVH) with dbGaP accession number phs000993; (4) Women’s Health Initiative Study (WHI) with dbGaP accession number phs001237; (5) Atherosclerosis Risk in Communities Study (ARIC) with dbGaP accession number phs001211; (6) Mayo Clinic Venous Thromboembolism Study (Mayo_VTE) with dbGaP accession number phs001402. Clinical data corresponding to individual patients used for blood sample collection in this study are provided in the Source Data file, with appropriate de-identification to maintain patient confidentiality and follow IRB guidelines. Access to the public genotype dataset is subject to standard usage restrictions to protect participant privacy. No additional restrictions apply to the raw or clinical data provided in this manuscript. Source data are provided with this paper.

Competing interests

P.S. received funding (not relevant to the current study) as a part of sponsored research agreements with CSL Behring Inc, IHP Therapeutics and Novartis Inc. P.S. was also the recipient of 2021 Bayer Hemophilia Award Program (not relevant to the current study). Other authors have declared that no conflict of interest exists.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68396-2.