Distinct endothelial cell toxicities of Abl tyrosine kinase inhibitors lead to arterial thrombosis

Key Points

• •Dasatinib, ponatinib, and nilotinib damage ECs in different ways, all of which can contribute to arterial thrombosis.
• •Asciminib, the most specific Abl inhibitor, does not cause human coronary EC damage in vitro nor thrombosis in mouse models.

Visual Abstract

Abstract

Inhibitors targeting Abl kinase have dramatically improved survival in Philadelphia chromosome–positive leukemias. First-generation imatinib has minimal cardiovascular side effects, whereas newer agents such as dasatinib, ponatinib, and nilotinib are more effective cancer treatments but carry a high risk of arterial thrombosis. The allosteric Abl kinase inhibitor asciminib was recently approved without long-term cardiovascular follow-up. Previous studies reveal disparate effects of dasatinib, ponatinib, and nilotinib on platelets with consistent evidence of endothelial cell (EC) toxicity. Here, we explore prothrombotic endothelial toxicity mechanisms by comparing exposure to vehicle vs clinically relevant concentrations of imatinib, dasatinib, ponatinib, nilotinib, and asciminib on primary human coronary artery ECs (HCAEC) and mouse models of endothelial injury and vascular thrombosis. Dasatinib and ponatinib increased adhesion of human platelets to HCAEC, specifically when ECs, but not platelets, were exposed to drugs. Dasatinib, ponatinib, and nilotinib impaired HCAEC healing in vitro, whereas only nilotinib impaired healing in vivo and increased von Willebrand factor levels in mice. Dasatinib and ponatinib increased early platelet but not fibrin accumulation in the mouse cremaster arteriole laser injury–induced thrombosis model, and only ponatinib increased platelet-leukocyte aggregate formation. Asciminib had no toxic effect in any of these assays, similar to imatinib. These studies reveal novel and distinct mechanisms by which EC damage induced by dasatinib, ponatinib, and nilotinib contributes to a prothrombogenic EC state. The findings suggest the need for distinct side effect prevention strategies and provide preclinical data supporting vascular safety of asciminib, while awaiting long-term vascular safety follow-up results.

Introduction

The development of Abl tyrosine kinase inhibitors (Abl-TKI) that directly target the causative BCR::ABL (Philadelphia chromosome) translocation has dramatically improved survival in Philadelphia chromosome–positive leukemias, especially in chronic myeloid leukemia (CML). Thus, since the first Abl-TKI imatinib was introduced, the number of patients on chronic Abl-TKI therapy has risen substantially.¹ In CML, leukemic cells develop resistance to imatinib in up to 40% of patients, leading to the development of new agents, such as dasatinib, ponatinib, and nilotinib, with superior cancer treatment efficacy.^2,3 Although more effective at achieving CML remission, long-term follow-up revealed that dasatinib, ponatinib, and nilotinib are associated with a threefold to fourfold increased risk of arterial thrombotic events compared to imatinib, which limits their clinical utility and subjects cancer survivors to higher cardiovascular morbidity and mortality.² For example, in a trial of ponatinib plus chemotherapy for acute lymphoblastic leukemia, the dose of ponatinib had to be reduced after some patients experienced chest pain and sudden death, attributed to myocardial infarction.⁴ Asciminib, the first agent to target the myristoyl pocket of Abl kinase rather than the ATP-binding site, was recently approved for frontline therapy for CML based on improved efficacy compared to imatinib in the ASC4FIRST trial.³ Thus, asciminib use is expanding rapidly in the absence of long-term cardiovascular safety data. As with imatinib and all other Abl-TKI, resistance to asciminib is already emerging; hence, the agents associated with arterial thrombosis continue to be used clinically.⁵ It is thus imperative to understand the pathophysiologic mechanisms driving increased risk of arterial thrombosis with dasatinib, ponatinib, and nilotinib to better predict and prevent cardiovascular side effects and determine the cardiovascular safety of asciminib and other emerging Abl-TKI.

Initial studies examining the effects of dasatinib, ponatinib, and nilotinib on platelet function yielded disparate results that do not fully explain the high risk of arterial thrombosis. Dasatinib, for example, impairs platelet function and has a US Food and Drug Administration warning due to increased risk for gastrointestinal hemorrhage.^6,7 Consistent with this, in preclinical studies dasatinib reduced ferric chloride–induced thrombosis formation in mouse mesenteric arterioles compared to vehicle, nilotinib, and ponatinib.⁸ For ponatinib, some studies show an increase in human and mouse platelet activation and increased thrombosis and thrombotic microangiopathy in mice.8, 9, 10 However, other studies have shown that ponatinib inhibits signaling and function of human platelets.^11,12 Nilotinib has also been shown to increase thrombosis in mouse models,^8,13 yet, it does not appear to have a direct effect on platelet function ex vivo.^12,14 The reproducible inhibitory effects of dasatinib on platelet function in vivo and the conflicting data regarding the effects of nilotinib and ponatinib on platelets suggest that platelet-specific effects may not be the sole driver of the increased risk of arterial thrombosis. Rather, recent studies implicate a proinflammatory impact on the endothelium.15, 16, 17

Healthy endothelial cells (EC) are anti-inflammatory and antithrombotic, whereas bidirectional activation of EC, platelets, and leukocytes has been implicated in many proinflammatory and prothrombotic conditions, such as antiphospholipid antibody syndrome, lupus, sickle cell disease, and COVID-19 coagulopathy.18, 19, 20, 21 Markers of EC dysfunction, including increased plasma levels of soluble forms of inflammatory EC adhesion molecules, have been associated with an increased risk of arterial thrombosis and have been observed in patients on dasatinib, ponatinib, and nilotinib.^13,22 In addition, multiple studies have shown that dasatinib, ponatinib, and nilotinib also cause damage to EC in vitro, including induction of EC apoptosis, expression of proinflammatory adhesion molecules, actin remodeling, and cell-cell junction disruption.^16,17,23 The potential effect of Abl-TKI on EC are compounded by the fact that the average age at CML diagnosis is 64 years; hence, 64% of patients have a preexisting diagnosis of hypertension, diabetes, or atherosclerotic heart disease when starting Abl-TKI therapy, all of which also are associated with EC dysfunction.²⁴ Platelet-leukocyte aggregates (PLA) also have been identified as biomarkers of thrombotic risk in thromboinflammatory disorders.²⁵ Patients with myeloproliferative disorders including CML have increased PLA compared to healthy controls.²⁶ We recently demonstrated that ponatinib has proinflammatory effects on EC, inducing atherosclerotic plaque inflammation and rupture in mouse atherosclerosis models, causing tumor necrosis factor receptor–dependent induction of ICAM1 on EC, inducing EC-leukocyte interactions in vitro and in vivo, and increasing PLA formation in mice, whereas imatinib and asciminib did not.^15,16 However, the prothrombotic EC-driven mechanisms by which ponatinib, dasatinib, and nilotinib may activate platelets and induce thrombosis are understudied.

In this study, we explore the impact of Abl-TKI in vitro in primary human coronary artery EC (HCAEC) and in vivo in mice, assessing EC functions that can contribute to arterial thrombosis. Direct comparison was made between the impact of imatinib, dasatinib, ponatinib, nilotinib, and asciminib, each at concentrations consistent with the plasma levels in human patients with cancer, using primary human cells that are relevant to the clinical finding of increased coronary artery thrombosis. We show for the first time, to our knowledge, the impact of the 5 drugs compared to vehicle control on adhesion of human platelet to HCAEC in vitro, as well as on EC healing after wire injury, thrombosis after laser EC injury, and PLA formation in vivo. We identify distinct prothrombotic EC toxicities and mechanisms depending on the TKI used in these assays. These findings suggest the clinical phenotype of increased arterial thrombosis with dasatinib, ponatinib, and nilotinib may arise from distinct pathophysiologic mechanisms, which have important implications for prevention and treatment. We also show for the first time, to our knowledge, that asciminib does not increase platelet adhesion to HCAEC in vitro nor does it impair carotid EC healing or increase platelet or fibrin accumulation in a well-established mouse model of EC injury–induced thrombosis.

Methods

Detailed experimental methods

For detailed experimental methods for all studies, please see supplemental Methods.

In vitro and in vivo Abl-TKI dosing

For all in vitro assays, cells were treated with Abl-TKI at the steady-state peak concentration for each drug reported in human pharmacokinetic studies (imatinib, 4.0μM; dasatinib, 0.2μM; ponatinib, 0.1μM; nilotinib, 3.0μM; asciminib, 1.1μM).27, 28, 29, 30, 31 For all in vivo assays, mice were treated with imatinib 150 mg/kg daily, ponatinib 3 mg/kg daily, nilotinib 7.5 mg/kg daily, dasatinib 10 mg/kg twice daily, asciminib 7.5 mg/kg twice daily, or vehicle control twice daily via oral gavage at a volume of 6 μL/g per dose dissolved in 10% dimethyl sulfoxide (DMSO) in phosphate-buffered saline. We have previously validated this dosing strategy based on mouse pharmacokinetic studies quantifying steady-state plasma levels of imatinib, dasatinib, and ponatinib measured by liquid chromatography/mass spectrometry.¹⁵ For this study, we further determined oral dosing for dasatinib and nilotinib that achieves plasma levels consistent with the maximum concentration and total exposure in humans (supplemental Figure 1). All drug studies were compared to equal final concentrations of DMSO as the vehicle control for in vitro studies and DMSO in phosphate-buffered saline for in vivo studies.

Drugs and cells

All Abl-TKI were purchased from MedChemExpress. Primary HCAEC from 4 donors (3 male donors aged 51, 25, and 58 years from the Molecular Cardiology Research Institute cell line biobank, originally generated from deidentified tissue from the National Disease Research Interchange as previously described³²; and a female donor aged 36 years [American Type Culture Collection]) were used for all in vitro studies. Only for the platelet-EC adhesion assays, to limit the number of variables and focus on the potential effects of Abl-TKI on platelets from different donors, HCAEC from 1 donor (a 36-year-old woman) were used, with different healthy donor–derived platelets used for each of the 6 experiments.

Mouse studies

All mouse studies were approved by the Tufts University (carotid wire injury and blood collection) or the Beth Israel Deaconess Medical Center (laser-induced thrombosis in mouse cremaster arterioles) Institutional Animal Care and Use Committees. Male C57/BL6J mice (The Jackson Laboratory) were used for all studies to ensure consistency because cremaster arteriole thrombosis studies are limited to male mice.

Statistics

All statistical analyses were performed using GraphPad Prism 10. Shapiro-Wilk testing was used to determine normality, and if data were normally distributed, 1-way analysis of variance with Dunnett post hoc multiple comparison testing was used. For nonnormally distributed data, Kruskal-Wallis multiple comparison testing with Dunn post hoc test was used. For comparison of rates of thrombosis after carotid wire injury, Fisher exact test was used. All post hoc testing compared each Abl-TKI to DMSO control. In a separate analysis, statistical comparison of each TKI was made to imatinib, because all clinical studies revealing increased risk of thrombosis due to nilotinib, dasatinib, and ponatinib were compared to imatinib, given that placebo would be unethical because Abl-TKI are standard of care. This alternate analysis revealed no differences in the statistical findings whether DMSO or imatinib (not shown) was the comparator, consistent with the lack of significant difference between DMSO and imatinib in any assay. Significance was set at P value <.05.

Results

Exposure of HCAEC but not platelets to dasatinib or ponatinib induces platelet-EC adhesion in vitro

We first investigated whether Abl-TKI affect platelet adhesion to EC in vitro (Figure 1). Primary HCAEC were used because myocardial infarction is a life-threatening manifestation of arterial thrombosis due to Abl-TKI treatment.³³ When platelets alone were treated with Abl-TKI and incubated with untreated HCAEC (Figure 1A,D), there was no significant impact of any Abl-TKI on platelet-EC adhesion. When HCAEC alone were treated (Figure 1B,E) or when platelets and HCAEC were both treated with Abl-TKI (Figure 1 C,F), dasatinib and ponatinib significantly increased platelet adhesion compared to DMSO. Neither imatinib, nilotinib, nor asciminib significantly affected platelet-EC adhesion.

Figure 1.

Exposure of HCAEC but not platelets to dasatinib or ponatinib induces platelet-EC adhesion in vitro. (A-C) Representative images of fluorescently labeled healthy human platelets bound to HCAEC after treatment of platelets for 30 minutes before incubation with untreated HCAEC (A); HCAEC for 24 hours before adding untreated platelets (B); or both HCAEC and platelets (C) with Abl-TKI or DMSO (scale bar, 50 μm). (D-F) Quantification of fluorescent platelets adherent to HCAEC for each Abl-TKI relative to average of DMSO controls after treating; platelets only (D), HCAEC only (E), and both HCAEC and platelets (F); error bars represent median with 25% to 75% interquartile range (n = 6 independent experiments). Kruskal-Wallis multiple comparison testing with Dunn post hoc test was used, with comparison to DMSO. ∗P < .05; ∗∗P < .01.

Dasatinib, ponatinib, and nilotinib impair HCAEC scratch wound healing in vitro

Because the prothrombotic Abl-TKI appear to directly affect EC function, we next examined HCAEC healing after causing a scratch wound in vitro (Figure 2). Impaired EC healing increases the duration of exposed basement membrane as a potential nidus for clot formation. Dasatinib, ponatinib, and nilotinib all significantly decreased HCAEC healing after scratch injury compared to DMSO, whereas asciminib and imatinib did not (Figure 2B).

Figure 2.

Dasatinib, ponatinib, and nilotinib impair HCAEC scratch wound healing in vitro. (A) Representative images of HCAEC monolayers immediately after scratch wound (0 hours) and after 20 hours incubation with the indicated Abl-TKI or DMSO control (scale bar, 100 μm). (B) Quantification of percent of the wound healed at 20 hours (n = 8 independent experiments). One-way analysis of variance (ANOVA) with Dunnett post hoc test was used, with comparison to DMSO. ∗∗∗P < .001; ∗∗∗∗P < .0001.

Nilotinib impairs re-endothelialization after wire carotid injury in mice

Next, we compared the impact of Abl-TKI on healing after EC denudation in vivo using the wire carotid artery injury model in mice (Figure 3). Mice were randomized to treatment with each Abl-TKI for 2 days at doses that result in plasma levels consistent with exposure in clinical trials (supplemental Figure 1).¹⁵ Carotid wire injury was performed on day 0 (third day of drug treatment), and near-complete endothelial denudation was confirmed by Evan blue infusion to label the exposed basement membrane in blue. Mice were euthanized 3 or 5 days after injury, and EC healing was quantified by Evan blue. In contrast to our in vitro studies, only nilotinib significantly delayed EC healing compared to imatinib at both day 3 and day 5 after injury (Figure 3E-F). Ponatinib also appeared to slow wound healing, but this was highly variable and hence not significantly different from DMSO. Although the carotid wire injury is not a thrombosis model, we noted an increase in carotid artery thrombotic events in Abl-TKI–treated mice compared to DMSO or historical norms (Figure 3G). Although dasatinib, ponatinib, and nilotinib had the highest rates of thrombosis (partial and complete combined), there were no significant differences between groups. As such, we next switched to a thrombosis model for further evaluation.

Figure 3.

Nilotinib impairs re-endothelialization after wire carotid injury in mice. (A) Mice were treated with DMSO or the indicated Abl-TKI beginning 2 days before left carotid artery injury followed by terminal Evan blue perfusion and imaging of the carotid artery endothelial surface on days 0, 3, and 5 after wire injury. The exposed basement membrane stains blue, whereas the healed area covered in EC is white. (B-C) Representative images of vessels on day 3 (B) and day 5 after injury (C). (D-F) Graph (D) showing quantification of percent healed (mean ± standard error of the mean [SEM]) on days 0, 3, and 5, with statistical analysis on day 3 (E) and day 5 (F) after injury (n = 3-5 mice per group for day 0; n = 16-18 mice per group for day 3; n = 6-7 mice for day 5). One-way ANOVA with Dunnett post hoc test was used, with comparison to DMSO. ∗P < .05; ∗∗P < .001. (G) Representative image of post–carotid injury complete artery thrombosis in situ (red arrow), excised complete thrombosis, and excised partial thrombosis that was washed away during perfusion; percentage of mice (right) from all injuries with completely occluded (dark brown bars), partially occluded (light brown bars), and patent vessels (yellow bars). There were no significant differences between groups by Fisher exact test (P = .31). Asc, asciminib; Das, dasatinib; Ima, imatinib; Nil, nilotinib; Pon, ponatinib.

Ponatinib and dasatinib increase early platelet, but not fibrin, accumulation in thrombi after laser-induced EC injury in vivo

To interrogate the impact of Abl-TKI on thrombosis after EC injury in vivo, the mouse cremaster arteriole laser injury model was used, with fluorescent labeling of platelets in green and fibrin in red (Figure 4). This thrombosis model has been used extensively to study the effects of EC injury and platelet-EC interactions in arterial thrombosis.^34,35 Dasatinib and ponatinib both significantly increased platelet accumulation after 20 seconds (Figure 4D, left panel). Ponatinib had a continued impact on platelet adhesion, with a significant increase compared to vehicle also after 40 and 60 seconds (Figure 4D, right and middle panels). There were no significant differences in fibrin accumulation between any of the treatments (Figure 4E). Once again, asciminib did not affect platelet or fibrin accumulation relative to DMSO control.

Figure 4.

Ponatinib and dasatinib increase early platelet but not fibrin accumulation in thrombi after laser injury to EC. Mice were treated with indicated Abl-TKI or DMSO control for 48 hours before laser-induced cremaster arteriole endothelial injury. (A) Representative intravital microscopy images of platelet (green) and fibrin (red) accumulation in thrombi at 0, 15, 30 and 60 seconds after laser injury (scale bar, 20 μm). Median fluorescent intensity of platelet (B) and fibrin (C) accumulation over 60 seconds after laser injury. (D) Quantification of area under the curve (AUC) (on log₁₀ scale) for platelet accumulation for each injury after 20, 40, and 60 seconds. (E) Quantification of AUC for fibrin accumulation (on log₁₀ scale) for each injury after 20, 40, and 60 seconds (n = 26-32 injuries in 4-5 mice per condition). Kruskal-Wallis multiple comparison testing with Dunn post hoc test was used, with comparison to imatinib. ∗P < .05; ∗∗P < .01. AUC, area under the curve; RFU, relative fluorescence units.

Ponatinib increases PLA formation in mice, and nilotinib increases VWF levels in mouse plasma and in HCAEC lysates

Ponatinib has previously been shown cause a thromboinflammatory phenotype in mice, including increased PLA formation,³⁶ and von Willebrand factor (VWF) has been suggested as a biomarker for drug-induced EC damage³⁷ and plays a major role in platelet adhesion to exposed basement membrane. Thus, we compared the effects of the 5 Abl-TKI on PLA formation and on VWF levels (Figure 5). After 3 days of treatment in mice, ponatinib significantly increased circulating PLA measured by flow cytometry compared to DMSO (Figure 5A-B), suggesting that only ponatinib and not dasatinib or nilotinib increases adhesion of platelets to leukocytes (in addition to EC; as shown in Figure 1). In contrast, only nilotinib increased plasma VWF levels in mice (Figure 5C) and increased the amount of VWF in HCAEC lysates (Figure 5D). There were not significant differences in VWF secreted into supernatant after 24 hours (supplemental Figure 2).

Figure 5.

Ponatinib increases PLA formation in mice, and nilotinib increases VWF levels in mouse plasma and HCAEC. (A) Representative whole blood flow cytometry plots from mice treated for 3 days with DMSO or indicated Abl-TKI showing gating for CD45⁺ cells (leukocytes; left) and gating for CD41⁺ leukocytes (PLA) in ponatinib- (middle) and DMSO-treated mice (right). (B) Quantification of the percentage of PLA for each treatment (n = 4 mice per treatment group). One-way ANOVA with Dunnett post hoc test was used, with comparison to DMSO. (C) Quantification of mouse plasma VWF after 3 days of treatment with the indicated Abl-TKI (n = 20-22 mice per treatment). Kruskal-Wallis ANOVA with Dunn multiple comparisons test to DMSO was used. (E) Level of VWF protein in lysates from HCAEC after 24 hours of treatment. One-way ANOVA with Dunnett post hoc test was used, with comparison to DMSO (n = 4 independent experiments). ∗P < .05; ∗∗∗∗P < .001. SSC-H, side scatter pulse height.

Discussion

This study directly compared the impact of 5 approved Abl-TKI (imatinib, dasatinib, ponatinib, nilotinib, and asciminib) at clinically relevant concentrations, both in vitro using primary HCAEC and in vivo in mouse models of EC injury and thrombosis. The findings identify prothrombotic EC-dependent mechanisms that may underlie the high rate of arterial thrombotic events that limit the clinical utility of dasatinib, ponatinib, and nilotinib. Overall, our data demonstrate the following: (1) treatment of HCAEC with dasatinib or ponatinib, but not nilotinib, imatinib, or asciminib, increases platelet adhesion to EC in vitro, whereas platelet treatment alone is insufficient to increase platelet-EC adhesion; (2) dasatinib, ponatinib, and nilotinib, all of which are associated with thrombotic risk in patients, but not imatinib or asciminib, impair HCAEC scratch wound healing in vitro; (3) only nilotinib inhibits EC healing after wire carotid injury in mice and increases plasma VWF levels in mice and VWF expression in HCAEC; (4) dasatinib and ponatinib also increase platelet accumulation, but not fibrin accumulation, during early thrombogenesis induced by cremaster artery EC laser injury in mice; and (5) only ponatinib increases the formation of PLA in mice. Together, these data reveal that the 3 Abl-TKI that increase thrombosis risk in patients also induce EC damage that mediates platelet adhesion in vitro and prothrombogenic conditions after EC injury in mice, However, this occurs via drug-specific mechanisms. Conversely, imatinib, which does not induce thrombosis in humans, did not have any effect on HCAEC function nor did it promote EC damage or thrombosis in mice compared to DMSO control. Finally, asciminib, a newly approved allosteric Abl kinase inhibitor without long-term vascular follow-up data in humans, behaves similar to imatinib in all the assays tested, supporting that asciminib may have an arterial thrombosis safety profile similar to imatinib.

This study examined for the first time, to our knowledge, the effects of Abl-TKI on HCAEC healing in vitro and EC healing in vivo in mice. Prior studies have shown that dasatinib, ponatinib, and nilotinib all impair EC scratch wound healing in vitro; however, those studies used human umbilical vein EC (HUVEC), which may be less relevant to the concerning clinical outcomes in patients.^16,17 Here, we confirm that dasatinib, ponatinib, and nilotinib also inhibit scratch wound healing in HCAEC, a primary human cell line that is relevant to the clinical manifestation of myocardial infarction, one of the most morbid conditions resulting from arterial thrombosis with these agents. However, when examined in vivo, only nilotinib significantly decreased EC healing after wire injury in the carotid artery in mice. In prior studies in HUVEC, nilotinib uniquely increased EC necrosis, whereas dasatinib and ponatinib impaired HUVEC migration and viability,^16,38 suggesting that direct EC toxicity and necrosis, rather than impairment of EC proliferation and migration, may be the prominent factors affecting EC healing in vivo.

Prior studies examining the impact of Abl-TKI on thrombosis in mice have shown increased thrombosis using the ferric chloride and rose bengal models to initiate thrombus formation.^9,13 Both of those models cause EC death and denudation, with exposure of the basement membrane as a nidus for clot formation, which makes studying platelet-EC interactions more difficult in those models.^39,40 Here, the EC laser injury model of thrombosis in mouse cremaster arterioles was used to induce thrombus formation, because this model largely leaves the EC layer in place, allowing for better understanding of the effect of Abl-TKI on platelet-EC interactions in the setting of EC injury. The findings of enhanced early platelet accumulation in mice treated with dasatinib and ponatinib are consistent with the findings from our in vitro platelet-EC adhesion studies in HCAEC, showing translation in vivo. Furthermore, the discovery that dasatinib and ponatinib affect platelet accumulation but not fibrin accumulation further supports that platelet-EC interactions, rather than procoagulant effects, are a critical driver of the thrombosis risk associated with Abl-TKI treatment. These in vitro and in vivo results are novel and consistent with the clinical side effects in patients, in whom thrombotic adverse events are mostly arterial, manifesting as coronary, cerebral, or peripheral arterial occlusion, rather than venous thromboses.

Interactions between EC, platelets, and leukocytes play a prominent but complex role in the pathophysiology of thrombosis in a variety of clinical conditions.^41,42 Recently published data revealed that ponatinib acts on EC and leukocytes to promote vascular inflammation that contributes to atherosclerotic plaque inflammation and rupture.¹⁵ The data in this study advance our understanding of how Abl-TKI can alter platelet interactions in response to injury to the endothelium in ways that lead to increased risk of arterial thrombosis, independent of or in addition to atherosclerosis complications. Both dasatinib and ponatinib increased platelet-EC adhesion in vitro only when EC were treated, providing support that EC toxicity, rather than direct platelet activation, is a main driver of this component of the prothrombotic phenotype of these Abl-TKI. By directly comparing the impact of all 5 Abl-TKI on PLA formation, we determined that ponatinib alone significantly increased PLA formation in mice. Increased PLA have been seen in a diverse cardiovascular and prothrombotic disorders, including acute myocardial infarction, COVID-19 infection, and chronic thromboembolic pulmonary hypertension.43, 44, 45 Because there are currently no biomarkers available to predict the risk of thrombosis with ponatinib, measuring PLA might be examined as a potential method to monitor arterial thrombosis risk. Future clinical studies could determine whether PLA could be followed in patients who require ponatinib treatment and whether PLA-informed dose-reduction or treatment cessation could prevent devastating arterial thrombotic events. Conversely, other studies have revealed a platelet-inhibitory effect of dasatinib,^11,12,46 and consistent with that finding, our studies show that dasatinib did not directly affect platelet adhesion in vitro nor increase PLA in vivo. This suggests that antiplatelet therapy might not be the preferred way to prevent or treat arterial thrombosis in patients on dasatinib and rather treatments aimed at inhibiting platelet-EC interactions and improving endothelial function may be more beneficial. Indeed, therapeutics targeting EC function and inflammation continue to be developed for inflammatory and thrombotic diseases and could be tested as EC-protective agents in ponatinib- or dasatinib-treated patients with cancer.^47,48

Some limitations in these studies should be acknowledged. First, although we have used concentrations and doses consistent with maximum concentrations seen in patients on these medications, dose modifications are common in patients on later generation Abl-TKI,^49,50 and it may be that some of these effects are only seen at the clinical higher doses. This would be consistent with lower rates of arterial thrombotic events in patients on lower doses of ponatinib in patients with acute lymphoblastic leukemia.⁴ Furthermore, mouse pharmacokinetics of Abl-TKI are different from larger species such as monkeys,⁵¹ and it might be that the failure to recapitulate the in vitro findings of impaired EC wound healing in mice might be secondary to different EC exposure to drug over time compared to the consistent levels used in vitro, which may also differ from humans. Because there is no reported sex difference in arterial thrombosis risk in Abl-TKI–treated patients, we only used male mice in in vivo studies, some of which (cremaster arteriole studies) can only be performed in males. Future studies should include female animals to clarify whether sex differences might exist. Finally, we chose a short pretreatment course of 2 days before inducing EC injury in our mouse models because this was recently shown to be sufficient to achieve steady-state drug levels consistent with human patients and induce EC damage in vivo.¹⁵ However, patients with leukemia are treated for extended periods, and thus, although we observed differences in the prothrombogenic mechanisms for each Abl-TKI even with brief treatment in vivo, prolonged treatment before EC injury could provide additional information into the mechanisms linking EC dysfunction and thrombosis for each agent.

Despite these limitations, these results have important clinical implications. The findings support distinct mechanisms behind the risk of thrombosis with dasatinib, ponatinib, and nilotinib. Dasatinib-induced EC dysfunction resulted in increased platelet-EC adhesion, whereas ponatinib increases both EC-platelet adhesion and PLA formation, and nilotinib causes impaired EC healing in vivo and increases VWF expression in EC and in mouse plasma. The finding that distinct mechanisms may be leading to the same clinical side effects for these agents suggests that strategies to prevent or treat arterial thrombosis might need to be specific to each Abl-TKI. Because all agents inhibit Abl kinase in EC,¹⁶ these data suggest off-target kinases likely mediate the distinct EC toxicities of Abl-TKI, consistent with overlapping yet distinct array of kinases inhibited by each agent.¹⁷ Prior studies have shown that a global phosphoproteomic signature can differentiate Abl-TKI that are associated with the risk of thrombosis from those that do not have that side effect,⁵² and gene expression profiling demonstrates both common and distinct gene alterations induced by dasatinib, ponatinib, and nilotinib in HUVEC.²³ Future studies will be needed to further clarify the detailed molecular pathways leading to the distinct toxicity mechanisms to identify drug-specific protective strategies. Another important implication of this work is that asciminib did not increase platelet adhesion to HCAEC in vitro nor did it increase platelet accumulation during laser EC injury–induced thrombus formation in vivo. This is, to our knowledge, the first study to compare asciminib to other Abl-TKI in an in vivo thrombosis model. The results strengthen the link between Abl-TKI–induced EC dysfunction (or the lack of EC damage by asciminib) and the associated risk of thrombosis and show, to our knowledge, for the first time that asciminib does not induce a prothrombotic phenotype in vivo in mice. This is especially important because asciminib use is growing as a single-agent therapy, and investigation is underway for combination treatment with other Abl-TKI for high-risk patients, due to the novel mechanism of action via allosteric inhibition of Abl kinase.^53,54 Dasatinib, ponatinib, and nilotinib will remain as later-line treatment options, even if long-term follow-up confirms the vascular safety of asciminib, especially in patients with high-risk disease or patients who develop resistance or noncardiovascular side effects to asciminib or other Abl-TKI.⁵⁵ There are also Abl-TKI, such as olverembatinib, that are in development to treat patients with resistance to all currently approved Abl-TKI, including asciminib, and there is a growing interest in using multiple Abl-TKI in combination therapy.^53,54,56 Therefore, understanding the mechanisms behind the increased risk of arterial thrombosis with each distinct Abl-TKI will continue to have clinical relevance in predicting, preventing, and treating arterial thrombosis in patients and guiding individual or combination Abl-TKI decision-making in patients in the future.

Conflict-of-interest disclosure: C.S.C. is a founder and owns shares of Satellite Biosciences and Ropirio Therapeutics, which do not have interests in relation to the current study. The remaining authors declare no competing financial interests.