Introduction
Interference with vascular endothelial growth factor (VEGF) signaling and subsequent tumor neovascularization by the use of targeted agents has shown beneficial effects for patients with various tumor types [
1‐
5]. Specific antibody-based VEGF inhibitors, like bevacizumab, have been developed, as well as small-molecule antiangiogenic multityrosine kinase inhibitors (TKIs) interfering with VEGF signaling, like sunitinib and sorafenib. It was not expected that these antiangiogenic agents would cause severe toxicities, as merely non-quiescent endothelial cells (ECs) present in the tumor would be disturbed [
6]. However, several (major) bleeding complications, such as subungual splinter hemorrhage, epistaxis, and gastrointestinal, pulmonary, and intracerebral bleedings, have been clinically observed (all-grade bleeding events: sunitinib: 19.3%, sorafenib: 13.5%, bevacizumab: 25.0–30.4%; high-grade bleeding events: sunitinib: 3.0%, sorafenib: 2.2%, bevacizumab: 2.8–3.5%) [
7‐
14]. Of the observed bleeding events, splinter bleedings are more frequently seen during TKI treatment.
The underlying mechanisms of these treatment-related bleeding complications have not been elucidated yet.
For adequate coagulation, coagulation factors and enzymes, ECs and platelets (thrombocytes) are important [
15]. Platelets play a specific role in clot formation, because they immediately adhere to the damaged endothelial layer of the vessel wall upon detection of a breach, where they become activated, aggregate and release their content [
15,
16]. Relatively small differences in this orchestrated interplay may cause significant disturbances leading to bleeding complications.
It has become apparent that platelets not only play a role in arrest of bleeding but are also involved in vascular integrity and vessel repair [
17,
18]. Since VEGF is crucial for maintaining the integrity of the microvasculature, it has been postulated that blockade of the VEGF pathway leads to compromised capacity of ECs for cell repair [
17,
19,
20], which has been proposed as a mechanism for the observed vascular toxicity. Previously, we found that VEGF is transported by platelets and released upon platelet activation [
21]. In addition, it became apparent that bevacizumab is taken up by platelets and neutralizes circulating platelet-VEGF [
22]. Based on these findings and the clinically observed bleeding complications, we hypothesized that antiangiogenic treatment may disturb platelet function which may be the cause behind the bleeding complications. To study this hypothesis, we evaluated the influence of the antiangiogenic agents sunitinib, sorafenib, and bevacizumab on platelet function both in vitro and ex vivo using platelets from patients with renal cell carcinoma (RCC) and non-small cell lung carcinoma (NSCLC).
Materials and methods
Detailed information is described in supplementary Materials and methods.
Healthy volunteers
Informed consent was obtained before blood collection. Healthy volunteers did not take any medication in the prior 10 days. In vitro experiments were performed to study the effect of sunibitinib (10 or 20 μM), sorafenib (5, 10, 25 μM), and bevacizumab (50, 100 or 250 μg/ml) on platelet aggregation. Clinically achieved concentrations for sunitinib (50 mg/day) are approximately 65 ng/ml (0.16 μM) [
23], for sorafenib (400 mg bid) around 5 mg/l (10.7 μM) [
24,
25], and for bevacizumab (10 mg/kg) up to 300 μg/ml [
26]. We chose these high concentrations with regard to the relative short incubation time. Furthermore, the influence of the angiogenesis inhibitors was examined on platelet–EC adhesion (measured by real-time perfusion) and on surface P-selectin expression and fibrinogen binding to αIIbβ3 (measured by flow cytometry). Finally, tyrosine phosphorylation of Src was studied in the presence of both TKIs sunitinib and sorafenib (measured by western blotting).
Additional information regarding used reagents and antibodies and the methods used is described in supplementary materials and methods.
Patients
Blood was drawn by venipuncture from patients with RCC before and during treatment with sunitinib and in patients with NSCLC before and after a single administration of bevacizumab. The studies were approved by the institutional review boards. Patients were required to sign informed consent prior to participation.
Time points for collection of blood in patients treated with sunitinib were: pretreatment, 24 h, 3 weeks, and 6 weeks after start of treatment. Platelet aggregation experiments and measurements of the concentrations of sunitinib [measured by liquid chromatography–tandem mass spectrometry (LC–MS/MS)], of VEGF, and of the activation markers of platelets and ECs [measured by enzyme-linked immunosorbent assay (ELISA)] were performed at the different time points. Treatment-related toxicity was reported during the first 6 weeks of treatment with sunitinib. Common Terminology Criteria for Adverse Events (CTCAE) were used to grade bleeding events.
Treatment with chemotherapy or biologicals was not allowed in the previous 28 days before start sunitinib.
Time points for collection of blood samples in patients treated with bevacizumab were: before, and 5 h and 3 to 5 days after the administration of bevacizumab. Prior to their scheduled therapy, these patients received a single dose of 15 mg/kg bevacizumab as part of an imaging study [
27]. In this study, a week prior to infusion patients underwent a dynamic PET-CT study with both [15O]H2O and [11C]docetaxel. Platelet aggregation was performed at the different time points.
Additional information regarding the methods used is described in supplementary methods.
Statistical analysis
For statistical analysis, SPSS (version 21.0; SPSS INC., Chigaco, IL, USA) was used. Paired t tests were performed to compare the absolute values of aggregation levels, the concentration of platelet and EC activation markers, the VEGF concentrations and the platelet counts over time, and the expression of P-selectin on the platelet membrane and the fibrinogen binding to GPIIb/IIIa. Independent-sample t tests were performed to compare impaired platelet aggregation in patients with and without bleeding events. With the Pearson correlation coefficient, the concentrations of sunitinib in plasma and serum and the changes in platelet counts were correlated with impaired platelet aggregation. In addition, the concentrations of platelet activation markers were correlated with platelet counts. Defined significance level is p < 0.05. Reported p values are two-sided.
Discussion
In this study, we investigated the hypothesis that targeted agents inhibiting VEGF signaling may disturb the function of platelets, thereby contributing to the observed treatment-related bleeding complications. We found in vitro that platelet aggregation, induced by collagen or ADP, is reduced by the TKIs sunitinib and sorafenib. Sunitinib impaired platelet function in patients as well, most likely due to a direct effect on platelets as no correlation was found with decreased platelet count (a known side-effect) [
23]. In contrast, no effect of bevacizumab (monoclonal antibody against VEGF) on platelet aggregation was detected in patients, and only high concentrations had an inhibitory effect in vitro.
Within 24 h after patients started treatment with sunitinib, platelet aggregation was impaired by almost 50%. Approximately 35% of the patients (6 out of 17) experienced a bleeding complication within the first 6 weeks of sunitinib treatment. All-grade bleeding events observed in our study were higher than the percentages reported in the phase III study in patients with RCC (12%) [
3] and in the phase III trial in GIST patients (7%) [
4]. Interestingly, approximately 12% (2 out of 17) patients presented with a grade ≥ 3 bleeding event, which is remarkably higher than previously reported [
3]. For this increased risk of high-grade bleeding complications, several possible explanations can be suggested: the bias introduced by the limited number of participating patients, or conversely, our data reflect more the frequency among real-world patients. An additional factor may be the difference in grading of bleeding events [
12].
In vitro, no effect of sunitinib, sorafenib, or bevacizumab was observed on platelet aggregation induced by thrombin, arachidonic acid, or ristocetin/vWF. This might be explained by the fact that part of the collagen signaling pathway is routed via ADP signaling. Our results of the observed effect of sunitinib on platelet aggregation induced by collagen or ADP, but not by thrombin or arachidonic acid, are compliant with recently published data by Sabrkhany et al. [
29].
To examine whether the impaired platelet aggregation is caused by a defect in platelet aggregation or platelet secretion, we analyzed the amount of P-selectin expressed on the membranes of activated platelets and of fibrinogen binding to GPIIb/IIIa. Significantly impaired expression of P-selectin was observed in the presence of sorafenib and to lesser extent of sunitinib, as compared to control conditions. Fibrinogen expression was only significantly reduced by sorafenib after platelet activation by PAR1-activating peptide. Bevacizumab had no effect on P-selectin expression or fibrinogen binding. These results indicate that disturbed secretion of the platelet content contributes to the decline in platelet aggregation [
30]. In the patients treated with sunitinib, plasma concentrations of activation markers of platelets (P-selectin, beta-TG, RANTES) and of ECs (vWF) within the first six treatment weeks remained similar, while in contrast to the in vitro experiment, these platelets were not activated. These results might suggest an intact interaction between platelets and activated ECs, which is further supported by the in vitro observation that incubation of platelets with sunitinib or bevacizumab did not alter platelet adherence to ECs. This observation is indicative for the complex differential biology of platelet aggregation and platelet adherence to the endothelial cell layer of the vascular wall.
Interestingly, serum concentrations of sunitinib were higher compared to plasma at 24 h and 3 weeks after start of treatment. These results indicate that intracellular accumulation of sunitinib occurs not only in cancer cells [
31] but in platelets as well [
29]. Due to accumulation, it may exert direct effects on platelet signaling, which might resemble data concerning the effect of other TKIs on platelet signaling [
16,
32‐
36].
While high concentrations of bevacizumab had some inhibitory effect on platelet aggregation in vitro, no significant inhibition was observed with platelets of patients up to 5 days after administration of the highest approved dose of bevacizumab. These findings may provide further insight in antiangiogenic treatment-related toxicities, and in the interplay of platelets and ECs to maintain vascular integrity. The observation that bevacizumab did not affect platelet function is in line with a previous report [
37], as a dose range of bevacizumab in vitro failed to affect platelet aggregation. However, the effect of bevacizumab on platelet aggregation still remains controversial. Meyer et al. [
38] concluded that bevacizumab can induce platelet aggregation and consequently degranulation through complex formation with VEGF and activation of the platelet FcγRIIa receptor, postulating that this might explain the observed thrombotic events. One possible explanation for these contradictory results could be artificial differences and bias introduced by the different pre-analytical and laboratory methods used for the analyses, such as used type of needle, type of anticoagulant, type of centrifuge and assay method, temperature during preparation, and the center performing the analysis [
39,
40]. These discrepancies underline the undoubtedly intricate and multifactorial mechanisms that can cause vascular events in patients with cancer.
Significant changes in VEGF concentrations in patients were observed during a treatment cycle of sunitinib, which is in line with a previous report [
41], in which the authors speculate that this effect is secondary to increased activity of HIF-1α, leading to treatment-related increases in tumor hypoxia. We have previously reported that VEGF is almost completely neutralized in PRP of patients after a single administration of bevacizumab [
27]. Now we report that bevacizumab does not significantly inhibit platelet aggregation in these patients, indicating that VEGF has a limited role in the process of platelet aggregation. This finding is supported by previous reports indicating that only very high VEGF concentrations can contribute to platelet aggregation [
42]. In contrast, several TKIs targeting other pathways than the VEGF signaling pathway have been shown to exert a negative effect on platelet aggregation as well [
16,
32‐
36]. In one of these reports, it was suggested that inhibition of Src family kinases (SFKs), which are among other signaling proteins involved in platelet activation, might play an important role by affecting immunoreceptor tyrosine-based activation motif (ITAM) signaling [
34]. Sunitinib is a very potent c-Src inhibitor, with approximately 60% inhibition at a concentration of 0.5 μM [
43]. This effect might be further potentiated by increased intracellular concentrations, resulting in even more profound inhibition and therefore pronounced impairment of platelet aggregation. We detected lower tyrosine phosphorylation in platelets upon TKI incubation compared to control conditions. More studies are required to elucidate whether the influence of TKIs on tyrosine phosphorylation might underlie impaired platelet aggregation resulting in bleeding events.
To conclude, treatment with sunitinib significantly inhibits platelet aggregation in patients already from the first day of treatment, while no significant inhibition was observed by bevacizumab. We found potential evidence that impaired platelet function due to antiangiogenic TKI treatment might play a role in clinically observed bleedings. This insight might contribute to the development of new targeted agents with a reduced risk of treatment-related toxicity.