Comparison of inhibitory effects of irreversible and reversible Btk inhibitors on platelet function

Abstract All irreversible Bruton tyrosine kinase (Btk) inhibitors including ibrutinib and acalabrutinib induce platelet dysfunction and increased bleeding risk. New reversible Btk inhibitors were developed, like MK‐1026. The mechanism underlying increased bleeding tendency with Btk inhibitors remains unclear. We investigated the effects of ibrutinib, acalabrutinib and MK‐1026 on platelet function in healthy volunteers, patients and Btk‐deficient mice, together with off‐target effects on tyrosine kinase phosphorylation. All inhibitors suppressed GPVI‐ and CLEC‐2‐mediated platelet aggregation, activation and secretion in a dose‐dependent manner. Only ibrutinib inhibited thrombus formation on vWF‐co‐coated surfaces, while on collagen this was not affected. In blood from Btk‐deficient mice, collagen‐induced thrombus formation under flow was reduced, but preincubation with either inhibitor was without additional effects. MK‐1026 showed less off‐target effects upon GPVI‐induced TK phosphorylation as compared to ibrutinib and acalabrutinib. In ibrutinib‐treated patients, GPVI‐stimulated platelet activation, and adhesion on vWF‐co‐coated surfaces were inhibited, while CLEC‐2 stimulation induced variable responses. The dual inhibition of GPVI and CLEC‐2 signalling by Btk inhibitors might account for the increased bleeding tendency, with ibrutinib causing more high‐grade bleedings due to additional inhibition of platelet‐vWF interaction. As MK‐1026 showed less off‐target effects and only affected activation of isolated platelets, it might be promising for future treatment.

platelet activation, and adhesion on vWF-co-coated surfaces were inhibited, while CLEC-2 stimulation induced variable responses. The dual inhibition of GPVI and CLEC-2 signalling by Btk inhibitors might account for the increased bleeding tendency, with ibrutinib causing more high-grade bleedings due to additional inhibition of platelet-vWF interaction. As MK-1026 showed less off-target effects and only affected activation of isolated platelets, it might be promising for future treatment.

K E Y W O R D S
bleeding, Bruton tyrosine kinase, CLEC-2, glycoprotein VI, platelets, von Willebrand Factor
Recently, multiple other Btk inhibitors have been developed, including the irreversible Btk inhibitor acalabrutinib and the reversible Btk inhibitor MK-1026 (formerly known as ARQ-531) [13,14]. Opposed to data suggesting that acalabrutinib shows less off-target effects, it still causes an increased bleeding tendency in patients, albeit showing mainly mild bleeding [15]. Studies investigating GPVI inhibitors as potential antithrombotic drugs have shown that inhibition on receptor level does not impair haemostasis [16]. Also, GPVI-deficient patients exhibit only a mild bleeding diathesis [17], and many potential patients are undiagnosed as they remain asymptomatic [18]. In agreement, studies in mice have shown that GPVI deficiency alone does not impair haemostasis [19][20][21], while combined GPVI and CLEC-2 deficiency does [21]. Furthermore, Btk was shown to be involved in GPIb-vWFmediated platelet adhesion, as well as integrin α IIb β 3 outside-in activation which were inhibited by ibrutinib [6,22,23]. Therefore, the current mechanism of increased bleeding tendency in patients using Btk inhibitors is still unclear and underlines the importance for treatment decisions and the development of novel Btk inhibitors. We therefore investigated the effect of multiple Btk inhibitors on platelet function pathways in healthy volunteers, patients and Btk knock-out (KO) mice.

MATERIALS AND METHODS
Materials and methods are available in Data S1.

Ibrutinib and acalabrutinib inhibit platelet activation and aggregation by GPVI and CLEC-2 in a dose-dependent manner
Treatment with ibrutinib, but also acalabrutinib, can increase bleeding tendency [6,15]. To assess the underlying pathways, we investigated platelet activation mediated by multiple agonists in presence of ibrutinib or acalabrutinib.
Both ibrutinib and acalabrutinib inhibited aggregation of washed platelets and platelet-rich plasma induced by 1 μg/ml collagen in a dose-dependent manner ( Figure 1A). In isolated platelets, a significant inhibition was observed at 1 μM and higher for both compounds ( Figure 1Ai). In the presence of plasma, a higher concentration of 5-10 μM was required to significantly inhibit collagen-induced platelet aggregation ( Figure 1Aii). The IC 50 values for ibrutinib and acalabrutinib were 7.7 and 6.0 times lower in washed platelets as compared to PRP (Table S1). At a higher collagen concentration, aggregation was not affected by either inhibitor. ( Figure S1Aii). CLEC-2-induced platelet aggregation was also inhibited by both acalabrutinib and ibrutinib in a similar dose-dependent manner ( Figure 1B). GPVI-induced integrin α IIb β 3 activation, P-selectin expression and PS-exposure were also dose-dependently reduced by both inhibitors, although lower concentrations of ibrutinib (<1 μM) resulted in significant inhibition as compared to acalabrutinib ( Figures 1C and 1D). The IC 50 values of ibrutinib for the inhibition of integrin α IIb β 3 activation, P-selectin expression and PS exposure were 5.0, 8.2 and 2.2 times lower, respectively, than for acalabrutinib (Table S1). Furthermore, both compounds did not alter the peak level of intracellular Ca 2+ elevations, but the slope of this response was reduced in ibrutinib-treated platelets ( Figure 1E).

3.2
Ibrutinib, but not acalabrutinib, impairs thrombus formation in whole blood on surfaces co-coated with vWF We for the first time investigated the effects of ibrutinib and acalabrutinib on whole blood thrombus formation under flow, simultaneously over multiple surfaces [24]. This allowed systematic analysis of platelet activation and thrombus formation on collagen type I and III and vWF co-coated with laminin, rhodocytin, ristocetin or fibrinogen. In the presence of plasma, concentrations higher than 3.3 μM were required to significantly inhibit collagen-induced aggregation ( Figure 1A). Considering previous in vitro studies, as well as plasma concentrations of ibrutinib and acalabrutinib in patients for inhibition of B-cell carcinomas [25,15], we selected 1 μM to inhibit washed platelets and 5 μM to inhibit platelets in the presence of plasma.
Microscopic visualization of platelet adhesion, activation and thrombus formation on collagen showed that neither of the two inhibitors affected these processes, except for reduced PS-exposure by ibrutinib (Figure 2A). This was confirmed by image analysis resulting in five thrombus parameters (P1-P5) and three activation markers (P6-8, Figure S2). To systematically summarize the effects on all parameters, cumulative histograms were generated showing that ibrutinib, but not acalabrutinib, reduced PS-exposure on collagen type I and III ( Figure 2B, P8, decrease indicated in red). Upon blood perfusion over surfaces that trigger GPIb alone (vWF+ristocetin) or in combination with CLEC-2 (vWF+rhodocytin), α 6 β 1 (vWF+laminin) or α IIb β 3 (vWF+fibrinogen), only ibrutinib decreased almost all parameters of thrombus formation ( Figures 2B and S2). In comparison to control, thrombi were smaller and less compact in structure with ibrutinib and showed reduced expression of activated integrin α IIb β 3 and P-selectin (α-granule release) ( Figure S2). These results show that, in contrast to acalabrutinib, platelet adhesion to surfaces co-coated with vWF was strongly impaired by ibrutinib.

Effects of ibrutinib and acalabrutinib on wild-type and Btk-deficient mouse platelets
To further examine the effects of ibrutinib and acalabrutinib on platelets, thrombus formation and platelet activation were determined in blood from wild-type (WT) and Btk-KO mice. Blood samples were incubated with ibrutinib or acalabrutinib before perfusion over collagen type I. In control conditions (vehicle), thrombus formation was reduced in blood from Btk-KO mice with regard to platelet deposition, multilayer formation (P1-2, P5) and PS-exposure

GPVI-induced platelet activation by CRP-XL was strongly reduced
in Btk-KO platelets as compared to WT with regard to α IIb β 3 activation, α-granule release and PS-exposure, as determined by flow cytometry ( Figures 3D and 3E). This was in agreement with the results obtained from whole blood perfusion experiments. Interestingly, α IIb β 3 activation and P-selectin expression were also moderately reduced in Btk-KO mice upon PAR4 stimulation, while activation with ADP was not affected ( Figures S3B and S3C). When blood was preincubated with Btk inhibitors, we observed significant reduction of all platelet activation markers in WT platelets upon GPVI stimulation ( Figures 3D   and 3E). In platelets from Btk-KO mice, both inhibitors did not further reduce α IIb β 3 activation and P-selectin expression regardless of the agonist. We only observed significant reduction of PS-exposure in Btk-KO platelets in the presence of ibrutinib, as compared to vehicle ( Figure 3E).
Overall, preincubation of whole blood from Btk-KO mice with acalabrutinib and ibrutinib did not result in major additional effects on thrombus formation on collagen under flow, as well as on platelet activation.

Btk inhibitors reduce GPVI-induced phosphorylation of multiple platelet tyrosine kinases
To evaluate off-target effects of the Btk inhibitors, we performed a PamGene kinase assay to visualize which tyrosine kinases were regulated upon platelet stimulation via GPVI, as well as those tyrosine kinases that were subsequently influenced upon pre-incubation of the platelets with 1 μM ibrutinib or acalabrutinib. We observed that in isolated platelets, in total 73 tyrosine kinases were significantly phosphorylated by stimulation with CRP, 65 of which were significantly inhibited by ibrutinib and acalabrutinib preincubation ( Figure 4). Median kinase statistics showed that ibrutinib inhibited tyrosine phosphorylation of these proteins on average 4.4-fold, while acalabrutinib was less strong and reduced this response 1.8-fold, that is, 2.5-fold weaker as compared to ibrutinib.
In comparison to these irreversible inhibitors, we investigated if the reversible Btk inhibitor MK-1026 [14] showed a more favourable antiplatelet effect. First, we evaluated which platelet tyrosine kinases were affected by this compound upon GPVI stimulation. Only 18 tyrosine kinases were significantly inhibited by MK-1026 ( Figure 4). Median kinase statistics showed that MK-1026 inhibited tyrosine phosphorylation of these proteins on average 0.52-fold, which was 8.3-fold lower as compared to ibrutinib. As expected, all three inhibitors significantly inhibited the phosphorylation of Btk, and ibrutinib and acalabrutinib also inhibited Tec. Furthermore, ibrutinib and acalabrutinib also inhibited the phosphorylation of Src family kinases and other downstream proteins as previously described [11,12].
These results show that ibrutinib and acalabrutinib showed the same off-target effects on GPVI-induced tyrosine kinase phosphorylation in platelets, although the effects of acalabrutinib were less strong.
The reversible inhibitor MK-1026 appeared to have less off-target effects in platelets upon GPVI stimulation, in combination with a less strong inhibition profile as compared to the irreversible inhibitors.

The reversible Btk inhibitor MK-1026 shows only limited effects on human and mouse platelets in vitro
As we demonstrated that irreversible inhibition of Btk can impair platelet function, we also investigated the effects of the reversible Btk inhibitor MK-1026 on platelet function. First, we performed collagen-and rhodocytin-induced dose-response aggregation experiments in PRP and washed platelets. Similar to the irreversible Btk F I G U R E 2 Ibrutinib, but not acalabrutinib, affected thrombus formation in whole blood perfused over six different surfaces. Pre-incubated blood samples from healthy donors with vehicle (<0.1% DMSO), acalabrutinib (5 μM) or ibrutinib (5 μM) were recalcified in the presence of PPACK. Blood was perfused for 3.5 min at a wall shear rate of 1000 s -1 over six different surfaces. After perfusion, thrombi were stained with AF568-annexin A5, FITC-α-fibrinogen and AF647-α-CD62P to detect PS-exposure, integrin α IIb β 3 activation and P-selectin expression. (A) Representative images of thrombi formed on collagen type I and vWF co-coated with rhodocytin. (B) Histograms show the cumulative representation of scaled values from 0 to 10 for each parameter over different surfaces consisting of (i) collagen type I, (ii) collagen type III, (iii) vWF co-coated with laminin, (iv) vWF co-coated with rhodocytin, (v) vWF co-coated with ristocetin and (vi) vWF co-coated with fibrinogen in presence of acalabrutinib or ibrutinib. Colours reflect the adhesion parameters (P1-2; black, grey), aggregation parameters (P3-5; shades of green) and activation parameters (P6-8; yellow, orange and red). *p < 0.05, **p < 0.01, ***p < 0.001 versus vehicle F I G U R E 3 Acalabrutinib and ibrutinib reduced thrombus formation and PS-exposure in whole blood under flow, as well as activation of isolated platelets from WT but not Btk-KO mice. Pooled citrated whole blood samples from two wildtype (WT) or Btk-KO mice were pre-incubated with vehicle (<0.1% DMSO), acalabrutinib (5 μM) or ibrutinib (5 μM) in the presence of PPACK and fragmin for 10 min at 37 • C. After incubation, blood samples were recalcified and perfused for 3.5 min at a wall shear rate of 1000 s -1 over a collagen type I surface. Thrombi were post-stained with FITC-labelled anti-CD62P mAb, PE-labelled JonA mAb and AF647-labelled Annexin-A5 to detect P-selectin expression, integrin activation and PS-exposure, respectively.  (Table S1). Based on these results, in combination with reported in vivo plasma concentrations [26], we selected 5 μM for further experiments. Aggregation induced by ADP and U46619 were slightly but significantly reduced at this concentration of MK-1026, while responses to thrombin and ristocetin were unaffected ( Figures   S4A and S4B). Integrin α IIb β 3 activation and P-selectin expression were not inhibited by MK-1026 in platelets stimulated with ADP or thrombin ( Figure S4C). Also, MK-1026 did not alter platelet adhesion and thrombus formation on six different surfaces, as seen with acalabrutinib ( Figures 5D and S5).
We also investigated the effects of MK-1026 using blood from Btk-KO mice. Similarly as for acalabrutinib and ibrutinib, we did not observe additional inhibition of whole blood thrombus formation on collagen with MK-1026 in blood samples from WT or Btk-KO mice ( Figures 6A,   6B and S6A). Also, no inhibitory effect of MK-1026 was found upon platelet activation or PS-exposure from WT or Btk-KO mice with CRP-XL, ADP or PAR4 agonist ( Figures 6C, 6D, S6B and S6C).
These data show that the reversible Btk inhibitor MK-1026 inhibited GPVI-and CLEC-2-mediated aggregation of washed platelets but did not impair platelet adhesion or activation in whole blood perfused over collagen.

Ibrutinib inhibits thrombus formation under flow on multiple surfaces and GPVI-induced platelet activation responses in patients
Next, we assessed which platelet activation pathways were altered in patients using ibrutinib. Sixteen patients were included, of which seven did not receive ibrutinib, and nine were treated with 420-560 mg ibru-tinib once a day (Table 1). Baseline characteristics were largely similar between groups. Importantly, none of the patients were thrombocytopenic, except one patient with a platelet count of 50 × 10 9 /L (Table 1). Of the ibrutinib-treated patients, 66% (6/9) reported bleeding symptoms with a median ISTH-BAT score of 2 (range 1-7, Table 1).
Of patients without ibrutinib treatment, none reported bleeding symptoms.
Whole blood from patients was perfused over collagen type I or vWF plus rhodocytin or laminin. No differences in thrombus formation were observed between patients without ibrutinib and healthy controls ( Figure S7). On collagen, treatment with ibrutinib reduced thrombus contraction and height (multilayer) compared to non-treated patients. Furthermore, α IIb β 3 activation , P-selectin expression and PS-exposure were significantly reduced ( Figures 7A   and S7A). For thrombi formed on vWF plus rhodocytin or laminin, α IIb β 3 activation and α-granule secretion were reduced upon ibrutinib treatment ( Figures 7A, S7B and S7C).

Aggregation of isolated platelets from ibrutinib-treated patients
showed on average no significant inhibition upon stimulation with any agonist ( Figure 7B). However, platelets from five ibrutinib-treated patients were highly nonresponsive to collagen stimulation, while platelets from two patients responded normally to collagen. Also, with rhodocytin, platelets from four patients receiving ibrutinib did not aggregate, while platelets from four patients showed normal aggregation. When platelet activation was examined by flow cytometry, α IIb β 3 activation and P-selectin expression were inhibited in CRP-XLstimulated platelets from all ibrutinib-treated patients ( Figure 7C).
Upon ADP stimulation, platelets from patients on treatment showed a slight increase in α IIb β 3 activation, while no effects on secretion were observed ( Figure 7C). PS-exposure by thrombin plus CRP-XL, was unchanged for all patients' platelets ( Figure 7D).
As 66% of our patients reported bleeding symptoms, we compared patients with significant bleeding (>2 ISTH-BAT score) with patients with no/mild bleeding (0-1 ISTH-BAT score). Generally, no significant changes were observed in platelet aggregation and thrombus formation in patients with or without bleeding ( Figure S8).

DISCUSSION
Due to their promising effects on progression-free survival, Btk inhibitors are increasingly prescribed in haematological malignancies [4,5]. Btk inhibitors require lifelong treatment, and the high percentage of discontinuation due to side effects (including bleeding tendency) stresses the importance of understanding and managing these [6][7][8]. We found that ibrutinib and acalabrutinib inhibit both Other studies investigating effects of Btk inhibitors on platelet function have described different findings. Several papers have shown inhibition of the GPVI pathway by ibrutinib or acalabrutinib, either directly via Btk or due to off-target effects on Src kinases or other downstream proteins [11,12,23,[27][28][29][30][31]. Others have implicated that GPIb- [29,32] or α IIb β 3 -mediated signalling [28,31] were affected by ibrutinib. In agreement, in the present study ibrutinib also inhibited thrombus formation on ristocetin and fibrinogen co-coated with vWF. Levade et al have also shown that ibrutinib affected platelet adhesion to vWF only under flow but did not assess acalabrutinib [22].
Furthermore, we found that both ibrutinib and acalabrutinib impaired GPVI-mediated aggregation but not platelet adhesion on collagen under flow. Previous studies found conflicting results with regard to the effects of ibrutinib on collagen adhesion, with some who did find inhibition [12,33] and others that could not [34]. A possible explanation for this is that platelet adhesion to collagen is also regulated by integrin α 2 β 1 , which does not signal via Btk [34]. Our data now directly compare the differences between ibrutinib and acalabrutinib in whole blood thrombus formation under flow on six different surfaces and show for the first time that ibrutinib, but not acalabrutinib, impairs platelet adhesion surfaces co-coated with vWF. This might play a role in the increased incidence of major bleeding seen in ibrutinib treatment compared to acalabrutinib.
Although GPVI is important for platelet adhesion and activation on collagen, recent studies have shown that GPVI inhibitors do not impair haemostasis [16]. GPVI-deficient patients exhibit only a mild bleeding diathesis [17], and GPVI depletion in mice did not increase bleeding tendency [19,20]. Therefore, the bleeding tendency seen with Btk inhibitors cannot solely be mediated by GPVI. Depletion of both GPVI and CLEC-2 from mouse platelets did impair haemostasis [21]. In line with this, we found that Btk inhibitors inhibited both GPVI-and CLEC-2-mediated aggregation.
In line with this, we found decreased GPVI-induced platelet granule secretion and thrombus formation in Btk-KO mice. XLA patients demonstrated impaired GPVI-or CLEC-2-induced PLCγ2 phosphorylation and platelet aggregation [39,40]. As compared to XLA patients, ibrutinib has similar effects on GPVI, CLEC-2 and GPIb signalling.
However, XLA patients are not associated with an increased bleeding risk [12]. This suggests that Btk inhibitors affect other targets, most notably Tec, which can substitute for Btk [35] and is also inhibited by ibrutinib and acalabrutinib [12]. Also, off-target effects on Src kinases have been implicated in the bleeding tendency [11,12].
We assessed off-target effects on GPVI-induced tyrosine kinase phosphorylation in platelets by using the PamGene assay. These results showed that ibrutinib and acalabrutinib showed the same off-target effects, although the effects of acalabrutinib were less strong. Importantly both inhibitors had an off-target effect on Tec and Src kinases, questioning the previous conclusions that these were responsible for the increased severe bleeding seen with ibrutinib compared to acalabrutinib.
When comparing results found in vitro in human and mice, we found that in human samples, Btk inhibitors did not alter collagenmediated adhesion under flow in agreement with Zheng et al [31], whereas in mice, Btk inhibition (either by KO mice or acalabrutinib) reduced this response. Btk-dependent signalling is completely absent in KO mice, and therefore, this may result in stronger effects as compared to pharmacological inhibition. Btk proteins display 99.4% similarity between human and mice [41], so a species-dependent interaction with an inhibitor cannot explain the observed differences. A possibility could be different bioavailability between mouse and man.
This may contribute to the more pronounced effects of acalabrutinib in mouse blood as compared to human blood. We also observed differ- In human and mouse blood, MK-1026 did not impair collagen-induced thrombus formation under flow, in contrast to acalabrutinib which did affect this response in mouse blood. Hence, this reversible inhibitor had less inhibitory effects on platelets as compared to ibrutinib, in agreement with a recent study [31]. Hence, MK-1026 could be expected to show a slightly reduced or comparable bleeding tendency compared to acalabrutinib.
The present dose-response experiments, in line with previous studies [11,33], show that at similar dose, ibrutinib is a more effective platelet inhibitor compared to other Btk inhibitors. This has been attributed to the inhibition of drug efflux pumps by ibrutinib [33].
We observed that higher concentrations of inhibitors are required to inhibit platelet aggregation in the presence of plasma, as compared to washed platelets. This in line with previous observations with other TKIs [44,45]. This is most likely caused by binding of ibrutinib and acalabrutinib to albumin [46][47][48]. It has been reported that prolongation of the incubation time lowered the IC 50 values of ibrutinib and acalabrutinib for GPVI-mediated aggregation [49]. However, in that study a much lower collagen concentration was used, which may be more sensitive to longer incubation with low doses of inhibitors. Furthermore, that study did not include CLEC-2-dependent platelet responses [49].
Although inhibitor concentration could play a role in the extent of inhibition, we showed that ibrutinib and acalabrutinib influence additional platelet pathways as compared to MK-1026, suggesting that the extent of Btk inhibition might not be the reason for the increased bleeding tendency.
Previous studies assessing which platelet pathways are involved in bleeding tendency in patients using ibrutinib found conflicting results. Some showed a correlation with collagen-induced aggregation [12,22,50] and some with platelet adhesion to collagen [12], whereas others could not find this [22]. Bleeding tendency was also associated with adhesion to vWF under flow [22], and one study found a correlation with ristocetin-induced platelet aggregation [32], which another study could not replicate [28]. Ibrutinib can also induce shedding of GPIbα, GPIX and integrin α IIb β 3 in patients, but the correlation with bleeding remains unknown [51].
A recent study assessed platelet parameters in patients with CLL and mantle cell lymphoma (MCL) [52], showing correlations between bleeding tendency, thrombocytopenia and decreased ADP-induced platelet aggregation. However, a drawback of this study was that a large patient proportion was thrombocytopenic, which can directly influence platelet aggregation, as platelet concentration in PRP was not reported to be adjusted. This impaired the establishment if ibrutinib or platelet count affected aggregation response. In our patients with (in general) normal platelet counts, we also observed significant differences in platelet responses to several stimuli of patients using ibrutinib. Furthermore, 66% of the included patients with ibrutinib treatment reported bleeding symptoms. Although we could show that in patients, similar to healthy volunteers, ibrutinib inhibited GPVI signalling, with variable effects on CLEC-2, as well as reduced thrombus formation to surfaces co-coated with vWF, this could not differentiate for the bleeding tendency. Generally, the patients showed a large variation in measurement outcomes, which might be influenced by clinical factors. However, with patients on ibrutinib treatment, we have directly compared whole blood thrombus formation under flow on multiple surfaces, including vWF plus rhodocytin and laminin, which has not been reported with ibrutinib-treated patients thus far.
In conclusion, the present work demonstrated that ibrutinib, acalabrutinib and MK-1026 inhibited GPVI-and CLEC-2-mediated platelet aggregation, but only ibrutinib also inhibited GPVI-induced platelet activation and thrombus formation on surfaces co-coated with vWF.
The novel reversible BTK inhibitor MK-1026 might therefore be promising for future treatment in patients at risk for bleeding.