See Supplementary Information for details.

S. aureus Newman (a gift from Prof Steve Kerrigan, RCSI, Ireland) and E. coli RS218 (a gift from Prof Ian Henderson, University of Queensland, Australia) were cultured and prepared as described (29, 31) ( Supplementary Information ).

This study was performed in accordance with relevant ethics committees: Hull York Medical School (reference number 1501) and UK National Health Service Research Ethics (08/H1304/35). Informed consent was obtained from all participants.

Peripheral blood from CLL and X-linked agammaglobulinemia (XLA) patients was taken at the Departments of Haematology and Immunology & Allergy, respectively, at Castle Hill Hospital (Cottingham, UK). Blood was drawn using sodium citrate or acid-citrate-dextrose (ACD) vacutainers (see below), and shipped to the University of Hull within 4 hours of venepuncture for immediate testing. Ibrutinib-treated CLL patients were taking a daily dose of 420 mg, except for two patients who were taking 140 mg. Blood from healthy donors was collected at the University of Hull in syringes containing sodium citrate or ACD from volunteers over 18 years of age not treated with concurrent anti-platelet agents.

Platelet-rich plasma (PRP) and platelet-poor plasma (PPP) were obtained from sodium citrated blood, while washed platelets (WP) were prepared from blood in ACD, as detailed in Supplementary Information .

Platelet aggregation in PRP or WP was done under stirring conditions (1000 rpm) at 37°C using light transmission aggregometry with a CHRONO-LOG 490 aggregometer (CHRONO-LOG, Havertown, PA, USA). WP reactions were supplemented with 1 mg/ml human fibrinogen and 0.2 mg/ml human IgGs before experimentation. Platelets were stimulated with bacteria and other agonists, including crosslinked collagen-related peptide (CRP-xl), thrombin receptor activator peptide 6 amide (TRAP-6) and ADP. To crosslink FcγRIIA, platelets were incubated with anti-FcγRIIA monoclonal antibody (mAb) (4 μg/ml, clone IV.3) for 2 minutes, followed by the addition of crosslinking F(ab’)₂ rabbit anti-mouse IgG (30 μg/ml), altogether designated as IV.3-xl hereafter.

Supernatants from light transmission aggregometry reactions were collected and analyzed for human platelet factor 4 (PF4) by enzyme-linked immunosorbent assay (ELISA) (R&D Systems), as described in Supplementary Information (29).

PRP samples diluted to 2x10⁷ platelets/ml including 15% autologous plasma were incubated for 1 hour with S. aureus Newman or fibrinogen immobilized on glass coverslips. Cells were stained with Hoechst 33342 (to visualize bacteria) and TRITC phalloidin (to visualize platelets) and imaged using a Zeiss Axio Observer fluorescence microscope equipped with ApoTome structured illumination and an AxioCam 506 camera (Zeiss, UK). See Supplementary Information for details.

Flow cytometry was performed in PRP samples using a BD LSR Fortessa cell analyzer (BD Biosciences) as described in Supplementary Information . Data were analyzed using the FlowJO software (BD Biosciences).

WP (10x10⁸/ml) supplemented with 1 mg/ml fibrinogen plus 0.2 mg/ml hIgGs were stimulated as required in aggregation conditions. Samples were lysed 3 minutes after the start of aggregation, or at a parallel time point in the case of samples where inhibitors blocked aggregation, by adding an equal volume of 2x lysis buffer (29) and analyzed using a RayBio human phosphotyrosine Tec ELISA kit (RayBiotech. Georgia, USA) following the manufacturer’s instructions.

Protein detection in whole platelet lysates was done by SDS-PAGE and near-Infrared Western Blot Detection (LI-COR Biosciences) using lysates collected from PRP samples in the absence of αIIbβ3 inhibitors, as previously described (29) with minor modifications (see Supplementary Information for details).

GraphPad Prism was used to perform the statistical analysis, including normality Kolmogorov-Smirnov and Shapiro-Wilk tests. Data are presented as the mean ± SD. In case of comparisons between two groups, means were evaluated using Student t-test or Mann-Whitney U test, for parametric and non-parametric data, respectively. If more than two groups were considered, one-factor or two-factor analysis of variance (ANOVA) or Kruskal-Wallis test (for parametric and non-parametric data, respectively) followed by multiple comparison tests were performed. For categorical variables, p values were calculated using the chi-square test or Fisher’s exact test. P ≤ 0.05 was considered significant.

A total of 34 untreated and 32 ibrutinib-treated CLL patients were enrolled with their characteristics summarized in Table 1 . The median age was 72 years, and most patients received concurrent medications for co-morbidities ( Supplementary Table 1 ). The median number of drugs excluding ibrutinib was four for ibrutinib-treated, and three for untreated patients ( Supplementary Table 2 ). Antiplatelet and anticoagulants were present in 6% of ibrutinib-treated (e.g., aspirin) and 18% of untreated (e.g., aspirin, clopidogrel, warfarin) CLL patients ( Supplementary Table 1 ). Notably, 59% of ibrutinib-treated patients were taking prophylactic antibiotics, and 13% were on combined chemotherapy (venetoclax).

Bacteria interact with plasma proteins (including IgGs) and platelet receptors in a bacterial strain-specific manner (34). S. aureus Newman and E. coli RS218 are known to cause IgG-dependent FcγRIIA-mediated aggregation and granule secretion in platelets derived from healthy donors (29, 31). Therefore, we asked whether platelets from CLL patients would be similarly responsive to bacteria and, if so, whether ibrutinib treatment would interfere with platelet immune functions. We tested platelet activation induced by clustering of FcγRIIA via IV.3-xl, bacteria, and GPVI-specific agonist, CRP-xl, in the presence of plasma. We also stimulated platelets with TRAP-6 and ADP, which are agonists for G-protein coupled receptors that are not affected by ibrutinib (12–14).

Platelets derived from untreated CLL patients responded to FcγRIIA crosslinking (IV.3-xl) and CRP-xl at levels comparable to controls ( Figure 1A ). S. aureus Newman and E. coli RS218 induced full or nearly full aggregation in 8 and 9 out of 13 samples respectively, similarly to TRAP-6 and ADP, with aggregation to S. aureus Newman and TRAP-6 being significantly lower than in controls ( Figure 1A ). The antiplatelet agent, clopidogrel, was taken by two CLL patients, but their platelets aggregated fully to bacteria; however, two aspirin-treated CLL patients showed a reduction in aggregation to bacteria, especially S. aureus Newman ( Supplementary Figure 1 ).

Platelet aggregation to S. aureus Newman and E. coli RS218 is characterized by a lag time between bacterial stimulation and start of aggregation that contrasts with the immediate onset of aggregation caused by IV.3-xl (29, 31). Ibrutinib-untreated CLL platelets in which aggregation in response to bacteria was detected ( Figure 1A ) showed a statistically significant decrease in lag times compared to controls ( Figure 1C ). Overall, our data indicates that platelets from ibrutinib-free CLL patients aggregate to bacteria in the presence of autologous plasma, although a trend to decreased aggregation compared to FcγRIIA crosslinking is seen.

In contrast, platelets from ibrutinib-treated CLL patients showed strong inhibition of IV.3-xl-induced aggregation, and a significant decrease in aggregation to bacteria ( Figure 1A ). As opposed to IV.3-xl, heterogeneity in platelet aggregation to CRP-xl was observed in our ibrutinib-treated platelet group ( Figure 1A ). In previous studies, defects in collagen-induced platelet aggregation were detected in only a subset of ibrutinib-treated patients (12, 13); and the presence of ibrutinib low and high sensitivity groups of both healthy individuals and CLL patients regarding platelet responses to collagen has been reported (17). As published (12, 13), responses to TRAP-6 and ADP (which signal through G-protein coupled receptors) were mostly unaffected in our ibrutinib-treated group ( Figure 1A ). Differences in aggregation among the three clinical groups were not due to varying platelet concentrations in PRP samples ( Supplementary Figure 2 ).

We next measured release of PF4, a chemokine that can interact with bacteria directly (29, 37), as a marker of α-granule secretion. PF4 was secreted by ibrutinib-untreated CLL platelets upon IV.3-xl and bacteria stimulation ( Figure 1B ) to similar levels as healthy controls ( Figures 1B and 4C ) (29). In contrast, ibrutinib-treated CLL patients showed an inhibitory trend of PF4 secretion in response to bacteria, IV.3-xl and CRP-xl, with the latter two being significantly reduced compared to untreated CLL platelets ( Figure 1B ). PF4 release in response to TRAP-6 and ADP was not affected by ibrutinib treatment ( Figure 1B ).

To investigate if the CLL platelet responses observed above were mediated by FcγRIIA, aggregation and secretion were measured after incubating PRP with IV.3 mAb, which without the crosslinking F(ab’)₂ inhibits FcγRIIA activation. Consistent with previous observations in healthy donor platelets (29, 31), CLL platelet aggregation and PF4 secretion to S. aureus Newman and E. coli RS218 were inhibited by IV.3 mAb and by the Src inhibitor, dasatinib ( Figures 1D, E ).

Thus, our results support that platelet immune recognition of bacteria via FcγRIIA is impaired in CLL patients treated with ibrutinib.

As innate immune effectors, platelets scan the vasculature and collect bacteria (21). Therefore, we next investigated if CLL platelets retain the capacity to scavenge bacteria and whether this is impaired by ibrutinib. Platelets from healthy controls and CLL patients were found to scavenge S. aureus Newman in the presence of autologous plasma with an average (± SD) of 15 ± 8.5 and 13 ± 11.5 bacteria per platelet (bacteria clusters), respectively ( Figures 2A, B ). Pre-treatment of PRP samples with the integrin αIIbβ3 inhibitor, eptifibatide, abolished scavenging of bacteria by both control and CLL platelets ( Figures 2A, B ), pointing out to a key role for the integrin in this response. Scavenging was also dependent upon FcγRIIA ( Supplementary Figure 3 ). Remarkably, platelets from CLL patients taking ibrutinib could bind to S. aureus Newman, but showed no morphological changes indicative of activation as compared to controls and were not able to scavenge bacteria ( Figures 2A, B ).

The same platelet samples were tested for spreading on immobilized fibrinogen. During spreading, the actin cytoskeleton rearranges causing cellular protrusions including finger-like filopodia and laterally extended lamellipodia; the latter being characteristic of later stages of spreading (38). A statistically significant decrease in total number of CLL platelets binding to fibrinogen was detected compared to healthy controls ( Figures 2A, C ), however, both groups showed similar percentages of platelets with lamellipodia. In contrast, platelets from CLL patients taking ibrutinib bound to fibrinogen to similar levels as controls, but did not spread ( Figures 2A, C ), consistent with published studies performed with washed platelets (14, 39). These data show decreased efficiency in platelet bacteria-scavenging ability and spreading in CLL patients treated with ibrutinib.

The reduction in platelet responses to bacteria observed in ibrutinib-treated CLL patients could be due to heterogeneity in surface expression of important receptors like FcγRIIA and αIIbβ3 (40). Therefore, we performed flow cytometry analyses of platelets derived from our three clinical groups to measure membrane protein levels of αIIbβ3, FcγRIIA and GPVI in PRP samples ( Supplementary Figure 4 ). This analysis revealed no change in cell size, but a significant decrease in granularity/complexity in untreated CLL platelets ( Figure 3A ). We also observed significant reductions of αIIbβ3 (as detected by two different mAb clones against CD41, the αIIb subunit) and GPVI, but not FcγRIIA (CD32), in CLL platelets independently of ibrutinib treatment when compared to healthy control platelets ( Figure 3B ). This suggests that decreased levels of αIIbβ3 in blood circulating platelets are unlikely to be the major or only cause of inhibition of ibrutinib-treated CLL platelet responses to bacteria.

To investigate the direct effect of ibrutinib on platelet response to bacteria, we performed light transmission aggregometry with healthy control PRP stimulated in vitro with different concentrations of ibrutinib ( Figure 4A and Supplementary Figure 5A ). Pre-incubation with 5µM ibrutinib for 5 minutes significantly inhibited platelet aggregation to S. aureus Newman and E. coli RS218, but did not cause a significant reduction of aggregation in response to CRP-xl or TRAP-6. A lower dose of ibrutinib (2 µM) was able to abolish aggregation to FcγRIIA crosslinking (IV.3-xl) ( Figure 4A ). Longer pre-incubation times with ibrutinib (20 and 40 minutes) did not significantly reduce the concentration of drug needed to inhibit aggregation to bacteria (data not shown). As internal controls, IV.3 mAb alone (FcγRIIA inhibitor) and dasatinib (Src inhibitor) were used in PRP, and both inhibited aggregation to S. aureus Newman and E. coli RS218, but not TRAP-6 ( Supplementary Figure 5C ), as previously reported (29, 31). The integrin αIIbβ3 inhibitor eptifibatide, as expected, blocked aggregation in response to all agonists including TRAP-6 ( Supplementary Figure 5C ) (29, 31).

The high concentrations of ibrutinib needed to inhibit aggregation in PRP contrasted with previous studies done in washed platelets (12, 14, 16), thus we wondered if binding of ibrutinib to plasma proteins (41) may explain the difference. As seen in Supplementary Figure 6A , washed platelet aggregation in response to S. aureus Newman was inhibited by 0.5 µM ibrutinib, while 0.1 µM ibrutinib was sufficient to inhibit the response to E. coli RS218 and FcγRIIA crosslinking (IV.3-xl). A higher dose of ibrutinib (1 µM) was necessary to inhibit washed platelet aggregation stimulated with CRP-xl. None of the ibrutinib concentrations tested in PRP and washed platelets caused inhibition of TRAP-6 induced aggregation ( Figure 4A and Supplementary Figures 5A, 6A ).

The second generation iBtk, acalabrutinib, was also assessed for its ability to inhibit platelet responses to bacteria. Like ibrutinib, acalabrutinib binds irreversibly to Btk at C481 and is highly bound to plasma proteins (42, 43). In PRP, 10-15 µM acalabrutinib impaired aggregation after IV.3-xl and bacteria exposure, but not in response to CRP-xl and TRAP-6 ( Figure 4B and Supplementary Figure 5B ). In washed platelets, the inhibitory effects of acalabrutinib were more prominent. Aggregation to IV.3-xl was inhibited by 5-10 µM acalabrutinib, while for bacteria, 1µM and 10 µM inhibited aggregation upon exposure to E. coli RS218 and S. aureus Newman respectively ( Supplementary Figure 6B ). Washed platelet aggregation after stimulation with TRAP-6 and CRP-xl was unaffected or only partially inhibited (10 µM) by acalabrutinib respectively ( Supplementary Figure 6B ), as previously described (17, 44).

Ibrutinib and acalabrutinib also inhibited platelet α-granule secretion (PF4) in response to IV.3-xl and bacteria, but not to TRAP-6 ( Figure 4C ). As previously reported (29, 31), eptifibatide blocked PF4 release in response to bacteria, but not to IV.3-xl ( Figure 4C ). Moreover, PF4 release in response to bacteria (but not TRAP-6) was inhibited by blocking FcγRIIA with IV.3 mAb alone ( Figure 4C ). This confirms previous observations of the interplay between FcγRIIA and αIIbβ3 in mediating platelet activation, including secretion, to bacteria.

Hence, in keeping with the above results from our clinical cohort of ibrutinib-treated CLL patients, iBtks strongly and directly inhibit immune function even in platelets derived from healthy donors.

We showed that ibrutinib can inhibit platelet immune recognition of bacteria, mediated by FcγRIIA, whilst not impacting on the surface levels of this receptor, thus we sought to investigate downstream intracellular signaling events. The binding of ibrutinib and acalabrutinib to Btk C481 blocks Btk autophosphorylation in Y223 that is critical for its activation (5). FcγRIIA crosslinking with non-physiological agonists induces Btk Y223 phosphorylation (35), yet it remains unknown if this is also critical downstream of bacteria recognition. In healthy donors, phosphorylation of Btk Y223 was detected in PRP after IV.3-xl and bacteria exposure, although at lower levels than those induced by CRP-xl ( Figure 5A ). For S. aureus Newman and E. coli RS218, inhibition of FcγRIIA (with IV.3 mAb alone) and αIIbβ3 (with eptifibatide) reduced Btk phospho-Y223 signal to basal levels ( Figures 5B.1, B.4 and B.5 ), which indicates that both receptors control Btk activation in response to bacteria, as opposed to CRP-xl ( Figures 5B.1, B.2 ) and TRAP-6 ( Figure 5B.6 ). For all agonists tested, Btk phospho-Y223 was inhibited below basal levels by ibrutinib and acalabrutinib as expected ( Figures 5A, B ).

We wondered whether signaling through FcγRIIA, rather than GPVI, would be more reliant on Btk activity, and hence sensitive to ibrutinib inhibition. Therefore, we analyzed if there was a correlation between the ibrutinib concentration found to inhibit Btk Y223 phosphorylation and that needed to inhibit aggregation. Upon IV.3-xl, concentrations of ibrutinib that reduced Btk phospho-Y223 to basal levels (1 μM) or below (2 μM) were also able to partially decrease aggregation ( Figure 5C ), with total inhibition of aggregation requiring higher concentrations of ibrutinib (5 μM). In contrast, inhibition of Btk Y223 phosphorylation did not correlate with suppression of CRP-xl induced aggregation ( Figure 5C ), as previously reported (16). Thus, phosphorylation of Btk Y223 downstream of bacteria recognition uses both FcγRIIA and integrin pathways, with FcγRIIA potentially being more sensitive to ibrutinib treatment than GPVI, in healthy donor-derived platelets.

We next sought to investigate CLL samples and found that CLL platelets activated Btk in response to IV.3-xl, bacteria, CRP-xl and TRAP-6 to a similar degree as healthy controls (compare Figures 5A, D ). This contrasted with ibrutinib-treated CLL platelets in which no phosphorylation of Btk Y223 was detected in response to IV.3-xl, CRP-xl and TRAP-6 ( Figure 5D ).

Platelets express two members of the Tec family of tyrosine kinases, Btk and Tec. Although Btk has an approximate 8 to 10-fold greater level of expression in human platelets (45), studies suggest that ibrutinib-mediated inhibition of Tec has important consequences in platelet function (14, 16). Thus, we aimed to decipher if Tec activation takes place downstream of FcγRIIA-mediated platelet response to bacteria as seen for IV.3-xl (35). Due to the lack of phospho-specific antibodies to detect Tec autophosphorylation site Y206 (46), total levels of Tec tyrosine phosphorylation were measured by ELISA.

In healthy donor-derived platelets, Tec phosphorylation increased upon exposure to S. aureus Newman and E. coli RS218, and was inhibited by pre-treatment with ibrutinib ( Figure 6 ). Moreover, when CLL platelets were stimulated with S. aureus Newman, Tec phosphorylation reached similar levels as those observed in healthy controls ( Figure 6 ). We used CRP-xl to compare Tec phosphorylation across our three clinical groups. As seen in Figure 6 , healthy donor and untreated CLL platelets stimulated with CRP-xl showed an increase in phospho-Tec that was significantly decreased in platelets from ibrutinib-treated patients. Overall, these data suggest that ibrutinib, both at concentrations used in vitro and at therapeutic doses, affects platelet Tec activation.

Mutations in the Btk gene resulting in lack of Btk expression or function cause XLA, a condition characterized by markedly reduced or absent B cells and profound hypogammaglobulinemia (47). To investigate if Btk function is essential for FcγRIIA-mediated platelet responses to bacteria, we obtained PRP from two XLA patients receiving periodic intravenous immunoglobulin treatment. Platelets from both patients aggregated to S. aureus Newman and E. coli RS218 although with varying lag times ( Figure 7A ). Upon crosslinking of FcγRIIA (IV.3-xl), no or less aggregation was observed in patients 2 and 1 respectively ( Figure 7A ), suggesting that this pathway is more strongly affected by Btk deficiency than bacteria-induced activation. Importantly, XLA platelet responses to bacteria were still dependent on FcγRIIA, as aggregation was blocked by pre-incubating PRP with inhibitory IV.3 mAb alone ( Figure 7B ). Conversely, we detected normal XLA platelet aggregation in response to CRP-xl and TRAP-6 ( Figure 7A ), which was independent of FcγRIIA ( Figure 7B ). These results suggest that while Btk has an active role in FcγRIIA-mediated platelet function, this is not essential for FcγRIIA-mediated responses to bacteria.