Bevacizumab immune complexes activate ... - Wiley Online Library

14 downloads 3028 Views 913KB Size Report
Oct 21, 2008 - *Center for Thrombosis Research, Florida Hospital, Orlando, FL, USA; and .... tech (South San Francisco, CA, USA), Centocor (Horsham,.
Journal of Thrombosis and Haemostasis, 7: 171–181

DOI: 10.1111/j.1538-7836.2008.03212.x

ORIGINAL ARTICLE

Bevacizumab immune complexes activate platelets and induce thrombosis in FCGR2A transgenic mice T. MEYER,* L. ROBLES-CARRILLO,* T. ROBSON,* F. LANGER,  H. DESAI,* M. DAVILA,* M. AMAYA,* J . L . F R A N C I S * and A . A M I R K H O S R A V I * *Center for Thrombosis Research, Florida Hospital, Orlando, FL, USA; and  University Hospital Eppendorf, Hamburg, Germany

To cite this article: Meyer T, Robles-Carrillo L, Robson T, Langer F, Desai H, Davila M, Amaya M, Francis JL, Amirkhosravi A. Bevacizumab immune complexes activate platelets and induce thrombosis in FCGR2A transgenic mice. J Thromb Haemost 2009; 7: 171–81.

Summary. Background: Treatment with Bevacizumab has been associated with arterial thromboembolism in colorectal cancer patients. However, the mechanism of this remains poorly understood, and preclinical testing in mice failed to predict thrombosis. Objective: We investigated whether thrombosis might be the result of platelet activation mediated via the FccRIIa (IgG) receptor – which is not present on mouse platelets – and aimed to identify the functional roles of heparin and platelet surface localization in Bev-induced FccRIIa activation. Methods and results: We found that Bev immune complexes (IC) activate platelets via FccRIIa, and therefore attempted to reproduce this finding in vivo using FccRIIa (hFcR) transgenic mice. Bev IC were shown to be thrombotic in hFcR mice in the presence of heparin. This activity required the heparin-binding domain of BevÕs target, vascular endothelial growth factor (VEGF). Heparin promoted Bev IC deposition on to platelets in a mechanism similar to that observed with antibodies from patients with heparin-induced thrombocytopenia. When sub-active amounts of ADP or thrombin were used to prime platelets (simulating hypercoagulability in patients), Bev IC-induced dense granule release was significantly potentiated, and much lower (sub-therapeutic) heparin concentrations were sufficient for Bev IC-induced platelet aggregation. Conclusions: The prevailing rationale for thrombosis in Bev therapy is that VEGF blockade leads to vascular inflammation and clotting. However, we conclude that Bev can induce platelet aggregation, degranulation and thrombosis through complex formation with VEGF and activation of the platelet FccRIIa receptor, and that this provides a better explanation for the thrombotic events observed in vivo. Keywords: bevacizumab, heparin-induced thrombocytopenia, thrombosis. Correspondence: Todd Meyer, Center for Thrombosis Research, Florida Hospital, 2501 North Orange Avenue, Suite #786, Orlando, FL, USA. Tel.: +1 407 303 2440; fax: +1 407 303 2441. E-mail: todd.meyer@flhosp.org Received 7 July 2008, accepted 21 October 2008  2008 International Society on Thrombosis and Haemostasis

Introduction The anti-angiogenic drug bevacizumab (Bev; Avastin, Genentech, San Francisco, CA, USA) is associated with arterial thromboembolic events in colorectal cancer patients [1]. Bev is a monoclonal IgG antibody (mAb) that forms immune complexes (IC) with vascular endothelial growth factor (VEGF; see Table 1 for abbreviation list). Bevinduced thrombosis surprised investigators [2], in part because many animal studies with anti-VEGF mAb therapy failed to reveal thrombotic side effects [3]. Prevailing opinion on the molecular mechanism behind the Ôconcurrently high incidence of bleeding and thrombosisÕ in Bev therapy derives from the suggestion by Kilickap et al. [4] that tissue factordriven coagulation, secondary to vascular endothelial cell (EC) dysfunction, may cause thrombosis as a result of VEGF suppression by Bev. This often-cited hypothesis [5], described as Ôarterioprotective actions of VEGF,Õ [2] has not been demonstrated in vivo (experimentally), but finds modest support in reports of thrombotic events in other antiangiogenic therapies (e.g. thalidomide, SU5416 [6]). However, while SU5416 + chemotherapeutics enhanced EC procoagulant activity, other anti-angiogenic drugs, including Bev, did not increase EC procoagulant activity [7]. Indeed, Ôthere have been no preclinical models demonstrating an increased risk of bleeding or thrombosis associated with VEGF inhibition [8]Õ. Arterioprotection by VEGF was not seen in preclinical Bev studies in primates [9] and mice [10] in which prolonged VEGF suppression produced no adverse vascular effects [11]. Recently, results from a long-term toxicity study in mice having human VEGF were reported [12]. Animals receiving biweekly high-dose anti-VEGF mAb exhibited no histologic differences from controls in any major organ, except limited glomerular IC deposition with some anti-VEGF mAbs (but not Bev) [12]. Clearly, evidence from published studies precludes Bev-suppression of VEGF-dependent arterioprotection as an explanation for the high incidence of Bev-associated bleeding and thrombosis. Rather, other molecular processes – active in human patients but not laboratory mice – appear to be required.

172 T. Meyer et al Table 1 List of frequently used abbreviations Terms

Definitions

mAb Heparin or UFH HIT PF4 IC Bev VEGF VEGF165 VEGF121 HBD

monoclonal Antibody Unfractionated heparin Heparin-induced thrombocytopenia Platelet factor 4 Immune complexes Bevacizumab (Avastin), an anti-VEGF mAb Vascular endothelial growth factor 165 Vascular endothelial growth factor 165 Vascular endothelial growth factor 121 Heparin-binding domain (e.g. of VEGF165 or PF4) B6;SJL-Tg(FCGR2A)11MKZ/J mice (having FccRllA) The platelet IgG receptor Endothelial cells Adenosine 5Õ-diphosphate Serotonin release assay Platelet-rich plasma Mean fluorescence intensity

hFcR mice FccRlla EC ADP SRA PRP MFI

CD40 ligand (CD154, an inflammatory cytokine) is an important therapeutic mAb target in autoimmune disease. Previously, we reported that anti-CD154 immune complexes directly activate platelets [13]. We later observed that a therapeutic anti-CD154 mAb, withdrawn from clinical trials because of fatal thromboembolism, also directly activates platelets when bound to its platelet surface antigen (CD154). Key to this discovery was the use of sub-active concentrations of ADP as a platelet priming agent, which improved assay sensitivity and donor responsiveness. Importantly, subsequent pharmaceutical anti-CD154 mAb development has attributed thromboembolism in antiCD154 clinical trials to mAb-induced platelet aggregation [14]. The apparent mechanism behind anti-CD154 mAbinduced platelet activation is strikingly similar to that of heparin-induced thrombocytopenia (HIT) [13], a syndrome strongly associated with thrombosis [15]. In HIT (typically), IgG antibodies form IC with heparin + platelet factor 4 (PF4) antigen on the platelet surface and thus directly activate FccRIIa. FccRIIa is not activated by individual IgG molecules [16], but only by clustered IgG, such as those found in IC. IC-activated FccRIIa induces aggregation and procoagulant microparticle release. Thus, anti-CD154 and HIT antibodies are both associated with thrombosis and both target multivalent platelet surface antigen that clusters IgG and activates FccRIIa. Noting that Bev targets multivalent antigen and is clinically associated with thrombosis, we found that Bev + VEGF IC activate platelet FccRIIa, that heparin greatly enhances this activity, and that ADP-priming improves assay sensitivity [17]. VEGF, like the HIT antigen PF4, is a heparin-binding protein. Bev + VEGF IC exhibited a serotonin release assay (SRA) signature identical to that of HIT antibodies. Platelet aggregometry showed that Bev + VEGF IC exhibit HIT-like

dependence on both heparin and FccRIIa. Given the mechanistic and clinical similarities between Bev, HIT and antiCD154 antibodies, we aimed in this study to identify the functional roles of heparin and platelet surface localization in Bev-induced FccRIIa activation. Importantly, we have identified experimental conditions in which BevÕs prothrombotic activity can, for the first time, be replicated in a mouse model of thrombosis. Laboratory [wild type (WT)] mouse platelets lack FccRIIa [18]. All murine-Bev studies published prior to this report used FccRIIa-deficient mice. To create a murine model of HIT-like thrombosis, McKenzie and colleagues made mice transgenic for human FccRIIa [18]. Unlike WT mice, FccRIIa mice given anti-platelet antibodies develop severe thrombocytopenia [19]. These same researchers later developed a fully-validated model of HIT-like thrombosis by creating mice doubly transgenic for human PF4 and FccRIIa. Therefore, in order to replicate Bevinduced thrombosis in mice (this study), we used FccRIIa transgenic (hFcR) mice, which enabled FccRIIa-mediated platelet prothrombotic activity. Our findings provide a basis for further development of this model and may improve our understanding of the molecular processes driving Bev-associated thrombosis. Methods Reagents

IV.3, anti-FccRIIa mAb, was purified from hybridoma media (ATCC). Anti-CD9 and anti-FLAG mAbs M2 were from BD Biosciences (San Jose, CA, USA) and Sigma (St. Louis, MO, USA). Antibodies were stored in azide-free phosphate-buffered saline (PBS). Bev, cetuximab (3 lM, added to IC solutions as a non-specific IgG competitor), and VEGF were from Genentech (South San Francisco, CA, USA), Centocor (Horsham, PA, USA) and R&D Systems (Minneapolis, MN, USA). Porcine heparin (UFH) was from Sigma, enoxaparin from Sanofi-Aventis (Bridgewater, NJ, USA), and fondaparinux from GlaxoSmithKline (Philadelphia, PA, USA). F(ab) and F(abÕ)2 Bev fragments were generated using Pierce ImmunoPure kits (Rockford, IL, USA). Bev was labeled with Alexa-488 (Invitrogen, Carlsbad, CA, USA). IC were pre-assembled by mixing antibody + antigen (1:1 molar stoichiometry; 1 IgG/ antigen-multimer) in PBS for 5–60 min at room temperature (RT). Serotonin release and aggregation studies

Platelets for SRA and aggregation studies were prepared by centrifuging blood (150 · g/15 min in BD Acid-CitrateDextrose ÔAÕ vacutainers) from normal donors (n = 5, previously known HIT antibody responders), free of antiplatelet drugs and consented by Institutional Review Board approved protocol. Platelets were washed in TyrodeÕs buffer (pH 6.2) containing 1.5 U mL)1 apyrase (Sigma) and resuspended (250/nL) for assay in apyrase-free TyrodeÕs buffer  2008 International Society on Thrombosis and Haemostasis

Bevacizumab-induced thrombosis in transgenic mice 173

±2 mM CaCl2, pH 7.3. For the SRA, platelet-rich plasma (PRP) was incubated with 14C-serotonin (0.1 lCi mL)1, 45 min, 37 C; GE-Healthcare, Piscataway, NJ, USA). Platelets were washed and shaken for 60 min at RT in triplicates with IC or control reagents in microtiter plates. EDTA was then added, platelets were centrifuged (2500 · g/ 5 min), supernatants were diluted 1:50 in scintillation fluid and beta emission radioactivity was measured for 1 min. Washed platelet aggregation studies were conducted in Chrono-Log aggregometers (Havertown, PA, USA). Briefly, baseline was established, 30 lL of agonist was added to 0.27 mL of platelet suspension, and aggregation was monitored. In some experiments platelets were primed with ADP (0.1–1 lM).

Animal studies

Wild-type mice (B6;SJL) and mice transgenic for human FccRIIa [B6;SJL-Tg (FCGR2A)11MKZ/J (hFcR mice)] were obtained from The Jackson Laboratories (Jax). Animals were genotyped by Jax-defined PCR and maintained under Institutional Animal Care and Use Committee guidelines. IC or control reagents (0.2 mL) were injected intravenously into the lateral tail vein (n = 3–10/group). Blood (0.45 mL) was drawn by cardiac puncture from anesthetized mice into 0.1 mL of 3.2% trisodium citrate. Platelets were counted using a Coulter AcT.-diff cell counter (Fullerton, CA, USA). In some experiments, lungs were dissected en bloc (10 min post-injection), rinsed in PBS and formalin-fixed (24 h), and H&E sections were analyzed for evidence of thrombosis.

Flow cytometry

Platelet-associated fluorescence was quantified (25 000 events) with a Coulter flow cytometer (Fullerton, CA, USA). For Fab-dependent binding analysis, platelets (5 · 106) were diluted from PRP into 100-lL reactions and incubated for 10 min with combinations of ADP (0.1 lM), AlexaFluor 488labeled Bev (Bev488; 35.7 lg mL)1), VEGF165 (10 lg mL)1) or VEGF121 (6.5 lg mL)1), IV.3 (20 lg mL)1), Integrilin (20 lg mL)1) and UFH (0, 0.2, or 200 U mL)1). IC were incubated with platelets for 30 min, fixed, washed in PBS and analyzed by flow cytometry.

Statistics

SigmaStat was used for standard deviation (SD.; donor variance) and standard error (SE; assay replicates). T-tests or the Mann–Whitney Rank Sum test were used for two-group comparisons and the Student–Newman–Kuels method for multiple comparisons with unequal variance. Significance was designated at P < 0.05. Results Bev IC mimic HIT antibodies in platelet activity assays

Ouchterlony assay and HPLC

Equimolar amounts of Bev, VEGF165, or VEGF121 ± UFH were diffused for 2 days in agarose gels, washed and stained with GelCode Blue (Pierce Biotechnology, Rockford, IL, USA). Densitometry was performed using an automated densitometry system. HPLC size exclusion chromatography was on a 30 cm Tosoh-G6000PWXL (Montgomeryville, PA, USA) column with fluorescence/UV detection (535 nm/ 210 nm), and TOTALCHROME software (Perkin Elmer, Waltham, MA, USA). Sizing standards were IgM (900 kDa), thyroglobulin (669 kDa), apoferritin (443 kDa), and bevacizumab (150 kDa). Bev488 + VEGF165 ± UFH were run for 45 min at 0.5 mL-PBS/min. Microscopy

Unused cells from flow cytometry experiments were pelleted, immersed in anti-fade media, and layered onto slides. Images were acquired on an Olympus BX51 (Olympus America, Inc., Center Valley, PA, USA) (UPlan Apo 100X/1.35 Oil objective) and a DP70 digital camera (Olympus America, Inc.). DPCONTROLLER/MANAGER software (Olympus America, Inc.) created overlay images in an automated workflow. All images were processed using identical workflow (no image-specific or area-specific modifications). Slides (2 lm sections) from animal studies were stained with hematoxylin and eosin (H&E).  2008 International Society on Thrombosis and Haemostasis

The signature activity profile of HIT antibodies [low-high-low column triplet – meaning, inactive without UFH, active with therapeutic UFH (0.2 U mL)1), inactive with excess UFH] is seen using 500 nM Bev + VEGF165 ± UFH IC (Fig. 1A). IC with IV.3 or Bev-derived Fab/F(abÕ)2 fragments were inactive, confirming FccRIIa dependence (data not shown). Bev forms soluble IC with VEGF121, an isoform lacking the heparinbinding domain (HBD) of VEGF165. Using impedance aggregometry, 250 nM Bev + VEGF121 ± UFH IC (ÔBV121Õ) with ADP priming were inactive (Fig. 1B), indicating the VEGF165 HBD was required. While all study donors responded to Bev + VEGF165 + UFH, only one donor responded (weakly) to Bev + VEGF165 without heparin (Fig. 1B). IV.3 blocked Bev + VEGF + UFH-induced aggregation in PRP, while sub-aggregatory ADP priming increased Bev + VEGF + UFH (ÔBVHÕ) potency (Fig. 1C). Enoxaparin (but not fondaparinux) supported Bev + VEGF-induced aggregation (Supp. Fig. 1). The absence or excess of UFH resulted in lost activity in washed platelet aggregations (Fig. 1D). VEGF + UFH, Bev, VEGF165, VEGF121, or UFH alone were also inactive by aggregation and SRA studies (data not shown). Additionally, the following therapeutic antibodies (control IgGs) were found inactive by aggregation and SRA analysis: panitumumab, cetuximab, trastuzumab, rituximab, and alemtuzumab (data not shown). Figure 1 demonstrates that Bev + VEGF IC induce aggregation and granule release with heparin and FccRIIa dependence, similarly as observed with HIT antibodies.

174 T. Meyer et al

Fig. 1. Platelet activation induced by bevacizumab + VEGF165 IC shows heparin dependence comparable to that of heparin-induced thrombocytopenia (HIT) antibodies. (A) Serotonin release assay (SRA) comparison of Bev + VEGF165 + UFH (ÔBVHÕ) with HIT antibodies. Error bars are SE values. (B) [Bev + VEGF121 ± UFH] + ADP (ÔBV121HÕ) was inactive by impedance aggregometry. (C) IV.3 inhibition and ADP potentiation of Bev + VEGF165 + UFH-induced aggregation in Platelet-rich plasma (PRP). (D) ADP-primed platelets aggregated Bev IC, only with 0.2 U mL)1.

Heparin enhances Fab-dependent Bev IC platelet surface localization

Using flow cytometry, maximal Fab-dependent Bev488 platelet surface binding occurred only with VEGF165 + 0.2 U mL)1 heparin (Fig. 2A, column 5; Fig. 2B). Note that IC binding (not platelet activation) was measured in these experiments. Saturating IV.3 concentrations, present in all samples, excluded Bev-Fc binding to FccRIIa. Binding of Bev488 + VEGF121 + UFH0.2 was not detected (column 8), suggesting the VEGF165 HBD is required for heparin-enhanced surface localization (Fig. 2C) and further confirming that no Bev-Fc to FccRIIa binding occurred. Bev488 + VEGF121 + UFH0.2 (column 8) is the optimal control for this series, being most nearly identical to the maximally active IC (column 5) while lacking both the activity tested (surface binding) and heparin binding activity. Cytometry samples (Bev + VEGF165 + UFH at 0, 0.2 or 200 U mL)1;Fig. 2B) were analyzed microscopically for Alexa488 staining (Fig. 2D). Many platelet singlets and small aggregates, heavily stained with Bev488, were observed with Bev + VEGF165 IC with 0.2 U mL)1 UFH (middle row), but not in the absence or excess of heparin (upper and lower rows). Figure 2 indicates Bev + VEGF + UFH IC localize to platelet surfaces in a heparin- and Fab-dependent (i.e. antigen-dependent) manner, as is also the case with HIT antibodies [20]. Bev IC-induced thrombosis in vivo requires platelet FccRIIa and the VEGF165 heparin-binding domain

We hypothesized that the failure of preclinical murine studies to predict thromboembolism in patients could be explained by

the absence of FccRIIa receptors on mouse platelets [18]. We therefore challenged this hypothesis in vivo – not to replicate the pathology of human thromboembolism, but to determine mechanistically whether FccRIIa enables Bev-dependent prothrombotic platelet activity in vivo. We began by comparing hFcR and WT mice, injecting preformed Bev + VEGF ± UFH IC in sufficient quantities to produce immediate systemic (observable) responses. Only Bev + VEGF + UFH IC having both prothrombotic and surface binding activity in vitro caused severe thrombocytopenia in vivo (Fig. 3A, column 3). Bev + VEGF + UFH IC caused thrombotic thrombocytopenia in hFcR (column 3) but not WT mice (column 2), demonstrating FccRIIa was required. Thrombocytopenia occurred