Engagement of Platelet Toll-Like Receptor 9 by Novel Endogenous Ligands Promotes Platelet Hyper-Reactivity and Thrombosis

Preparation of 2-(ω-carboxyalkyl)pyrrole protein adducts and anti-CEP antibodies

The 2-(ω-carboxypropyl)pyrrole bovine serum albumin adduct (CPP-BSA), 2-(ω-carboxyethyle)pyrrole human serum albumin adduct (CEP-HSA), and other CAPs were generated by Paal-Knorr condensation reaction with γ-dicarbonyl compounds as described earlier.^{15, 21} The polyclonal anti-CEP antibody was generated in a Pasteurella-free, New Zealand white rabbit by subcutaneous inoculation of CEP-KLH along with complete Freund’s adjuvant, and the IgG fraction of anti-CEP antibodies was purified using immobilized protein G.¹⁵

Isolation of platelets

PRP and gel-filtered platelets were isolated as described.⁵ Platelet counts were assessed by a Cellometer^™ Auto-M10 (Nexcelom Bioscience Lawrence, MA).

Platelet aggregation

Platelet aggregation was monitored using a Chronolog Model 560VS aggregometer with AGGRRO/LINK version 5.1.9 software, at a stirring speed of 1000 rpm. Aliquots (200μL) of PRP were placed in cuvettes containing magnetic stirrer bars, warmed at 37°C, and stirred for 90 seconds to obtain a stable baseline. Platelet concentration in PRP was adjusted to 2×10⁸ platelets/ml using platelet poor plasma of the same genotype.

Flowcytometry

Human and murine platelet suspensions (2.5×10⁷/mL) prepared by gel-filtration in modified HEPES-Tyrode buffer. P-selectin expression and integrin-α_IIbβ3 activation were assessed as described previously.⁵ Data was acquired using a FACS Calibur instrument (Becton Dickinson, San Jose, CA) and analyzed using the FlowJo 9.0.2 software (Tree Star, Ashland, OR).

Human TLR9 activation assay in HEK-Blue hTLR9 cells

The HEK-Blue^™ hTLR9 cells were cultured in 96-well plate in the presence of indicated concentrations of agonists, and NF-κB-induced SEAP activity was assessed using QUANTI-Blue^™, and a reading at OD 620 nm.

Surface Plasmon Resonance (SPR)

Real-time protein-protein interactions were analyzed using a Biacore3000 (BIAcore AB, GE Healthcare). Functional grade TLR9 peptides (Novus Biological) were bound to CM5 biosensor chips (Biacore) at pH 5.5. Analyte binding to the immobilized ligand was recorded by measuring the variation of the SPR angle, and the results are expressed in resonance units (RU).

Intravital Thrombosis

Intravital thrombosis study was performed using an acute carotid artery injury model as previously described.⁵

Statistical analysis

Values are expressed as means ± standard errors of mean (SEM). The statistical significance was evaluated between two groups of data using two-tailed unpaired Student’s t test. We used two-tailed nonparametric Mann Whitney U test for analyzing the results of ‘intravital carotid thrombosis’, and ‘tail cut bleeding time’ studies. The p values less than 0.05 were considered as statistically significant.

2-(ω-carboxyalkyl)pyrrole protein adducts induce platelet activation

We first assessed the effects on platelets of 2-(ω-carboxyethyl)pyrrole protein (CEP) and 2-(ω-carboxypropyl)pyrrole (CPP) protein adducts (CAPs) of albumin as the most abundant target in plasma for modification by products of lipid peroxidation. In vitro studies demonstrated that these adducts induce significant integrin-α_IIbβ₃ activation and P-selectin expression in human platelets (Fig. 1a–b). In contrast, the sham-modified protein had no effect on platelets. In order to see whether the nature of the modified protein has any role in this effect, we studied adducts of a number of plasma proteins, including albumins from several species, IgG, as well as Keyhole limpet hemocyanin (KLH), a protein that is phylogenetically distant from mammalian proteins. 2-(ω-carboxyalkyl)pyrrole adducts (CAPs) of the tested proteins, but not native unmodified proteins, induced significant platelet activation responses (Fig. 1c–d), indicating that the nature of the modified protein plays no critical role. To test whether this effect is specific to 2-(ω-carboxyalkyl)pyrrole, we used platelet activation assay to test multiple protein adducts formed by oxidized lipids, which resembled hydroxy-ω-oxoalkenoic acids but were not capable of forming pyrroles (Online Table I). None of these non-pyrrole protein adducts had the ability to activate platelets, suggesting that the effect is specific for 2-(ω-carboxyalkyl) pyrrole modification of the proteins. We then tested whether the number of modification per protein plays a role in the effect of CAPs. CAPs with an increasing molar ratio of pyrrole per mole of protein were synthesized and tested in a platelet integrin-α_IIbβ₃ activation assay. We observed a direct correlation between the extent of protein modification by CAPs and the capacity to induce activation of human platelets (Online Figure I).

CAPs activate platelets via Toll-like receptors

CAPs belong to altered-self ligands that are commonly recognized by pattern recognition receptors.^{5, 14, 16} Several pattern recognition receptors are expressed in platelets, including class B scavenger receptors CD36 and SR-BI.⁴ Correspondingly, effects of CAPs were assessed using platelets from wild type mice (WT), CD36deficient (CD36^−/−) mice, and SR-BI deficient (SR-BI^−/−) mice. Platelets isolated from wild type mice were activated by CAPs similarly to human platelets (Fig. 2a). Activation of platelets of CD36^−/− and SR-BI^−/− mice was comparable to that of WT platelets (Fig. 2b, c), ruling out significant involvement of class B scavenger receptors in platelet activation by the CAPs. Platelets also express a number of TLR (Online Figure IIa) that can recognize various pathogen associated molecular patterns as well as altered-self ligands.²² The adaptor protein MyD88 is a common mediator of TLR signaling in cells and is present in platelets (Online Figure IIa). Therefore, to test the general involvement of TLR in CAPs induced platelet activation; we first used platelets from MyD88^−/− mice. CAPs failed to induce platelet activation in MyD88^−/− platelets, demonstrating an absolute requirement for functional TLR-MyD88 signaling. This requirement was specific for CAPs, as integrin-α_IIbβ₃ activation response of MyD88^−/− platelets to the physiological agonists like ADP (Fig. 2d) and thrombin (not shown) was similar to that of WT platelets.

CAPs induce platelet activation in a TLR9-dependent manner

We next tested the involvement of TLR 2, 4, 6, and 9, known to be expressed in platelets,^{7, 9} in platelet responses to CAPs. We used Fab fragments of anti human TLR2, TLR4, TLR6, and TLR9 antibodies, or the chemical inhibitor of TLR4 (CLI 095, data not shown) to block a specific receptor. Unexpectedly, anti-human TLR9 Fab inhibited CAPs induced platelet activation, while no significant effects of Fab fragments to TLR2, TLR4, and TLR6 was observed (Fig. 3a, d–f). The effect of the anti-TLR9 Fab was specific, as it did not affect platelet activation by physiological agonists, such as thrombin or ADP (Fig. 3c). TLR9 expression on human platelets has been previously reported.^{7, 8} Using RT-PCR and Western blot analyses, we detected the expression of TLR9 mRNA and protein in purified murine and human platelets, in the human megakaryocyte cell line Meg-01 and Meg-01 derived platelet like particles (Online Figure IIa). Immunofluorescence microscopy revealed granular staining of permeabilized human and murine platelets with anti-TLR9 antibody and lack of staining of platelets from TLR9 deficient mice (Online Figure IIb). Low levels of TLR9 expression were detected on non-permeabilized resting platelets; however, platelet activation by physiological agonists resulted in a significant increase of TLR9 expression in platelets (Online Figure IIc). To further demonstrate the specific involvement of TLR9 in response to CAPs, we used platelets from TLR9 deficient mice. Response of platelets from TLR9^−/− mice to CAPs was negligible as compared to WT platelets (Fig. 3g–j, Online Figure III), demonstrating that TLR9 is a major mediator of the effects of CAPs on platelets. As an additional control, we tested the effects of CAPs using platelets from TLR2^−/− and TLR6^−/− mice and observed no significant effect on respective integrin-αIIbβ3 activation by CAPs (Online Figure IV).

CAPs are novel and unconventional ligands for TLR9

To demonstrate that CAPs represent new non-canonical ligands for TLR9, we used several approaches. We first studied direct interaction of CAPs with TLR9 by ‘surface plasmon resonance’ (SPR) using human TLR9 immobilized on the surface of a CM5 biosensor chip. As shown in Figure 4a, CAPs (CEP-BSA) interact with the immobilized TLR9 in a concentration dependent manner at high binding affinity (KA = 8.53 × 10⁵) and low dissociation (KD = 1.17 × 10⁻⁶) constants. Interaction increased with the increase in the extent of protein modification by CAPs, i.e the number of pyrroles per protein molecule (Fig. 4b). Pretreatment of immobilized TLR9 with anti-TLR9 antibody completely blocked the binding of CAPs to a TLR9 coated surface (Online Figure Va). To test whether TLR9 binding is specific for 2-(ω-carboxyalkyl)pyrrole modification, we used SPR assay to test control protein adducts formed by oxidized lipids, which resembled hydroxy-ω-oxoalkenoic acids but are not capable of forming a pyrrole. We observed very weak and completely reversible binding of control adducts in comparison to that of CAPs (Online Figure Vb, data for OHdiA-BSA are shown), suggesting that the effect is specific for 2-(ω-carboxyalkyl)pyrrole modification of the proteins. To further demonstrate direct interaction of CAPs and TLR9 in platelets, we incubated platelet lysate with CAPs and performed co-immunoprecipitation assay as described in the Methods section. TLR9 was specifically immunoprecipitated with CAPs, demonstrating the direct interaction of platelet TLR9 and CAPs (Fig. 4c). To demonstrate that CAPs can induce cellular responses via TLR9, even though this receptor is usually localized intracellularly, we used an alternative system - the HEK-Blue^™ hTLR9 cell line that express the human TLR9 (Fig. 4d) and an NF-κB-inducible reporter gene. CAPs induced strong, concentration-dependent and TLR9-mediated signaling in this cell line. The magnitude of the response was comparable to that of established TLR9 specific ligand unmethylated CpG oligodeoxynucleotide (ODN2006) (Fig. 4e, f). Interestingly, while CAPs co-localize with platelet TLR9, we observed a lack of co-localization with the early endosomal marker Rab4 in platelets and partial co-localization with the late endosomal marker Rab9 (Online Figure VIa–c). These data are consistent with the recent finding of TLR9 localization to a novel electron-dense tubular system-related compartment.²³ In nucleated cells TLR9 undergoes proteolytic cleavage upon endosomal acidification, a step required for TLR9 signaling. We observed that an inhibitor of endosomal acidification chloroquine significantly inhibited activation of platelets by CAPs (Online Figure VId–e) but not by physiological platelet agonists ADP, Thrombin, or convulxin, suggesting that the key event leading to activation via TLR9 is similar in platelets and nucleated cells. Taken together, these results demonstrate that CAPs is a new and unconventional ligand for TLR9 that could reach the endosomal TLR9, initiating the downstream signaling cascade.

CAPs induce platelet aggregation and promote platelet hypereactivity in vitro

We next tested whether CAPs can induce platelet aggregation in vitro. We observed that only protein with high degree of modification (≥ 5 pyrroles/per molecule of protein) can induce significant platelet aggregation (Fig 5a). Since TLR9 in platelets is expressed intracellularly in specialized granules, and since we observed that TLR9 expression can be induced by physiological agonists, we tested effects of proteins with level modification of less than 5 pyrroles per molecule on platelet activation and aggregation using platelets primed by either suboptimal concentrations of thrombin receptor activating peptide (TRAP), or weak physiological agonist ADP. Threshold concentrations of TRAP and ADP induced only reversible or no aggregation and activation responses (Fig. 5b, e, h). We observed irreversible and significantly accelerated platelet aggregation in TRAP and ADP primed human (Fig. 5b, c) and murine (Fig. 5e, f) platelets in response to proteins with lower level modification by CAPs. Correspondingly, while CAPs with lower level of modification induced only modest activation of platelets we observed a strong increase in P-selectin expression (Fig. 5d), and integrin-α_IIbβ3 activation (Fig. 5g) when platelets were primed with physiological agonist. This accelerated activation response was TLR9 dependent since pre-incubation of PRP with anti-human TLR9 Fab significantly reduced platelet aggregation induced by CAPs in ADP primed platelets (Fig. 5h, i). Taken together, these data suggest that presence of carboxyalkylpyrrole protein adducts in vivo may promote platelet hyperactivity and aggregation response, especially when physiological platelet agonists are present.

CAPs accelerate thrombosis in vivo in a MyD88 and TLR9 dependent manner

To test whether the presence of CAPs in circulation accelerates thrombosis in vivo and whether MyD88/TLR9 pathway is involved, we compared vessel occlusion times using a ferric chloride induced carotid artery thrombosis model on sex and age matched groups of MyD88^−/−, TLR9^−/−, and corresponding WT mice, which received intravenous injections of either CAPs or sham-modified proteins prior to the vascular damage. Occlusion times were similar in MyD88^−/−, TLR9^−/−, and corresponding WT mice receiving sham-modified proteins. In addition, tail cut bleeding time was similar in TLR9^−/− and WT mice (Online Figure VII). These results and normal activation responses of platelets from MyD88^−/− and TLR9^−/− to physiological agonists in activation assay indicates that there are no major deficiencies in thrombosis in these two knockout mice. Nevertheless, time to complete thrombotic occlusion was significantly shortened in the WT mice that received CAPs as compared to the WT mice that received sham modified protein. In contrast, the presence of CAPs in circulation of MyD88^−/− and TLR9^−/− mice had no significant effect on the occlusion times (Fig. 6a, b). These results demonstrate that the presence of CAPs in vivo can accelerate thrombosis. They also demonstrate that TLR9 modulates thrombosis in vivo when specific ligands for TLR9 are present.

We then tested whether CAPs are present in hyperlipidemic ApoE^−/− mice. Our data show that after intravenous injection, CAPs are rapidly removed from circulation (Online Figure VIII). Nevertheless, concentration of CAPs (data for CEP-adduct are shown) was increased in ApoE^−/− mice on chow diet as compared to WT mice (Fig. 6c). Western diet feeding led to a significant accumulation of CAPs in the plasma of ApoE^−/− mice (Fig. 6c). Platelets are capable of binding of CAPs; therefore, we tested whether platelets can accumulate CAPs in vivo. Indeed, we found a dramatic increase in platelet associated CAPs in ApoE^−/− mice fed a Western type diet (Fig. 6d-e). Immunostaining of the atherosclerotic plaques of aortic root in hyperlipidemic ApoE^−/− mice also revealed the presence of CEP in atherosclerotic plaque, but not in surrounding tissue, (Fig. 6f) indicating that CAPs can also accumulate locally in hyperlipidemia. Taken together, this data confirm that CAPs accumulate in vivo in oxidative stress and demonstrate that presence of CAPs in circulation can modify platelet responses to physiological agonists and thrombosis.

CAPs induce TLR9/MyD88/IRAK1 signaling and require PI3K and Src kinases for platelet activation

Platelets express many components of a TLR signaling pathway downstream of MyD88²⁴, including IRAK4 (S. Panigrahi and E. Podrez unpublished). IRAK1 is one of the key mediators of the TLR signaling pathway downstream of MyD88 and IRAK4. CAPs alone and CAPs in ADP primed human platelets induced phosphorylation of IRAK1 (Fig. 7a). The connection between TLR9/MyD88/IRAK1 activation and platelet integrin activation is not known. MyD88 signaling may involve cGMP-dependent protein kinase. However, we found no effect of a PKG inhibitor DT2-oligopeptide in a concentration range of 50–500nM on CAPs induced platelet activation (Online Figure IX). The PI3K/AKT pathway plays a key role in platelet activation,^25–27 and has been linked to the TLR9 signaling in other cell types.^27–29 We found that in human platelets, CAPs could specifically induce AKT (Ser473) phosphorylation (Fig 7b). PI3K/AKT kinase inhibitor Ly294-002, Syc kinase inhibitor Bay61, Src kinase inhibitor PP2, but not the respective control PP3, also specifically blocked AKT1 phosphorylation in human platelets induced by CAPs at very low concentrations (Fig. 7c). Furthermore, CAPs induced P-selectin expression and integrin-α_IIbβ₃ activation of human platelets could be almost completely blocked by the Src kinase inhibitor PP2 (Fig. 7d). Three AKT isoforms are expressed in platelets. Deletions of specific AKT isoforms in the mouse suggest that they play distinct and redundant roles in platelet responses to physiological agonists.^{26, 30, 31} We observed that CAPs induced platelet activation was significantly reduced in AKT1 and in AKT2, but not in AKT3 deficient platelets (Fig 7e), suggesting that, at least in murine platelets, AKT1 and AKT2 are the major isoforms involved in signaling events downstream of TLR9 and MyD88. Taken together, these findings demonstrate the role of PI3K/AKT and Src family kinases in platelet signaling induced by new unconventional ligands via TLR9/MyD88.