Apolipoprotein A-IV binds αIIbβ3 integrin and inhibits thrombosis

Platelet αIIbβ3 integrin and its ligands are essential for thrombosis and hemostasis, and play key roles in myocardial infarction and stroke. Here we show that apolipoprotein A-IV (apoA-IV) can be isolated from human blood plasma using platelet β3 integrin-coated beads. Binding of apoA-IV to platelets requires activation of αIIbβ3 integrin, and the direct apoA-IV-αIIbβ3 interaction can be detected using a single-molecule Biomembrane Force Probe. We identify that aspartic acids 5 and 13 at the N-terminus of apoA-IV are required for binding to αIIbβ3 integrin, which is additionally modulated by apoA-IV C-terminus via intra-molecular interactions. ApoA-IV inhibits platelet aggregation and postprandial platelet hyperactivity. Human apoA-IV plasma levels show a circadian rhythm that negatively correlates with platelet aggregation and cardiovascular events. Thus, we identify apoA-IV as a novel ligand of αIIbβ3 integrin and an endogenous inhibitor of thrombosis, establishing a link between lipoprotein metabolism and cardiovascular diseases.

A polipoprotein A-IV (apoA-IV) is a 46 kDa exchangeable plasma protein that shares structural features with other apolipoproteins 1 . It is synthesized in the small intestine and can rapidly increase 3-5 fold in response to the absorption of dietary or biliary fat, particularly unsaturated fats 2,3 . After secretion into the intestinal lymphatics, apoA-IV primarily associates with chylomicrons and enters the blood circulation via the thoracic duct. Unlike other apolipoproteins (e.g., apolipoprotein A-I), apoA-IV binds to lipoprotein particles weakly and can be readily displaced by other apolipoproteins 4 . As a result, approximately 80% of circulating apoA-IV is lipid-free 3,5 . Although abundant in the blood (150-370 µg/mL) 6 , the exact function of apoA-IV is still unclear. Roles postulated for apoA-IV to date include regulation of lipoprotein metabolism and reverse cholesterol transport, anti-oxidation, anti-inflammation, and control of food intake 3,7 . Several studies in different human populations have consistently demonstrated that the plasma levels of apoA-IV are inversely correlated with cardiovascular diseases (CVDs) [8][9][10] , but its roles in platelet activity and thrombosis, the major cause of heart attack and stroke, and the leading cause of mortality and morbidity worldwide 11,12 , are unknown.
Platelets are anucleate cells in the blood that play a key role in thrombosis and hemostasis [13][14][15][16] . They are also actively involved in inflammation 17,18 , immune responses [19][20][21] , tumor metastasis 22,23 , and contribute to the initiation of atherosclerosis through interaction with other cells including endothelial cells, leukocytes, and endothelial progenitor cells 24,25 . Recently, the role of platelets in atherosclerosis has been highlighted since increased platelet production accelerates atherogenesis 25 . Upon rupture of the atherosclerotic lesion, platelet adhesion, and subsequent aggregation at the site of injury may lead to thrombosis and vessel occlusion, resulting in myocardial and/or cerebral infarction.
It has been well documented that platelet integrin αIIbβ3 is the dominant integrin expressed on platelets, which plays a key role in platelet adhesion and is required for platelet aggregation 26,27 . Although fibrinogen (Fg), a major ligand of platelet αIIbβ3, was considered to be essential for platelet aggregation (and contributing in part to platelet adhesion), Fg-independent thrombosis occurs [27][28][29][30][31] . This suggests that other ligands of αIIbβ3 exist which are involved in platelet aggregation and thrombosis, but little information is available regarding what they are and how they regulate thrombosis and hemostasis; two opposing but critical biological processes.
Here we show that apoA-IV is a novel ligand of platelet αIIbβ3 integrin. Through competing with Fg and other prothrombotic ligands, it attenuates platelet aggregation, thrombosis, and postprandial platelet hyperactivity, and is an important endogenous protective factor against CVDs.

Results
ApoA-IV is a novel ligand of αIIbβ3 integrin. To search for the unknown ligands of αIIbβ3 integrin, we coated latex beads with RGD (arginine-glycine-aspartic acid) peptide-activated human platelet β3 integrins, and incubated them with human blood plasma. Protein(s) associated with β3 integrin-coated beads were electrophoresed and apoA-IV was identified by mass spectrometry. We found that apoA-IV bound to the surface of platelets activated by adenosine diphosphate (ADP) or collagen, but not to quiescent or β3 integrin-deficient platelets in platelet-rich plasma (PRP) (Fig. 1a). Furthermore, apoA-IV-platelet association was completely blocked by a specific anti-β3 integrin monoclonal antibody (mAb) M1 30 (Fig. 1a). Thus, binding of apoA-IV to platelets is dependent on activated β3 integrin.
Interestingly, apoA-IV bound to activated platelets, but was not internalized by β3 integrins since no apoA-IV was detected in resting platelets (Fig. 1b), which is different from either Fg or fibronectin 29,32 . To confirm whether apoA-IV is a ligand of β3 integrin, we generated biotinylated recombinant apoA-IV and detected the binding between apoA-IV and platelet β3 integrins using a Biomembrane Force Probe (BFP) 33,34 (Fig. 1c-e, Supplementary Fig. 1a, Methods). The BFP brings an αIIbβ3 integrin-coated bead or a platelet target into contact with an apoA-IV bearing force probe ( Supplementary Fig. 1a), and detects the single-molecular adhesions from the pico-force signal measured upon target retraction ( Supplementary Fig. 1b), while a zero force indicates a non-adhesion event 33 ( Supplementary  Fig. 1c). At 1 mM Ca 2+ , apoA-IV bound to ADP-treated platelets with high frequency and had minimal interactions with resting platelets (Fig. 1c). ApoA-IV also bound to Chinese hamster ovary (CHO) cells that express αIIbβ3, but not to the cell lines that express GPIb-IX or αMβ2 integrin (Fig. 1d). Furthermore, at 1 mM Ca 2+ , apoA-IV bound to purified αIIbβ3, but not αVβ3 and α5β1 integrins (Fig. 1e). Importantly, apoA-IV and αIIbβ3 interactions in these experiments were completely blocked by anti-β3 integrin mAb M1 (Fig. 1c-e). Kinetic measurements further showed that the effective two-dimensional (2D) affinity of apoA-IV-αIIbβ3 is approximately 40% of the affinity between Fg and αIIbβ3 (Fig. 2a). Notably, recombinant apoA-IV competitively inhibited the Fg-αIIbβ3 interaction in a dosedependent manner in BFP (Fig. 2b), enzyme-linked immunosorbent assay (Fig. 2c) and a flow cytometric assay (Fig. 2d). Fg also inhibited apoA-IV-αIIbβ3 interaction ( Supplementary  Fig. 1d-e), indicating that apoA-IV and Fg inhibit one another in a reciprocal fashion. Furthermore, recombinant apoA-IV decreased Fg-mediated platelet adhesion (Supplementary Fig. 2a-b). These findings suggest that the occupancy of αIIbβ3 by apoA-IV inhibits Fg-αIIbβ3 interactions. We further found that RGD peptides, the canonical integrin binding motif, decreased apoA-IV binding to αIIbβ3 in a dose-dependent manner ( Supplementary Fig. 3). Thus, apoA-IV is identified as a novel ligand of platelet integrin αIIbβ3.
ApoA-IV inhibits platelet aggregation in vitro. Fg-αIIbβ3 interaction plays an important role in platelet aggregation (Supplementary Fig. 4) and is required for platelet aggregation in anticoagulated blood 27 . Since apoA-IV can inhibit Fg-αIIbβ3 interaction, we first examined the role of apoA-IV in platelet aggregation using apoA-IV-deficient (apoA-IV −/− ) mice 6 and wild-type (WT) littermate controls (apoA-IV +/+ ). Platelet aggregation was significantly enhanced in the PRP of apoA-IV −/− mice following stimulation with various agonists including ADP, collagen, and thrombin receptor activating peptide (TRAP, AYPGKF-NH 2 ) (Fig. 3a). To distinguish whether this enhancement was due to plasma apoA-IV deficiency, but not alteration of platelet function, we switched apoA-IV −/− and apoA-IV +/+ plasma (platelet poor plasma, PPP), and found the enhancement of platelet aggregation was specific to plasma apoA-IV deficiency (Fig. 3b). This was further confirmed using plasma-free gel-filtered platelets, in which plasma was removed during Sepharose 2B column filtration, and no differences were observed in platelet aggregation between apoA-IV −/− and apoA-IV +/+ mice (Fig. 3c). This is consistent with our data demonstrating that apoA-IV was not internalized by platelets (Fig. 1b) and therefore does not affect platelet granule release-related platelet aggregation. Considering the anti-inflammatory roles of apoA-IV 7 , possible alterations of inflammatory factors and other plasma proteins in apoA-IV −/− mice may alter platelet aggregation, thus recombinant human and mouse apoA-IV were used directly to confirm these results. We found that recombinant apoA-IV inhibited both WT mouse and human ADP-induced platelet aggregation in PRP in a dose-dependent manner (80-320 µg/mL) (Fig. 3d, e). Recombinant apoA-IV also reduced platelet aggregation induced by thromboxane A 2 analogue (U46619), calcium ionophore (AY23187) and various doses of collagen (Supplementary Fig. 5a-b). Furthermore, apoA-IV inhibited Fg/von Willebrand Factor (VWF)-independent platelet aggregation ( Supplementary Fig. 5c), which is known to be mediated by β3 integrin 27,[29][30][31] .
To further evaluate the role of endogenous apoA-IV in platelet function, we depleted human plasma apoA-IV with an anti-apoA-IV polyclonal antibody. This depletion enhanced ADPinduced platelet aggregation in healthy blood donors (Fig. 3f). Furthermore, we employed transgenic mice overexpressing mouse apoA-IV (apoA-IV-Tg), which have an approximately 3fold increase in plasma apoA-IV compared to WT levels 35 . Consistently, apoA-IV-Tg mice had markedly attenuated platelet aggregation (Fig. 3g).
To test whether apoA-IV affects other physiological antagonist pathways, we examined platelet cAMP levels and found that they were not significantly altered by apoA-IV treatment (Supplementary Fig. 6a). ApoA-IV also did not directly affect collagenand ADP-induced WT platelet P-selectin expression (Supplementary Fig. 6b). However, it indirectly inhibited the late phase P-selectin expression in an αIIbβ3-dependent manner, suggesting it inhibits αIIbβ3 outside-in signaling-mediated platelet activation and granule release 36 (Supplementary Fig. 6c) and apoA-IV −/− platelets had enhanced P-selectin expression ( Supplementary  Fig. 6d). Consistently, it inhibited shear-induced platelet phosphatidylserine (PS) exposure ( Supplementary Fig. 7a), which is important for cell-based thrombin generation and blood coagulation ( Supplementary Fig. 7b).
These in vitro studies demonstrated that apoA-IV inhibits both Fg-dependent and -independent platelet aggregation via blocking αIIbβ3-ligand interactions, which subsequently inhibits αIIbβ3 outside-in signaling-mediated platelet activation. These suggest its physiological roles in humans and animals.
ApoA-IV aspartic acids 5/13 are required for its inhibition. To understand the structural basis of this antiplatelet effect and localize the potential binding site in apoA-IV for platelet αIIbβ3 integrin, we first deleted both the N-terminal 38 amino acids (Δ1-38) and C-terminal 41 amino acids (Δ336-376), leaving the core helical bundle region we had identified previously 37 (residues 39-335) intact (Fig. 4a). This deletion mutant of recombinant apoA-IV (160 µg/mL) lost platelet-inhibitory function (Fig. 4b). Similar results were observed when only the N-terminus was deleted (Δ1-38, Fig. 4c, d). Interestingly, deletion of the C-terminus (Δ336-376) in the presence of an intact N-terminus enhanced the inhibitory function of apoA-IV (Fig. 4c, d). This suggests that the N-terminus plays a key role in the interaction with αIIbβ3 and the inhibition of platelet function, whereas the * R e s t i n g p l a t e l e t s A c t i v a t e d p l a t e l e t s P l a s m a Fig. 1 ApoA-IV is a novel ligand of αIIbβ3 integrin. a Flow cytometry detection of apoA-IV on activated wild-type platelets (n = 4). b Western blot detection of plasma apoA-IV on activated platelets (n = 4, Supplementary Fig. 16). c-e Biomembrane Force Probe adhesion frequency assays. ApoA-IV bound to activated platelets with high frequency and interacted minimally with resting platelets (c), bound to CHO cells that express αIIbβ3, but not to the cell lines that express GPIb-IX or αMβ2 integrin (d) and bound to purified αIIbβ3, but not αVβ3 and α5β1 integrins (e). Complete inhibition by mAb M1 was indicated. BSA was used as control (n = 50 for each bead/target pair, >3 pairs per bar or data point for c-e). WT wild-type, WB western blot, MFI mean fluorescence intensity, M1 a specific anti-β3 integrin monoclonal antibody, β3 −/− β3 integrin-deficient, CHO Chinese hamster ovary. Unpaired, two-tailed Student's t-test. Mean ± SEM. NS not significant, *P < 0.05, **P < 0.01 and ***P < 0.001 C-terminus negatively regulates this process. Our previous studies have demonstrated that the C-terminus of apoA-IV can interact with the N-terminus, which alters the lipid binding affinity of apoA-IV 38,39 . These new data suggest that this intra-molecular interaction also regulates the exposure of the apoA-IV N-terminus for binding to platelet αIIbβ3.
Notably, apoA-IV does not contain an RGD sequence, the classical binding motif of many integrin ligands. However, through sequence analysis we observed two conserved aspartic acid (D) residues at positions 5 and 13. The D residues in some non-RGD integrin ligands have been reported to be essential for integrin binding 40 . We replaced these residues individually with glutamate (E), a conservative mutation that preserves the negative charge, but increases the side-chain volume of the amino acid. As expected, a mutation in D5E or D13E attenuated the inhibitory function of apoA-IV (Fig. 4e). We then  Contact time, t c (s) Fig. 2 ApoA-IV competes with Fg binding to αIIbβ3 integrin. a Effective 2D affinity of apoA-IV-αIIbβ3 versus Fg-αIIbβ3 within contact area. The adhesion frequency (P a ) depends on the densities of receptor and ligand, therefore the weaker ligand apoA-IV required four-fold higher density (40 µm −2 ) than Fg (8 µm −2 ) yet only generated approximately one-fold higher P a . Affinity to αIIbβ3 for apoA-IV (2.13 × 10 −4 μm 4 ) is 43% of that for Fg (4.99 × 10 −4 μm 4 ) (Grey, Fg-αIIbβ3; dark grey, apoA-IV-αIIbβ3). b Dose-dependent inhibition of Fg-αIIbβ3 by recombinant apoA-IV in Biomembrane Force Probe assay. c, d Dose-dependent inhibition of Fg-αIIbβ3 by recombinant human apoA-IV in ELISA (c) and a flow cytometry assay (d). The inhibitory effect of varying apoA-IV concentrations (0-10 µM) on the interaction between immobilized αIIbβ3 (1 µg) and Fg (1 µM) (c, left). The concentration-response curves were fit to the half maximal inhibitory model (methods) to determine the IC50 for apoA-IV. The apoA-IV IC50 for the inhibition of Fg-αIIbβ3 was found to be 1.32 ± 0.42 µM. Control experiments confirm that the ELISA signal observed is indeed from specific binding interactions and not from non-specific or off-target binding (c, right). FITC-conjugated human Fg binding to the human gel-filtered platelets that were incubated with increasing doses of human recombinant apoA-IV or BSA control was analyzed by flow cytometry (d). Platelets were activated by TRAP (250 µM). P a adhesion frequency, A c contact area, m r αIIbβ3 density, m l Fg or apoA-IV density, k off 2D off-rate, A c k on 2D effective on-rate, rH-ApoA-IV recombinant human apoA-IV. n = 4. Unpaired, two-tailed Student's t-test. Mean ± SEM. NS not significant, *P < 0.05, **P < 0.01 and ***P < 0.001 generated recombinant human and mouse double D mutant (DM) apoA-IV with mutations at both of these residues and found that DM apoA-IV abrogated these inhibitory effects (Fig. 4f, i). Consistently, BFP adhesion frequency assay showed that DM apoA-IV did not significantly bind to αIIbβ3 ( Supplementary Fig. 8), resulting in the loss of inhibitory function of apoA-IV (Fig. 4f, i).
To determine whether the double D mutations affected the structure of apoA-IV, we performed circular dichroism (CD) spectroscopy and thermal denaturation assays. CD wavelength scans revealed excellent superimposition between the WT and DM apoA-IV CD spectra, suggesting no loss of secondary structural elements as a result of the mutation ( Supplementary  Fig. 9a). In addition, both the WT and DM apoA-IV exhibited  Fig. 9b). Taken together, these findings strongly suggest that double D mutations at the apoA-IV N-terminus do not affect the global structure and stability of apoA-IV. Furthermore, the synthetic peptide of the N-terminal 20 amino acids was capable of inhibiting platelet aggregation, while the D5E, D13E, or DM mutations on the peptides attenuated or abrogated its inhibition (Fig. 4j). These studies demonstrate that D5 and D13 at the apoA-IV N-terminus are required for its inhibitory function and are potential binding sites for αIIbβ3 integrin.
Recombinant mouse apoA-IV also markedly inhibited mouse thrombus growth under shear rates that mimic stenotic arteries (e.g., 5000 s −1 ; Supplementary Fig. 10c). Human thrombus growth under both low (Fig. 6a) and high ( Fig. 6b) shear rates in the perfusion chambers was also inhibited by recombinant human apoA-IV. These ex vivo results suggest that apoA-IV is a significant inhibitor of thrombosis at both arterial and venous shear rates. Interestingly, we observed that the inhibitory effect of apoA-IV increased with increasing shear rate (Figs. 5c, d and 6, Supplementary Fig. 10-11). Considering the results obtained with C-terminal truncated apoA-IV in platelet aggregation (Fig. 4c, d), it is possible that high shear stress may disrupt the previously reported intra-molecular contact between the C and N termini 38 , increasing the exposure of N-terminal residues that bind to platelet αIIbβ3. The enhanced platelet inhibition with increasing shear rate may be a critical property of apoA-IV opposing the formation of occluding plugs at the apex of growing thrombi, thus preserving blood supply to the downstream tissues.
ApoA-IV inhibits thrombus growth in vivo. To examine whether apoA-IV affects thrombus growth in vivo, we next utilized two complementary intravital microscopy thrombosis models and a large artery thrombosis model. Thrombus growth and vessel occlusion induced by FeCl 3 injury 29-31,41 were significantly accelerated in apoA-IV −/− mice versus apoA-IV +/+ littermate controls (Fig. 7a). To exclude the possibility that apoA-IV in vivo inhibition was due to anti-oxidant activity 42 as opposed to blockade of platelet αIIbβ3 integrins, we used a cremaster arteriole intravital microscopy thrombosis model which does not involve oxidative stress as it induces mild vascular injury with a laser 43,44 . We found that thrombus growth was also markedly accelerated and thrombus dissolution was slower in apoA-IV −/− mice than apoA-IV +/+ controls (Fig. 7b). Interestingly, we observed that thrombi could reach occlusive size in apoA-IV −/− mice (Fig. 7b), which has not been previously reported in this laser injury model 43 . These findings indicate that endogenous apoA-IV could inhibit thrombus growth and promote thrombus dissolution. Intravenous infusion (9 μg/g) of recombinant mouse apoA-IV attenuated the thrombus growth in apoA-IV −/− mice (Fig. 7b). To examine whether apoA-IV also inhibits thrombus growth in large arteries, a carotid artery thrombosis model 45,46 was used. ApoA-IV −/− mice indeed exhibited accelerated vessel occlusion in carotid arteries (Fig. 7c). We further demonstrated that thrombus growth in WT mice was significantly inhibited by intravenous infusion of recombinant apoA-IV, but not DM apoA-IV, in mesenteric arterioles (Fig. 8a), cremaster arterioles, and carotid arteries Fig. 8b). ApoA-IV significantly prevented stable vessel occlusion in carotid arteries, suggesting its critical inhibitory roles in high shear flow environments (Fig. 8b). These data clearly demonstrate that apoA-IV is an endogenous antithrombotic factor.
To further evaluate the therapeutic potential of recombinant apoA-IV, we compared apoA-IV treatment (9 μg/g) with aspirin (20 μg/g) and clopidogrel (12 μg/g) treatments. The inhibitory effect of apoA-IV was similar to aspirin or clopidogrel treatments, although there was a synergistic effect when apoA-IV and aspirin, or apoA-IV and clopidogrel were used in conjunction ( Supplementary Fig. 12). Importantly, there was no significant alteration of bleeding time or blood loss between apoA-IV −/− mice and littermate controls, or in mice transfused with apoA-IV ( Supplementary Fig. 13). No bleeding tendency has been observed during surgery on these mice in the intravital microscopy experiments after apoA-IV infusion (9 μg/g), although infusing a higher dose (>45 μg/g, i.e., >5 times of physiological level) of recombinant apoA-IV prolonged the bleeding time. These data are consistent with the observations in heterozygous β3 integrin deficient (β3 +/− ) mice that express approximately 50% αIIbβ3 on platelets without obvious bleeding disorders but do have impaired thrombosis, particularly at high shear in vitro and in vivo ( Supplementary Fig. 14a-d). In addition, plasma lipid and cholesterol components in apoA-IV transfused mice were not significantly altered (Supplementary Table 1). Therefore,   We observed that a HFD induced postprandial platelet hyperactivity in WT mice (Fig. 9a). Transgenic mice overexpressing mouse apoA-IV, with an increase in plasma apoA-IV concentration following HFD, markedly attenuated postprandial platelet aggregation (Fig. 9a). Correspondingly, deficiency or decrease of apoA-IV levels significantly enhanced postprandial platelet function (Fig. 9b). Thus, apoA-IV is a postprandial regulator of platelet activity in vivo.   Human apoA-IV inversely correlates with platelet aggregation. Serious life-threatening cardiovascular events peak in the morning, likely related to the increased thrombosis in critical vessels [48][49][50] . Besides changes in apoA-IV postprandial concentrations, whether there are diurnal changes in apoA-IV levels in humans that might contribute to the increased incidence of heart attacks in the morning is unknown. We therefore studied the relationship between levels of apoA-IV and platelet aggregation at 6-h intervals for 24 h in 20 healthy volunteers. We found that humans exhibit the strongest platelet aggregation in the morning around 06:00 (Fig. 10, Supplementary Fig. 15a). Interestingly, plasma apoA-IV levels in humans also exhibited a circadian rhythm, with a peak at midnight and nadir around 06:00 (Fig. 10, Supplementary Fig. 15b). Importantly, our study revealed that the levels of apoA-IV and platelet aggregation are, in most cases (16/20), inversely correlated throughout the day (Fig. 10, Supplementary Fig. 15a-d). Therefore, the lowest level of apoA-IV in the morning likely contributes to enhanced platelet aggregation, leading to more adverse cardiovascular and cerebrovascular events.

Discussion
Thrombotic disorders, such as heart attack and stroke, are the leading causes of mortality and morbidity worldwide. Platelet aggregation mediated by αIIbβ3 integrin and its ligands is essential for the progression of these disorders. In this comprehensive study, using a series of in vitro and in vivo methods (including the single-molecule force spectroscopy technique BFP), as well as conducting human studies in healthy subjects, we identified that apoA-IV is a novel ligand of platelet αIIbβ3 integrin and an endogenous inhibitor of platelet aggregation and thrombosis. Through the long-term moderate but significant attenuation of platelet hyperactivity, particularly during the postprandial period, apoA-IV may also slow down the chronic process of atherosclerosis and other inflammatory diseases 7,35,51 without impairing hemostasis. Our results demonstrate that endogenous apoA-IV can block 20-40% of interactions between αIIbβ3 and Fg or other prothrombotic ligands under static or low shear rate conditions. Notably, clinical experience has not revealed pronounced bleeding effects of a 50% reduction in αIIbβ3 expression (heterozygous individuals who are carriers for Glanzmann's thrombasthenia). The same holds true in β3 +/− mice. However, thrombosis is different from hemostasis and the antithrombotic effect could be achieved before bleeding occurs, as has been well documented in both animal models and antithrombotic drug developments in human clinical trials [52][53][54][55] . Such is the case in our thrombosis models, in which thrombus growth was markedly decreased in β3 +/− mice in vitro and in vivo. Since there are limited human cases of Glanzmann's thrombasthenia, it is currently unknown whether the 50% reduction in αIIbβ3 can also reduce their risk for thrombosis. This concept is consistent with the positive correlation between levels of Fg (αIIbβ3 ligand) and platelet aggregation in vitro, and thrombotic events in patients in vivo 56,57 .
Of note, apoA-IV has an enhanced antiplatelet and antithrombotic effect with escalating shear stress at the apical zone of growing thrombi under flow conditions, likely due to high shear stress that may disrupt apoA-IV intra-molecular interactions and increase exposure of the apoA-IV binding site for αIIbβ3. This may be critical for preventing vessel occlusion at the sites of stenosis. In addition, competitive blockade of the Fg-αIIbβ3 interaction and αIIbβ3 outside-in signaling can subsequently reduce platelet granule release and PS exposure, which may decrease amplification of platelet aggregation, and thrombin generation/clot formation in non-anticoagulated blood 58,59 . These mechanisms may explain its stronger antithrombotic effects in vivo than in vitro. Thus the blockade of αIIbβ3 by apoA-IV is biologically significant in that this level of inhibition does not significantly impair hemostasis, but can predominantly attenuate thrombosis, while a low level of plasma apoA-IV increases the risk for CVDs 8,9 . Indeed, our finding that circadian rhythm of apoA-IV in most blood donors had a negative correlation with platelet aggregation throughout a 24 h period, is suggestive and might at least partially explain the high incidence of heart attack and stroke in the morning compared to the evening.
Recent studies have shown that the high density lipoprotein (HDL) inhibits platelet function by promoting cholesterol efflux, as well as by attenuating agonist-induced platelet activation signaling pathways upon binding to platelet HDL receptors [60][61][62] . Although a small fraction of apoA-IV in circulation is associated with HDL particles, the inhibitory effects observed in our in vitro study do not likely result from other HDL components, but rather via apoA-IV directly antagonizing prothrombotic ligand-αIIbβ3 binding. Recombinant apoA-IV in buffer (i.e., in the absence of HDL components) directly bound to αIIbβ3 in single molecule BFP assays and inhibited gel-filtered platelet aggregation. Double mutations of apoA-IV D5E and D13E, which do not affect global structure and likely do not affect affinity of lipid binding 39 , completely abolished the binding to αIIbβ3 and abrogated the inhibitory effects on platelet function in vitro and in vivo. Furthermore, apoA-IV did not significantly alter platelet cAMP levels. However, the current study could not completely exclude other mechanisms mediated by apoA-IV, such as possible interactions with HDL, LDL or membrane lipid components 25,62 , which may synergistically contribute to the inhibitory effect on platelet aggregation and thrombus growth. It is also conceivable that interactions with HDL and other plasma components may increase the sensitivity of apoA-IV to shear stress, which facilitate the accessibility of the N-terminus to αIIbβ3 integrin. These possibilities merit future investigation.
The displacement of platelet β3 integrin ligands has broad implications not only in thrombotic disorders, but also in platelet-mediated inflammation. Platelet β3 integrins contribute to platelet-leukocyte and platelet-endothelial cell interactions, and the effect of β3 integrin inhibitors on reduction of intimal hyperplasia has been well demonstrated in animal models 63 . We previously reported that platelet Fg-αIIbβ3 integrin interaction can induce platelets to de novo synthesize P-selectin 64 , an important mediator of inflammation, leukocyte rolling and the Th1 immune response, which promotes atherogenesis 17,21,65 . In the current study, although we did not observe that recombinant apoA-IV directly altered platelet activation and P-selectin expression, apoA-IV indirectly inhibited the late phase Pselectin expression by interfering with αIIbβ3-Fg/prothrombotic ligand interactions and the subsequent αIIbβ3 outside-in signaling, which is important for platelet granule release and P-selectin expression 36 . Notably, in the absence of αIIbβ3, apoA-IV did not significantly inhibit the late phase P-selectin expression on thrombin-treated β3 −/− platelets. Consistently, enhanced Pselectin expression was observed on ADP-treated platelets from apoA-IV −/− PRP, where Fg and other prothrombotic ligands can more easily bind to αIIbβ3 and deliver outside-in signaling, which may increase both platelet P-selectin synthesis and releasing. Our data may explain why transgenic mice overexpressing apoA-IV were protected from atherosclerosis 51 .
It has also been reported that postprandial spikes in glucose and lipids may generate excess free radicals, induce inflammation and enhance platelet-endothelial cell and platelet-leukocyte interactions, which may play an important role in atherogenesis and thrombogenesis 25,66 . In our study, we also observed that a HFD induced postprandial platelet hyperactivity. The rapid increase in apoA-IV secretion after lipid intake is therefore important to attenuate postprandial platelet hyperactivity and inflammation, and thus may be physiologically critical to slow the progression of atherosclerosis and thrombogenesis. Furthermore, since platelets have recently been revealed to contribute to tumor metastasis 23 , in which αIIbβ3-mediated platelet-tumor cell heterotypic aggregation is involved 23 , whether apoA-IV can thus attenuate this process remains to be investigated.
This study provides the first direct link between apoA-IV and platelet activity, and the novel mechanisms as to why apoA-IV is a protective factor for CVDs. We demonstrate that apoA-IV, by antagonizing platelet αIIbβ3 integrin activity, attenuates platelet aggregation and thrombosis, postprandial platelet hyperactivity, and may also slow the progression of atherosclerosis (i.e., both early and late stages of atherothrombosis). Furthermore, we identified a circadian rhythm of apoA-IV in humans, which negatively correlates with platelet aggregation and the risk of cardiovascular events. Control of food consumption (e.g., increase of unsaturated fat) 2 , which raises apoA-IV synthesis and secretion, may have a significant impact on the prevention of the early developmental stages of atherosclerosis. Infusion of recombinant apoA-IV may be able to directly intervene in thrombosis and control CVDs and stroke. Hynes (Massachusetts Institute of Technology, USA). Both female and male mice were used with a ratio approximately 1:1 unless where indicated for the intravital microscopy thrombosis models that only male mice can be used 29,43 . Genotypes of all experimental animals were confirmed by polymerase chain reaction (PCR) analysis of tissue DNA. All mice were housed in the research vivarium of St. Michael's Hospital in Toronto, or in each participating institution. Mice or platelets from mice were randomly allocated to experimental and control groups. General conditions of mice were monitored, and the blood flow was evaluated before in vivo experiments. Mice with conditions such as dehydration, or reduced activity were excluded from the study. All animal studies were approved by the Animal Care Committees of St. Michael's Hospital, Toronto, Canada, University of Cincinnati, Cincinnati, USA and The Scripps Research Institute, La Jolla, USA.

Methods
Mouse platelet and plasma preparation. Mice (6-8 weeks old) were anaesthetized and bled from the retro-orbital plexus using heparin-coated glass capillary tubes 29,30 . Blood was collected into a tube containing sodium citrate or ACD (38 mM citric acid, 75 mM trisodium citrate, 100 mM dextrose). PRP was obtained by centrifugation at 300 × g for 7 min. Gel-filtered platelets were then isolated from the PRP using a Sepharose 2B column in PIPES buffer (PIPES 5 mM, NaCl 1.37 mM, KCl 4 mM, glucose 0.1%, pH 7.0). Platelet-poor plasma (PPP) was prepared by centrifugation at 1500 × g for 20 min. The PPP was further centrifuged at 10,000 × g for 5 min to remove the remaining cells.
Human blood preparation. Human blood samples were drawn from antecubital veins of healthy volunteers after providing informed consent. Blood was obtained by venipuncture into Li-Heparin Vacutainers and was allowed to rest at 37°C for 10 min prior to preparation of PRP. Whole anticoagulated blood was spun at 300 × g for 7 min. The PRP was transferred to a fresh tube and stored at 37°C before experiments. Gel-filtered human platelets were prepared as described above for mouse platelets. Human blood samples with significant hemolysis or visible clot formation were excluded from this study. Human platelets were randomly allocated to recombinant apoA-IV treatment or control groups. All experimental procedures using human blood samples were approved by the Research Ethics Board of St. Michael's Hospital, Toronto, Canada, the Institutional Review Board of the Georgia Institute of Technology, and Guangdong Provincial Hospital of Chinese Medicine, Guangzhou, China.
Isolation of β3 integrin ligands. Human β3 integrins were coated on the latex beads and the presence of active β3 integrins on the beads was confirmed by flow cytometry with PAC-1 or anti-β3 integrin anti-sera using a FACScan™ flow cytometer (BD Biosciences, USA). Ligands were precipitated from fractionated, heparin anticoagulated human blood spin filters (Centricon YM-100 and YM-50, Millipore, Canada) according to the manufacturer's protocol. Beads were then washed, resuspended in protein loading buffer, and subjected to 2D electrophoresis. The protein spots of interest were identified by MALDI Quadrupole time-offlight mass spectrometry.
Detection of apoA-IV on activated platelets. PRP was obtained from WT or β3 −/− mice and were pre-incubated with anti-β3 integrin mAb M1 (where indicated) for flow cytometry. Resting or activated (20 µM or 50 µM ADP or 5 µg/mL collagen) platelets were fixed with 4% paraformaldehyde at 4°C. Samples were incubated with goat anti-human apoA-IV antibody (N-20, Santa Cruz), which also detects mouse apoA-IV, and then incubated with fluorescein isothiocyanate (FITC)-conjugated donkey anti-goat IgG (Santa Cruz) at room temperature in the dark. Phosphate-buffered saline (PBS, 0.2 mL) was added to the sample immediately before acquisition. All samples were analyzed by flow cytometry using a calibrated FACScan™ flow cytometer (BD Biosciences). For Western blot analysis, resting or activated (0.5 U/mL thrombin) gel-filtered mouse platelets (4 × 10 7 to 1 × 10 8 ) were lysed, separated by SDS-polyacrylamide gel electrophoresis (PAGE), transferred onto polyvinylidene fluoride (PVDF) membrane, and probed with polyclonal goat anti-human apoA-IV antibody and rabbit anti-goat IgG HRP conjugated antibody (Bio-Techne). Following apoA-IV detection, actin was probed as a loading control after stripping the blots and the actin bands were visualized using HRP conjugated goat anti-mouse actin antibody (Supplementary Fig. 16).
Protein expression and purification. Recombinant human and mouse apoA-IV were expressed in Escherichia coli (E. coli) and purified by immobilized metal affinity chromatography (IMAC) columns (His GraviTrap, GE Healthcare) 67 . Briefly, full-length human and mouse apoA-IV were subcloned into the pET30 E. coli expression plasmid containing 6× His tag and tobacco etch virus (TEV) protease cleavage site. The uncut protein was expressed by 1-thio-β-D-galactopyranoside induction and purified by affinity chromatography utilizing the vector's His tag and Ni 2+ resin columns followed by overnight dialysis in PBS buffer at 4°C. The tag was then cleaved using the TEV protease overnight at 4°C and the cleaved tag was separated using the Ni 2+ columns and collecting the eluent containing pure cut protein.
Site-specific biotinylation of apoA-IV. Site-specific cysteine incorporation for biotinylation was accomplished by site-directed mutagenesis to introduce a cysteine residue at the C-terminus for plasmid construction 68 . The target gene fragment was first PCR amplified and subcloned into pET 30 E. coli expression vector. Using human apoA-IV as the template, the target gene fragment was amplified by PCR with upstream primers containing TEV protease site (GAGAACCTGTACTTCCAGGGCGAG) with an Nco I site, and a downstream primer including C-terminal linker GlyGlyCys (GGGGGCTGC) with a hind III site. The final construct (TEV-apoAIV-GGC) was then inserted into a pET 30 expression vector using restriction endonucleases Nco I and Hind III. All mutations were verified through sequence analysis. The expression plasmid TEV-apoA-IV-GGC was next transformed into E. coli BL21 cells. Protein expression and purification were performed as described above. For in vitro biotin labeling, the Cterminal biotinylation of apoA-IV-GGC was performed by adding maleimide-PEG2-Biotin (Thermo Scientific) into apoA-IV-GGC in PBS (pH7.4) for 30 min at room temperature to produce a C-terminal biotinylated apoA-IV. SDS-PAGE and Western blot analysis were used to verify apoA-IV biotinylation.
Biomembrane-Force-Probe detection of apoA-IV-αIIbβ3 integrin interactions. Purified proteins (i.e., Fg and αIIbβ3) were covalently pre-coupled with maleimide-PEG3500-NHS (JenKem, USA) in carbonate/bicarbonate buffer (pH 8.5). To coat proteins onto the glass beads, 2 μm silanized borosilicate beads (Thermo Scientific) were first covalently coupled with mercapto-propyl-trimethoxysilane (Sigma-Aldrich), then covalently linked to maleimide modified streptavidin (SA, Sigma-Aldrich) and proteins of interest in PBS (pH 6.8) by overnight incubation 33,34 . To coat beads with apoA-IV, SA alone pre-coupled beads were prepared as above. Subsaturating biotinylated apoA-IV was coupled to SA beads by 2 h incubation at room temperature. After resuspending in PBS with 1% BSA, beads were ready for immediate use in BFP experiments. Please refer to the published protocols for details 33,34 .
For the BFP adhesion frequency assay, in each BFP cycle, the target (a receptorbearing bead, cell, or platelet, i.e., αIIbβ3) was driven to approach and contact the probe (a ligand-bearing bead, e.g., apoA-IV or Fg) with a~15 pN compressive force for a certain contact time (t c ) (0.2 s by default, 0.1-1 s for adhesion frequency assay, 1 s for competition assay) that allowed for bond formation and then retracted for adhesion detection. During retraction, tensile force signified an adhesion event and no tensile force indicated a non-adhesion event. Adhesion and non-adhesion events were enumerated to calculate an adhesion frequency (P a ) in 50 cycles for each probe-target pair. As P a depends on t c , the 2D effective on-rate (A c k on ) and 2D off-rate (k off ) can be derived by fitting the P a vs t c curve with the model P a ¼ 1 À expfÀm r m l A c k on ½1 À expðÀk off t c Þ=k off g 34 , where m r and m l are the respective surface densities of receptors and ligands measured by flow cytometry. The 2D effective affinity (A c K a ) is calculated as A c k on /k off . Three to five probetarget pairs were measured at each t c . For the competition assay, P a of Fg-αIIbβ3 interaction was measured with different concentrations of apoA-IV in solution.
Flow cytometry detection of platelet-bound Fg on human platelets. Resting human gel-filtered platelets (1 × 10 8 ) were incubated with increasing doses of human recombinant apoA-IV or BSA control, then incubated with FITCconjugated human Fg (15 min, room temperature). Platelet activation was induced by TRAP (250 μM, 5 min, room temperature) in the dark. The samples were then fixed using 4% PFA. Fg binding to the platelet surface was analyzed by flow cytometry.
In vitro platelet aggregation assays. PRP and PPP from human and mice (apoA-IV −/− , apoA-IV +/− , apoA-IV +/+ control, apoA-IV transgenic mice and WT mice) were obtained by centrifugation of sodium citrate anticoagulated whole blood. Platelet concentration in PRP or PIPES buffer was adjusted to 3 × 10 8 platelets/mL with the use of autologous PPP or PIPES buffer 27,30,41 . Platelet aggregation was performed at 37°C with a sample stir speed of 1000 rpm using a computerized aggregometer (Chrono-Log Corp, Havertown, USA). Platelet aggregation in mouse PRP was stimulated by 1 μM ADP, 1 μg/mL collagen, and 250 μM TRAP. Platelet aggregation in human PRP was stimulated by 2.5 μM ADP and 5 μg/mL collagen. Platelet aggregation in gel-filtered platelets was stimulated by 10 μg/mL collagen and 0. 5U thrombin. For platelet aggregation after switching plasma from the opposite genotype 40 , gel-filtered apoA-IV −/− and apoA-IV +/+ control platelets were suspended in PPP (3 × 10 8 platelets/mL) from either the same or opposite genotype, and aggregation was induced by 250 μM TRAP. In another set of experiments, gel-filtered platelet aggregation in PIPES buffer was induced by 250 μM TRAP. Platelet aggregation induced by various agonists was also measured after adding indicated doses of human or mouse recombinant apoA-IV (incubation for 2 min). The change in light transmission was monitored and recorded for at least 10 min.
Depletion of human plasma apoA-IV. Goat anti-human apoA-IV antibody (C-20, Santa Cruz) was used to deplete apoA-IV in human PPP. Antibody was first coated on the protein G Sepharose beads. The same amount of non-specific goat IgG (Santa Cruz) was used as control. Human PPP was prepared as described above, and added to the beads. The eluate was collected and used for western blot (Supplementary Fig. 17) and platelet aggregometry. For platelet aggregometry, one third volume of normal human PRP was combined with two third volumes of apoA-IV-depleted human PPP or non-specific goat IgG-treated control PPP.
Site-directed mutagenesis. Depletion and point mutations of human and mouse apoA-IV were created by site-directed mutagenesis 37 . Human or mouse apoA-IV DNA was ligated into the pET30 expression vector using the NcoI and Hind III sites with a cleavable N-terminal 6×His tag as described above. We generated the single mutants Δ1-38 apoA-IV, Δ336-376 apoA-IV, D5E apoA-IV, D13E apoA-IV and double mutants 39-335 (Δ1-38, 336-376 apoA-IV), DM (D5E, D13E apoA-IV). Point mutagenesis was performed in the expression vector using the Quick Change Site-Directed Mutagenesis Kit (Agilent Technologies, Canada). All mutations were verified through sequence analysis.
Circular dichroism spectroscopy (CD) and thermal denaturation assays. WT and DM apoA-IV recombinant proteins were purified by size exclusion chromatography on a BioRad SEC650 column equilibrated in PBS. Protein concentration and purity were quantified by A280 and SDS-PAGE analysis, respectively. CD spectral scans and thermal melts of WT and DM apoA-IV (concentrations ranging from 0.4 to 0.5 mg/mL) were acquired on a Jasco J-810 spectropolarimeter using a 1 mm quartz cuvette (Helma). CD wavelength scans were collected between 190 and 250 nm and averaged over five accumulations. Raw data was then converted to molar ellipticity (deg.cm 2 .dmol −1 ). Thermal denaturation assays were carried out at a single wavelength (209 nm) by monitoring the change in ellipticity as the temperature was raised from 20°C to 90°C at a rate of 5°C/min. All thermal denaturation data were baseline corrected, normalized between 0 (folded) and 1 (unfolded) and were fit to a nonlinear biphasic sigmoidal curve in GraphPad (GraphPad Software, San Diego, CA, USA).
Quantitative dynamics of platelet fluorescence intensity were acquired by Slide-Book software (Intelligent Imaging Innovations Inc., USA).
Intravital microscopy thrombosis models. The experiments on intravital microscopy thrombosis models were performed after injuries induced by ferric chloride (FeCl 3 ) 27,29,41,45 or laser 30,31,41,43 . The person conducting the experiment was unaware as to the genotype or treatments received by the mice.
For the mesenteric arterial thrombosis model, thrombus formation in mesenteric arterioles was monitored in 3-to 4-week-old mice 27,29,41,45 . ApoA-IV −/− , apoA-IV +/+ , WT control, and WT mice treated with recombinant or DM apoA-IV were used. Mice were injected with donor-matched fluorescently labeled platelets and visualized under a Zeiss Axiovert 135-invertedfluorescent microscope (Zeiss, Germany). Briefly, blood was collected into ACD from the donor-matched mice. Gel-filtered platelets were prepared and labeled with Calcein AM (1 mg/mL, Invitrogen, Canada) at room temperature for 20 min. Platelets were then injected into the experimental mice via the tail vein with control saline buffer, mouse recombinant apoA-IV (9 μg/g) or DM apoA-IV (9 μg/g). Mice were then anesthetized and the mesentery was externalized. A single mesenteric arteriole of 100-120 μm diameter was chosen and injury was induced by topical application of 30 μL of 250 mM ferric chloride. The time required for formation of the first 20 μm thrombus and the time to complete vessel occlusion was recorded. Images of thrombus formation and dissolution were visualized with a fluorescence microscope.
For the laser-induced cremaster arterial thrombosis model 30,31,41,43 , apoA-IV −/− , apoA-IV +/+ control, and WT control mice (male, 6-8 weeks old) were anesthetized and a tracheal tube was inserted to facilitate breathing. The cremaster muscle was prepared under a dissecting microscope and superfused throughout the experiment with preheated bicarbonate-buffered saline. Antibody, control saline buffer, mouse recombinant apoA-IV (9 μg/g) or DM apoA-IV (9 μg/ g) were administered where indicated by a jugular vein cannula. Platelets were labeled by the rat anti-mouse CD41 antibody (Leo.A1; EMFRET Analytics, Germany; 0.1 μg/g) injection. Multiple independent upstream injuries were induced on a cremaster arteriole using an Olympus BX51WI microscope with a pulsed nitrogen dye laser. The dynamic accumulation of fluorescently labeled platelets within the growing thrombus was captured and analyzed using Slidebook software.
Ferric chloride-induced carotid artery thrombosis. Experiments were performed in >8-week old, 20-35 g, isoflurane-anesthetized C57BL/6J mice of both genders 45,46 . Ten minutes before inducing arterial injury, WT or mutant apoA-IV (9 μg/g), or an equal volume (100 μL) of PBS were injected through a catheter into the right jugular vein. After measuring the baseline blood flow rate (volume/time) for 20 s (s), the left common carotid artery was injured by applying fresh FeCl 3 •6H 2 O solution (0.7 μL, 8%, 0.30 M Fe 3+ ion) onto the exposed adventitia for 180 s. The vessel was then rinsed with warm physiologic saline solution and the flow rate was measured for an additional 1800 s (1980 s total from beginning of injury). Three parameters were calculated: (1) time to first occlusion (O-1), until the flow rate was ≤0.1 mL/min (maximum zero flow error of the probe); (2) time to stable occlusion (O-S), until the flow rate was ≤0.1 mL/min lasting at that level for ≥10 min; and (3) flow index (FI), the ratio of the volume of blood flowed through the artery after-injury (calculated by integrating post-injury flow rate values sampled every second for 1800 s) to the volume flowed through in an uninjured vessel (calculated from the average pre-injury flow rate). For the carotid artery thrombosis model in apoA-IV −/− and apoA-IV +/+ control mice, the right carotid artery was dissected following anesthetization and held with a miniature Doppler flow probe (TS420 transit-time perivascular flowmeter, Transonic Systems Inc., USA). Carotid artery injury was induced with a strip of Whatman filter paper saturated with 10% ferric chloride. Blood flow was monitored until complete vessel occlusion was observed.
Bleeding time assay. ApoA-IV −/− , apoA-IV +/+ , and WT mice (6-8 weeks old) were injected with recombinant apoA-IV where indicated via the tail vein 40 min before injury. Mice were anesthetized and maintained at 37°C on a heating pad during the experiment. The tip of the tail (5 mm) was cut off with a sharp scalpel, and the tail was immediately placed into warm saline at 37°C. Bleeding time was recorded as the time to cessation of blood flow (bleeding stopped for >10 s). Blood loss was calculated by counting the red blood cells in the PBS fraction.
Postprandial platelet function assay. C57BL/6 J WT, apoA-IV transgenic (apoA-IV-Tg), apoA-IV −/− , apoA-IV heterozygous (apoA-IV +/− ), and apoA-IV +/+ mice (6-8 weeks old) were fasted for 12 h, followed by a HFD for 3 h. Blood was then collected from these mice following fasting or HFD condition. Gel-filtered platelets (2.5 × 10 8 /mL) and PPP were prepared respectively. To minimize interference in the light transmission of PRP, which is commonly caused by postprandial chylomicrons and other lipid particles 69 , 125 µL gel-filtered platelets were incubated with its PPP (125 µL) for 20 min under 37°C. The same volume of PPP and PIPES buffer were used as blank control. Platelet aggregations were then induced by collagen (10 µg/mL).
Enzyme-linked immunosorbent assay to detect αIIbβ3 binding. A 96-well plate (Nunc MaxiSorp) was coated with purified human αIIbβ3 or control proteins (BSA and β3 −/− platelet lysate) by incubation of 0.01 µg/µL protein in binding buffer NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05806-0 ARTICLE (TRIS buffered saline with 0.05 % TWEEN-20 and 1 mM each of MgCl 2 , MnCl 2 and CaCl 2 ) at 4°C overnight. Incubation of 3% skim milk (ED Millipore) and 2% TWEEN-20 for 1 h at 37°C was used for blocking. For Fg binding experiments, wells were incubated with the appropriate Fg concentration in binding buffer for 1 h at 37°C. Wells were then incubated with anti-fibrinogen mouse IgG (Sigma Aldrich) followed by incubation with anti-mouse IgG-HRP (Santa Cruz), both at 37°C for 1 h. For apoA-IV binding experiments the desired protein concentration in binding buffer was added to each well and incubated 1 h at 37°C followed by the incubation of streptavidin-horseradish peroxidase (HRP) conjugated protein (Abcam), at 37°C for 1 h. Peroxidase substrate, o-Phenylenediamine dihydrochloride (OPD, Sigma Aldrich) was prepared at 0.4 mg/mL in 0.05 M phosphate citrate buffer at pH = 5.0 with 0.4 µL/mL of 30 % H 2 O 2 . The OPD peroxidase reaction was stopped after 1 h with 2 M H 2 SO 4 and the absorbance was read at 492 nm. Between incubation steps the plate was washed with copious amounts of TRIS buffered saline with 0.5% TWEEN-20.
Human study to evaluate the correlation between apoA-IV and platelet aggregation. Twenty healthy volunteers taking no medications (mean (range), 25.8 (20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31) years; body mass index, 23.6 (19.9-29.6) kg/m2; 10 women and 10 men) were recruited in this study. All participants gave written informed consent in accordance with the Declaration of Helsinki. The meals, sleep, physical activity, room temperature, and light the volunteers received were all controlled. Blood samples were collected prior to meals via venipuncture at 12:00, 18:00, 24:00, and 06:00 across a 24-h period. PRP and PPP were prepared as described above. Plasma samples to be assayed for the concentration of apoA-IV were stored at −80°C until analyzed by ELISA. Platelet aggregation was induced by ADP (6 µM) and assays were performed immediately using fresh PRP and PPP within 1 h after blood samples were collected.
Enzyme-linked immunosorbent assay to measure apoA-IV plasma level. Plasma apoA-IV concentrations were determined using an ELISA that employs affinity purified goat anti-human apoA-IV polyclonal antibody as the capture antibody (apoA-IV N-20; Santa Cruz). ApoA-IV N-20 was raised against a peptide mapping near the N-terminus of apoA-IV of human origin. Recombinant human apoA-IV proteins with known concentrations (determined by Nanodrop) served as a standard. ApoA-IV N-20 (1 µg/mL) was coated on the surface of ELISA plates (Nunc MaxiSorp) at 4°C for overnight. 3% Skim Milk (Millipore) with 2% Tween-20 in PBS buffer were used for blocking the wells at 37°C for 1 h. We then incubated recombinant human apoA-IV or human plasma samples (100 times dilution) at 37°C for 3 h, followed by incubation with the detection antibody, apoA-IV G-8, a mouse monoclonal antibody raised against a peptide mapping near the C-terminus of apoA-IV of human origin (2ug/mL; Santa Cruz) at 37°C for 2 h. The mouse IgG kappa binding protein (m-IgGκ BP) conjugated to HRP (1ug/mL; Santa Cruz) was then incubated in the wells at 37°C for 1 h. OPD substrate and the stop solution were used as described above. Three to five times washing were performed between each step. Read absorbance at 492 nm and 650 nm in a plate reader.
Statistical analysis. Data are presented as mean ± SEM. Statistical significance was assessed by unpaired, two-tailed Student's t-test and non-parametric Kruskal-Wallis one-way analysis of variance followed by Dunn's test for multiple paired comparisons. The sample size of each experiment was estimated based on our previous experience and the publications from others 32,33,37,47,50 .

Data availability
The data that support the findings of this study are available from the first author and the corresponding author upon reasonable request.