Apolipoprotein E Inhibits Platelet Aggregation through the L -Arginine:Nitric Oxide Pathway

We have previously reported that plasma apolipopro- tein (apo) E-containing high density lipoprotein particles have a potent anti-platelet action, apparently by occupying saturable binding sites in the cell surface. Here we show that purified apoE (10–50 (cid:109) g/ml), complexed with phospholipid vesicles (dimyristoylphos- phatidylcholine, DMPC), suppresses platelet aggregation induced by ADP, epinephrine, or collagen. This effect was not due to sequestration of cholesterol from platelet membranes; apoE (cid:122) DMPC chemically modified with cyclohexanedione (cyclohexanedione-apoE (cid:122) DMPC) did not inhibit aggregation but nevertheless removed similar amounts of cholesterol as untreated complexes, about 2% during the aggregation period. Rather we found that apoE influenced intracellular platelet signaling. Thus, apoE (cid:122) DMPC markedly increased cGMP in ADP-stimulated platelets which correlated with the re- sulting inhibition of aggregation ( r (cid:53) 0.85; p < 0.01, n (cid:53) 10), whereas cyclohexanedione-apoE (cid:122) DMPC vesicles had no effect. One important cellular mechanism for up-reg- ulation of cGMP is through stimulation of nitric oxide (NO) synthase, the NO generated by conversion of L - arginine to L -citrulline, binds to and activates guanylate cyclase. This signal transduction pathway was impli- cated by the finding that NO synthase inhibitors of distinct structural and functional UK). Other chemicals were from Sigma (Dorset, UK). ApoE (cid:122) DMPC Complexes— ApoE was purified from triglyceride-rich lipoproteins by heparin-Sepharose affinity chromatography (11) and was incorporated into small, unilamellar vesicles of dimyristoyl phos-phatidylcholine (DMPC) liposomes at a 3.75:1 phospholipid:protein ra- tio (12). Protein and DMPC concentrations were measured using com-mercially available methods (Bradford reagent (Bio-Rad Laboratories, Hemel Hempstead, UK) and phospholipid B kit (Wako, Neuss, Ger-*

Human apolipoprotein E (apoE) 1 is a 299-residue protein of molecular mass 34 kDa found in the surface of circulating triglyceride-rich lipoproteins (very low density lipoprotein and chylomicrons, or their remnants) and certain HDL particles (1). Its major function is to mediate hepatic clearance of lipoproteins through interaction with two receptors, the low density lipoprotein or B,E receptor and an apoE-specific receptor, most probably the low density lipoprotein receptor-related protein (2). When the apoE polypeptide is dysfunctional or absent severe hyperlipidemia and atherosclerosis in humans or animal models ensues (1,(3)(4)(5). Although apoE is synthesized predominantly by the liver, macrophages also secrete apoE; this appears important for facilitating local cholesterol redistribution, for reverse cholesterol transport, and for restricting development of atherosclerotic lesions (6). Indeed, atherosclerosis in apoE-deficient (apoE Ϫ/Ϫ ) mice can be prevented by transplantation of normal murine bone marrow cells (5), by macrophage-specific expression of the human apoE transgene (7), or by adenovirus-mediated gene replacement (8).
Recently, we proposed an additional anti-atherogenic role for apoE. We found that HDL-E, the minor apoE-containing subclass of bulk plasma HDL was a powerful inhibitor of agonistinduced platelet aggregation, apparently through interaction with saturable binding sites in the platelet surface membrane (9). Indirect evidence implicated apoE as the active constituent; chemically modifying apoE blocked binding and its anti-platelet action (9), whereas abnormal apoE-enriched HDL from patients with hepatic cirrhosis had a highly potent anti-aggregatory effect that correlated with apoE content (r ϭ 0.70, p Ͻ 0.001) (10). In the present study, we infer that apoE exerts its anti-platelet effect by enhancing production of endogenous nitric oxide (NO); apoE markedly elevated platelet NO synthase activity and intraplatelet levels of cGMP, whereas NO synthase inhibitors restricted its inhibitory action.
Platelet Preparations-Blood was withdrawn with minimal stasis from the antecubital vein of normal volunteers and was mixed with one-ninth volume of 3.8% sodium citrate. Platelet-rich plasma (PRP) was obtained by centrifugation at 200 ϫ g for 15 min at room temperature; platelet-poor plasma was prepared by centrifugation of PRP at 2000 ϫ g for 10 min. Washed platelets were prepared from prostacyclinstabilized PRP (9).
Platelet Aggregation-PRP or washed platelets (80 l) were preincubated for 30 s with Tyrode's buffer (20 l) and aggregation was initiated by addition of increasing concentrations of ADP, epinephrine, collagen, or thrombin at 37°C in a Payton dual-channel aggregometer fitted with 0.1-ml cuvettes (9). The minimum amount of each agonist required to induce secondary aggregation within 3 min was determined, and this "threshold concentration" was used in subsequent assays substituting the Tyrode's buffer for increasing amounts of free apoE, apoE⅐DMPC, cyclohexanedione-modified apoE⅐DMPC (CHD-apoE⅐DMPC), or DMPC liposomes alone. In experiments using NO synthase inhibitors, PRP was preincubated for 10 min at 20°C with 300 M of the chemical analogs of L-arginine, L-NMMA, or L-NAME or with 100 nM of hemoglobin. In the other inhibitory experiments, PRP was preincubated with 100 nM DPI or 3 M Ethyl-ITU for 1 min at 37°C prior to addition of ADP. To examine effects of the soluble guanylate cyclase inhibitor, ODQ, washed platelets (3 ϫ 10 8 /ml) were preincubated at 20°C with 100 nM ODQ for 30 min before addition of either 200 nM GSNO or 50 g of protein/ml of apoE⅐DMPC. Some variation in the extent of apoE⅐DMPC-induced inhibition was evident, but this presumably reflected differences in the individual platelets used or in the particular apoE⅐DMPC preparation, or a combination of these. However, all experiments were repeated several times and were qualitatively reproducible.

Preparation of [ 3 H]Cholesterol-labeled Platelets-A [ 3 H
]cholesterol/ albumin emulsion was prepared as before (13). Platelets were pelleted from prostacyclin-stabilized PRP, resuspended in the [ 3 H]cholesterol/ albumin emulsion for 1 h, and treated again with prostacyclin (300 nM). The mixture was then diluted 50-fold with Tyrode's buffer, centrifuged at 750 ϫ g for 20 min, and the platelet pellet resuspended in buffer. The platelets were used for cholesterol depletion studies immediately after recovering from the prostacyclin effects.
Cholesterol Removal-Aliquots of [ 3 H]cholesterol-labeled platelet suspensions (600 l, 3 ϫ 10 8 cells/ml) were incubated in the aggregometer with buffer, apoE⅐DMPC, or CHD-apoE⅐DMPC at 37°C. At defined time intervals up to 10 min, a portion (100 l) was removed and rapidly centrifuged (12,000 ϫ g, 30 s), and the [ 3 H]cholesterol released into the supernatant (80 l) was measured by liquid scintillation counting. In addition, at zero time an aliquot of the total platelet suspension (100 l) was dissolved in 1 M NaOH, neutralized, and was counted similarly.
Cyclic Nucleotide Assays-Intraplatelet cGMP and cAMP concentrations were measured in PRP with and without a 10-min preincubation at 20°C with the phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine (IBMX) (1 mM). Aggregation was terminated after 3 min by addition of 40 l of 20% HClO 4 ; the samples were then neutralized with 1.08 M K 3 PO 4 (80 l), centrifuged (2000 ϫ g for 15 min at 4°C), and after acetylation were assayed for cGMP and cAMP contents by commercial radioimmunoassay kits (Amersham Int. Plc). All samples were corrected for the cGMP and cAMP content of platelet-poor plasma.
NO Synthase Activity-Washed platelets (10 9 cells) were incubated with or without apoE⅐DMPC (50 g of protein/3 ϫ 10 8 cells) for 10 min at 37°C in a final volume of 1 ml, and the reaction was stopped by addition of 100 l of 10 ϫ homogenization buffer (250 mM Tris-HCl, pH 7.4, containing 10 mM EDTA and 10 mM EGTA). The cells were pelleted in a microcentrifuge for 30 s, resuspended in 100 l of 1 ϫ homogenization buffer, and lysed by two cycles of freezing in liquid nitrogen and thawing on ice. NO synthase activity was measured by the conversion of L-[ 3 H]arginine to L-[ 3 H]citrulline using the NOSdetect assay kit (Stratagene) and expressed as pmol/h per 10 9 platelets. Briefly, 25 l of platelet extract was incubated with 25 l of substrate buffer (50 mM Tris-HCl, pH 7.4, containing 1 mM NADPH, 6 M tetrahydrobiopterin, 2 M flavin adenine dinucleotide, 2 M flavin adenine mononucleotide, 0.2 M calmodulin, 1.2 mM CaCl 2 , and L-[ 3 H]arginine (200,000 dpm)) for 1 h at 37°C, terminating the reaction by addition of 400 l of stop buffer (50 mM HEPES, pH 5.5, containing 5 mM EDTA) and 100 l of cationic resin (Dowex AF 50W-X8). The mixture was transferred to a spin filter, microcentrifuged for 30 s, and the L-[ 3 H]citrulline in the eluate was measured by liquid scintillation counting. Nonenzymic formation of citrulline was controlled for by addition of the specific NO synthase inhibitor, L-NAME (1 mM), to a parallel set of tubes.
Statistics-Results are expressed as means Ϯ S.E., and statistical differences were determined by Student's two-tailed unpaired t test.

Inhibition of Platelet Aggregation by ApoE⅐DMPC-Aqueous
solutions of apoE had no effect on agonist-induced aggregation using either PRP or washed platelets but were highly potent anti-platelet agents when complexed with phospholipid (DMPC) vesicles. Thus, ADP-induced aggregation was inhibited in a dose-dependent manner using a physiological range (10 -50 g/ml) of apoE ( Fig. 1), and similar findings were seen with epinephrine and collagen as agonists. By contrast, little inhibition was noted with thrombin as the aggregating agent, unless either very high amounts of apoE were added (500 g/ml) or the incubations were prolonged (up to 30 min).
Ability of ApoE⅐DMPC to Remove Cholesterol from Platelet Membranes-As cholesterol-deficient platelets respond poorly to agonists (14), we investigated whether the inhibitory action of apoE⅐DMPC might reflect extraction of cholesterol from platelet membranes. Incubation of washed [ 3 H]cholesterol-labeled platelets with apoE⅐DMPC and CHD-apoE⅐DMPC (both 50 g of protein/ml) released similar amounts of cholesterol as a function of time, corresponding to Ͻ1% after our standard 30-s preincubation period and only about 2% after a further 3 min when aggregation studies would be completed ( Fig. 2A). However, CHD-apoE⅐DMPC was an ineffective inhibitor of platelet aggregation (Fig. 2B) compared with its unmodified control (Ϫ2 Ϯ 5% versus 63 Ϯ 4% inhibition, respectively; p Ͻ 0.001), implying that an alternative mechanism to cholesterol sequestration must explain the anti-platelet effect of apoE.
Effects of ApoE⅐DMPC on Intraplatelet cGMP and cAMP Levels-Because attenuation of platelet responsiveness is frequently accomplished by changes in intraplatelet cGMP or cAMP levels (15), we measured the influence of apoE on these cyclic nucleotides during ADP-induced aggregation of PRP. The basal levels of platelet cGMP and cAMP (3.7 Ϯ 1.1 and 11.7 Ϯ 1.9 pmol/10 9 platelets, respectively) were not significantly altered by incubation with 50 g of protein/ml of apoE⅐DMPC vesicles (1.9 Ϯ 0.7 and 9.5 Ϯ 1.5 pmol/10 9 platelets, respectively). However, with threshold concentrations of ADP, the same apoE⅐DMPC complexes produced marked dose-dependent increases in both cGMP (33.9 Ϯ 3.2 versus 13.6 Ϯ 1.5 pmol/10 9 platelets at 50 g of protein/ml of apoE⅐DMPC; p Ͻ 0.001, n ϭ 3) and cAMP (23.5 Ϯ 3.3 versus 7.4 Ϯ 1.1 pmol/10 9 platelets at 50 g of protein/ml of apoE⅐DMPC, p Ͻ 0.001, n ϭ 3) (Fig. 3); these correlated with the observed concomitant inhibition of aggregation after 3 min (r ϭ 0.85 and 0.81 for cGMP and cAMP, respectively; both p Ͻ 0.01, n ϭ 10). No changes in cyclic nucleotides were noted with CHD-apoE⅐DMPC (Fig. 3A), free apoE or DMPC alone.
Effects of IBMX and ODQ on ApoE⅐DMPC-treated Platelets-Levels of cGMP and cAMP are controlled directly by the activities of their synthesizing enzymes, guanylate cyclase and adenylate cyclase, respectively, and catabolizing enzymes, cGMP and cAMP phosphodiesterases (16,17). When platelets were preincubated with the general phosphodiesterase inhibitor, IBMX (1 mM) (18), basal cAMP levels were increased 2-fold (14.8 Ϯ 3.0 versus 7.4 Ϯ 1.1 pmol/10 9 platelets) as a consequence of inhibition of platelet cAMP phosphodiesterase. However, addition of apoE⅐DMPC vesicles to the system did not invoke a further rise in cAMP (14.7 Ϯ 3.4 versus 14.8 Ϯ 3.0 pmol/10 9 platelets at 50 g of protein/ml of apoE⅐DMPC; Fig.  3B), implying that apoE had no effect on adenylate cyclase. By contrast, apoE⅐DMPC still elicited a dose-dependent rise in cGMP levels in the presence of IBMX (10.8 Ϯ 0.8 versus 2.6 Ϯ 1.1 pmol/10 9 platelets at 50 g of protein/ml; Fig. 3A), indicating an apoE-induced increase in synthesis of this cyclic nucleotide rather than a decrease in its catabolism. Further support for this pre-eminent role of cGMP was obtained by use of ODQ, a potent and specific inhibitor of soluble guanylate cyclase (19,20). This reagent impaired, as expected, the anti-aggregatory action of the NO donor, GSNO, but also effectively reversed the anti-platelet effect of 50 g of protein/ml of apoE⅐DMPC vesicles (7.5 Ϯ 8.9 versus 68.7 Ϯ 4.4% inhibition; p Ͻ 0.001, n ϭ 3) (Fig. 4).
Effects of NO Synthase Inhibitors on the Aggregation of ApoE⅐DMPC-treated Platelets-One important cellular mechanism for up-regulation of cGMP is through stimulation of NO synthase (21,22); this enzyme acts on L-arginine to produce NO which then binds to, and hence activates, soluble heme-containing guanylate cyclase, its physiological target (23,24). When we preincubated platelets with L-NMMA or L-NAME, chemical analogs of L-arginine and competitive inhibitors of NO synthase (21,22), the anti-platelet action of apoE⅐DMPC vesicles was essentially blocked (Fig. 5). These inhibitory reactions were enantiomer-specific since D-NMMA was ineffective. Moreover, two additional inhibitors of NO synthase, Ethyl-ITU (25,26) and DPI (27,29), also reversed the anti-platelet effect of apoE (Fig. 5B) without discernible effect themselves on aggregation. Finally, hemoglobin, a competitor for NO binding (29), suppressed the anti-aggregatory effect of apoE⅐DMPC (Fig. 5A).
Stimulation of Platelet NO Synthase Activity by ApoE⅐DMPC Vesicles-Platelet NO synthase is the constitutive, strictly Ca 2ϩdependent form of the enzyme which generates only trace amounts of NO (21,22,30). Indeed, a highly specialized, noncommercial porphyrinic microsensor was required to detect the transient release of low pmol amounts of NO following agonist activation, whereas basal synthesis and release of NO by human platelets was not detected (30). We monitored, therefore, release of NO by measuring the stoichiometric production of  (Fig. 6). This was considered specific NO synthase activity since L-NAME (1 mM) negated any increase.  Fig. 1. The aggregation traces shown are from one experiment but were reproduced in two independent assays.

DISCUSSION
This study substantiates our previous work implicating apoE as the active anti-platelet constituent of HDL-E. The metabolic activity of apoE is sensitive to its lipid environment; purified apoE does not interact with the low density lipoprotein receptor (12), whereas the apoE in surfaces of very low density lipopro-tein and chylomicrons is also inactive unless these lipoproteins are from hypertriglyceridemic subjects (31) or have undergone substantial lipolysis to form remnant particles (32). Similarly, apoE in free solution did not inhibit platelet aggregation, although apoE-phospholipid complexes, which mimic the form secreted by macrophages (1), were potent inhibitors; presumably the DMPC allowed the apoE polypeptide to assume an appropriate orientation and conformation for biological activity (12). Recently, apoE⅐DMPC complexes were reported to inhibit collagen-or thrombin-induced aggregation (33), apparently by avidly extracting cholesterol from platelet plasma membranes and hence impairing arachidonate release and conversion to thromboxane A 2 (35). However, sequestration of cholesterol in our experiments was much lower, Ͻ1% versus 10%, reflecting our shorter preincubation times (30 s versus 30 min), an amount insufficient to be anti-aggregatory with ADP as agonist (13). Indeed, we showed this explanation to be implausible by use of chemically modified apoE; CHD-apoE⅐DMPC failed to suppress platelet aggregation but still removed the same amount of platelet cholesterol as inhibitory apoE⅐DMPC.
Calcium is central to control of platelet reactivity, interacting with diverse second messengers through a myriad of complex, but tightly regulated, signaling pathways (35). Two important control elements for suppression of platelet activation are the cyclic nucleotides, cAMP and cGMP, and agents that increase their intraplatelet levels exert anti-aggregatory effects both in vitro and in vivo. For example, prostacyclin and adenosine limit platelet activation by raising cAMP (36,37), while NO or NO-donating compounds are similarly restrictive by increasing cGMP (19,38). Although apoE induced increases in both cAMP and cGMP, additional experiments implicated a specific stimulation of guanylate cyclase activity and a rise in cGMP as prerequisites for the anti-platelet action of apoE. Thus, ODQ, a potent and selective inhibitor of soluble guany-FIG. 3. ApoE⅐DMPC complexes increase intraplatelet cGMP and cAMP levels in a dose-dependent manner, but preincubation with the phosphodiesterase inhibitor, IBMX, abolishes only the rise in cAMP. PRP (2-3 ϫ 10 8 cells/ml) was preincubated in the absence or presence of IBMX (1 mM) for 10 min and then incubated with 20 or 50 g of protein/ml of apoE⅐DMPC or CHD-apoE⅐DMPC for 10 min at 20°C. Aggregation was initiated by addition of the threshold concentration of ADP at 37°C and allowed to proceed for 3 min, at which point the platelets were immediately processed for cGMP (A) and cAMP (B) measurements. Each assay was carried out in quadruplicate, and the results shown are means Ϯ S.E. of two independent experiments. As expected IBMX treatment increased cAMP levels in the absence of apoE⅐DMPC, but paradoxically levels of cGMP were diminished. However, this finding is readily explained by the rise in cAMP causing a reduced Ca 2ϩ flux; platelet NO synthase is a constitutive form of the enzyme, requiring activation by Ca 2ϩ , and impaired Ca 2ϩ mobilization is known to limit NO release and hence production of cGMP (15,21,22). late cyclase (19,20), was able to reverse the anti-aggregatory action of apoE. Similarly, studies with the general phosphodiesterase inhibitor, IBMX, supported a primary role for cGMP since this reagent abolished the apoE⅐DMPC-induced rise in cAMP but not the dose-dependent increase in cGMP. Interestingly, these findings were consistent with second messenger "cross-talk" (15), namely that the increase in cGMP invoked by apoE in the absence of IBMX may cause inhibition of cAMP phosphodiesterase (39,40), permitting levels of cAMP to rise ( Fig. 3B and Fig. 7).
Several experiments implicated platelet NO synthase in mediating the anti-aggregatory action of apoE through production of NO, a physiological activator of guanylate cyclase. Thus, NO synthase inhibitors of distinct structural and functional types, the amino acid analogs of L-arginine, L-NMMA, and L-NAME, a nonamino acid analog, ethyl-ITU, and the flavoprotein inhibitor, DPI, all reversed the anti-platelet action of apoE. These chemicals inhibit cellular production of NO by binding to NO synthase to displace either L-arginine or, in the case of DPI, essential cofactors (26). Because all reagents were used at concentrations close to their quoted IC 50 values for specific NO synthase inhibition, it seems improbable that they act via diverse NO-independent mechanisms to block the apoE⅐DMPC effects. Indeed, we demonstrated direct involvement of the L-arginine:NO pathway by measuring platelet NO synthase activity; lysates from platelets pretreated with apoE⅐DMPC had a 4-fold increased ability to convert L-arginine to L-citrulline which could be abolished by L-NAME. Furthermore, hemoglobin, which strongly inhibits NO activation of guanylate cyclase by forming a hemoglobin-NO adduct, suppressed the antiplatelet action of apoE. As hemoglobin does not penetrate platelets (41), this implies that apoE generated sufficient NO for secretion; presumably binding of NO by hemoglobin in the extracellular medium created a concentration gradient to reduce NO inside platelets (28) and hence allow aggregation. Interestingly, our observation that apoE⅐DMPC vesicles do not inhibit thrombin-stimulated aggregation provides indirect support for the functional importance of platelet NO synthase; although the reasons are obscure, NO synthase is not activated during platelet aggregation induced by this agonist (21,30).
In summary, as indicated in Fig. 7, we believe that our findings provide clear evidence for the L-arginine:NO signal transduction pathway as the mechanism by which apoE exerts FIG. 5. NO synthase inhibitors prevent the anti-aggregatory action of apoE⅐DMPC complexes. A, aliquots of PRP (2-3 ϫ 10 8 cells/ml) were preincubated with 300 M L-NAME, L-NMMA, D-NMMA, or 100 nM hemoglobin (Hb) for 10 min at 20°C and then for a further 10 min with 10 g of protein/ml of apoE⅐DMPC. Aggregation measurements were carried out in triplicate as described in Fig. 1. The results shown are the means Ϯ S.E. of one experiment but were reproduced in two independent assays. B, aliquots of PRP (2-3 ϫ 10 8 cells/ml) were preincubated with 20 g of protein/ml of apoE⅐DMPC for 9 min at 20°C and then for a further 1 min at 37°C with 100 nM DPI or 3 M Ethyl-ITU. Aggregation measurements were carried out as before and are the means Ϯ S.E. of three independent experiments. its anti-platelet effect. Less clear are the initial events required to activate this pathway. However, the ability of platelets to bind saturably HDL-E (9) and apoE⅐DMPC complexes (33) suggests a distinctive apoE receptor in platelet surface membranes, a proposal consistent with the noninhibitory action of CHD-apoE⅐DMPC. Note also that the failure of apoE to increase basal levels of cGMP was expected; platelet NO synthase requires Ca 2ϩ -calmodulin for activation, and an agonist such as ADP is essential to supply the initial burst of Ca 2ϩ needed for up-regulation (21). Intriguingly, this pathway allows us to speculate that occupation of specific receptors by apoE may "prime" platelets to help attenuate activation when challenged by agonists or other agents.
Although NO synthesized by the constitutive enzyme NO synthase is now well-recognized as an important regulatory mechanism for platelet hemostasis (21)(22)(23)(24), and hence prevention of thrombosis, our finding that apoE substantially augments its production is novel and, potentially, important. Clinically, it may be directly relevant to reports that platelet reactivity increases prevalence (42) and incidence (43) of coronary heart disease and that low apoE and HDL-E are important risk factors (44 -46). But we can also speculate on wider implications. First, that the extracellular NO released from platelets by apoE stimulation of NO synthase activity has the potential to function in a paracrine manner. Second, that apoE might act directly on other cell types to elicit production of NO, including neuronal cells (47)(48)(49) and T-lymphocytes (50,51). Clearly, further experiments are needed to ascertain the biochemical and clinical consequences of this newly identified link in platelets between apoE and NO, two molecules increasingly implicated in cardiovascular disease and neurological disorders (47,52). FIG. 7. Proposed mechanism for apoE-mediated inhibition of agonistinduced platelet aggregation. Occupation of putative cell-surface receptors (apoE-R) by apoE causes up-regulation of the constitutive enzyme NO synthase (cNOS) when an agonist-induced burst of Ca 2ϩ -calmodulin (Ca 2ϩ -CAM) occurs. Some of the NO generated acts on soluble guanylate cyclase to produce inhibitory cGMP, the concomitant rise in cAMP most probably occurring by second messenger "cross-talk." The remainder of the NO produced rapidly diffuses out of the cell and, since extracellular hemoglobin restricts the apoE inhibitory effect, appears to function in a paracrine manner to sustain the dampening influence on platelet activation.