Amyloid Precursor Protein and Amyloid b Peptide in Human Platelets ROLE OF CYCLOOXYGENASE AND PROTEIN KINASE C*

The main component of Alzheimer’s disease (AD) senile plaques is amyloid- b peptide (A b ), a proteolytic fragment of the amyloid precursor protein (APP). Platelets contain both APP and A b and may contribute to the perivascular amyloid deposition seen in AD. However, no data are available concerning the biochemical mechanism(s) involved in their formation and release by these cells. We found that human platelets released APP and A b following activation with collagen or arachidonic acid. Inhibition of platelet cyclooxygenase (COX) reduced APP but not A b release following those stimuli. In contrast, activation of platelets by thrombin and calcium ionophore caused release of both APP and A b in a COX-independent fashion. Ex vivo studies showed that, despite suppression of COX activity, administration of aspirin did not modify A b or APP levels in serum or plasma, suggesting that this enzyme plays only a minor role in vivo . We examined the regulation of APP cleavage and release from activated platelets and found that cleavage requires protein kinase C (PKC) activity and is regulated by the intracellular second messengers phosphatidylinositol 2-phosphate and Ca 2 1 . Our data provide the first evidence that in human platelets COX is a minor component of APP secretion whereas PKC plays a major role in the secretory cleavage of APP. By contrast, A b release may represent secretion of preformed peptide and is totally independent of both COX and PKC activity. of amyloid b -peptide form to partici-pate was temperature. previously Platelet Aggregation— Platelet aggregation was studied in a light transmission aggregometer PA) in with at Threshold of various platelet agonists were our

Platelet Isolation-Human platelets were isolated as previously described (13). Briefly, whole venous blood was collected from healthy volunteers (all males, 22-40 years old) without stasis in a syringe containing one part of sodium citrate (3.8%) as anti-coagulant and nine parts of whole blood, using a 19-gauge needle to reduce any possible * This work was supported in part by grants from NIA, National Institutes of Health Grant AG11542 and the American Heart Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Recipient of a Medical Scientist Training Program Predoctoral Fellowship from the National Institutes of Health.
Ex Vivo Study-Four healthy volunteers (mean age 30 Ϯ 8 years, 2 male and 2 females, all non-smokers) who had not taken any medication during the previous 2 weeks were each given 1 g of aspirin. Blood samples were taken prior to and 3 h following aspirin ingestion using the methods described above. For plasma preparation, whole venous blood was anti-coagulated with sodium citrate (3.8% v/v) prior to ultracentrifugation at 3000 rpm for 30 min at room temperature. For preparation of serum, venous blood was collected without anti-coagulant and incubated in glass tubes at 37°C for 1 h prior to centrifugation at 3000 rpm for 30 min. Serum and plasma were analyzed for thromboxane B 2 (TxB 2 ), A␤ 1-40 , A␤ 1-42 , and APP levels.
Platelet Aggregation-Platelet aggregation was studied in a light transmission aggregometer (Chrono Log, Haverton, PA) at 37°C in plastic cuvettes with constant stirring at 1000 rpm, as previously described (13). Threshold concentrations of various platelet agonists were determined by measuring the lowest concentration of agonist that caused irreversible platelet aggregation with a trace amplitude of 65% to 85% of maximal light transmission. In our experiments, aggregation was induced by using threshold concentrations of collagen (5 g/ml), arachidonic acid (50 M), thrombin (1 IU/ml), the calcium ionophore A23187 (1 M), or PMA (100 nM). Based on previous published studies, inhibitors were used at the following concentrations: indomethacin (10 M), GF109203X (5 M), wortmannin (100 nM), LY294002 (20 M), BAPTA (200 M), EGTA (2 mM), calpain inhibitor I (25 M), calpain inhibitor II (25 M). Platelet aggregation was continuously measured for 15 min following the addition each agonist. At the end of the incubation time, the reaction was stopped by placing samples in ice. They were then centrifuged at 4°C for 10 min at 12,000 ϫ g to separate platelet pellet from supernatant.
Serotonin and PF 4 Release-Supernatant samples were assayed for release of dense body [ 3 H]serotonin and ␣-granule platelet factor 4 (PF 4 ) as previously described (14). Platelets were incubated for 30 min at room temperature with 0.2 Ci/ml tritiated serotonin, washed twice, and resuspended in Hanks' balanced salt solution containing 5 M imipramine to inhibit serotonin re-uptake. Platelets were treated with agonist for 15 min, and 700 l of the samples was then transferred to a vial containing 70 l of 5 mM EDTA, 5 mM theophylline, 500 M aspirin, and 0.2 g/ml PGE 1 . After centrifugation for 2 min at 12,000 ϫ g, 100 l of supernatant was transferred to a scintillation vial and 2 ml of aqueous scintillation fluid was added. The amount of tritium released was determined by using a scintillation counter and expressed as a percentage of the total amount of tritiated serotonin accumulated by the platelets. Each experiment was performed in duplicate. PF 4 release was determined by a standard ELISA kit, according to the manufacturer's protocol. PF 4 concentration is expressed as a percentage of the total amount released from control platelets after their lysis by sonication. Each experiment was performed in duplicate.
Platelet TxB 2 Formation-Platelet thrombane (Tx) A 2 formation was measured in the supernatant as its hydrolysis product TxB 2 by gas chromatography/mass spectrometry assay, as previously described (13).
Release of Lactate Dehydrogenase-Lactate dehydrogenase (LDH) release was measured using an enzymatic assay (14). Aliquots were taken 15 min after the addition of the agonist, and the supernatant was obtained by centrifuging for 3 min at 12,000 ϫ g. LDH in the supernatants was compared with total LDH released from control platelets after lysis by sonication.
Measurement of A␤  , A␤ 1-42 -Sandwich-ELISA was performed as described previously using mAbs specific for different species of A␤ (15,16). BAN-50 (a monoclonal antibody specific for the first 10 amino acids of A␤) was used as a capturing antibody, and horseradish peroxidaseconjugated BA-27 (a monoclonal antibody specific for A␤  ) and horseradish peroxidase-conjugated BC-05 (a monoclonal antibody specific for A␤  ) were used as reporter antibodies. To calibrate the sensitivity of the ELISA for detecting A␤, synthetic A␤ 1-40 and A␤ 1-42 peptides (Bachem Bioscience Inc., King of Prussia, PA) were used to generate standard curves. The BAN-50, BA-27, and BC-05 mAbs were prepared and characterized as described previously (17).
APP Western Blots-Platelet-associated APP was measured following lysis of platelet pellets in 100 l of radioimmune precipitation buffer (0.5% sodium deoxycholate, 0.1% SDS, 1% Nonidet P-40, 5 mM EDTA in TBS, pH 8.0). 10 l of lysate was resolved on 7.5% polyacrylamide gels for immunoblotting with 6E10 mAb. Immunoreactivity was quantitated by PhosphorImager analysis following application of anti-mouse 125 I secondary antibody. To measure APP released by platelets, and APP levels in serum and plasma, samples were mixed with sample buffer and loaded directly on 7.5% polyacrylamide gels. APP was quantitated by immunoblotting with 6E10 as described above.
Statistical Analysis-Data were analyzed using analysis of variance. Pair-wise comparisons were made using Student's t test where appropriate. Data are displayed as means Ϯ S.E.

APP and A␤ Formation and Release by Human Platelets-
Isolated human platelets incubated for 15 min at 37°C with constant stirring did not show any significant increase in light transmission, indicating no detectable aggregation. Supernatants from these samples had undetectable levels of TxB 2 , serotonin, and PF 4 (Table I) and contained only low levels of APP (Fig. 1A) and A␤ (Fig. 1D). As expected, lysates of unstimulated platelets contained high levels of APP (Fig. 1B). Because APP was detected in these lysates using both an APP ectodomain antibody (Fig. 1B) and an APP cytoplasmic-domain antibody (data not shown), the APP contained in platelets lysates must exist in a full-length, uncleaved form. The addition of collagen (5 g/ml) or arachidonic acid (50 M) to platelets caused irreversible platelet aggregation (Table I), with significant production of TxB 2 , and increased release of serotonin and PF 4 (Table I), but no change in levels of LDH (data not shown). Similar results were obtained when thrombin (1 IU/ml) or calcium ionophore A23187 (1 M) were used (Table I).
Platelets activated by each of these agonists released higher levels of soluble APP (sAPP) than unstimulated platelets (Fig.  1A). Concomitantly, recovery of intracellular full-length APP was decreased by ϳ50% ( Fig. 1, B and C). Because these agonists decreased levels of full-length APP inside platelets and increased levels of soluble, cleaved APP in supernatants, platelet activation must result in cleavage and secretion of APP. Platelet activation also resulted in increased release of A␤ 1-40 , as detected by sandwich-ELISA (Fig. 1D). Although both A␤ 1-40 and A␤ 1-42 levels increased with platelet activation, A␤ 1-42 levels were typically 10-fold lower than A␤ 1-40 levels. These low levels of A␤ 1-42 precluded accurate quantitation by sandwich-ELISA and are therefore not reported.
To test the correlation between platelet activation, APP, and A␤ release, we treated platelets with 0.1-20 g/ml collagen and measured platelet aggregation, TxB 2 generation, and APP and A␤ release. Increasing doses of collagen resulted in a dose-dependent increase of platelet aggregation (light transmission) and TxB 2 , A␤, and APP formation and release (Fig. 2,  A-D).
Role of Platelet Cyclooxygenase-To test the role of cyclooxygenase (COX) in the formation and release of A␤ and APP by human platelets, we utilized two different COX inhibitors, indomethacin and aspirin (18). Inhibition of platelet COX was confirmed by blockade of TxB 2 formation and the effects on platelet aggregation and activation. As expected, incubation of platelets with indomethacin inhibited TxB 2 formation following stimulation with collagen, arachidonic acid, thrombin, and A23187 (Table I). Indomethacin inhibited also platelet aggre-gation and activation following platelet treatment with collagen or arachidonic acid. In contrast, it did not reduce platelet aggregation or serotonin or PF 4 release following activation by thrombin or A23187 (Table I). Similar results were obtained with 100 M aspirin (not shown).
Inhibition of platelet COX activity by indomethacin in collagen-and arachidonic acid-stimulated platelets significantly reduced secretion of APP (Fig. 3A), which was associated with an increased recovery of full-length APP in the lysates (data not shown), suggesting that COX inhibition prevented APP cleavage and not simply APP release. In contrast, indomethacin did not inhibit release of A␤ (Fig. 3B). Indomethacin also did not influence A␤ or APP release in thrombin-and A23187-activated platelets (Fig. 3, A and B). Similar results were obtained with 100 M aspirin (not shown). These results suggest that human platelets release A␤ in a complete COX-independent manner, whereas depending on the stimulus, they can cleave and release APP in either a COX-dependent or COX-independent manner.
Although threshold concentration of collagen induces a COXdependent platelet aggregation, higher levels of collagen can restore full platelet aggregation even when COX is inhibited (19). Thus, to further confirm that human platelets can secrete APP independently from COX activity, we asked if higher concentrations of collagen can restore cleavage and release of APP despite inhibition of COX. As predicted, higher concentration of collagen (4-fold threshold concentration) in the presence of 10 M indomethacin induced a full aggregation response even in the absence of TxB 2 generation (Fig. 4A, trace c). In the same samples serotonin release increased by ϳ40% while PF 4 levels did not change significantly compared with threshold concentration of collagen in the presence of indomethacin (not shown). Similar results were obtained using 100 M aspirin (not shown). Interestingly, high concentration of collagen, in the presence of indomethacin, fully restored APP release in the supernatant (Fig. 4, B and C), confirming our hypothesis that APP cleavage and secretion can occur also via a COX-independent pathway.
Ex Vivo Study-At baseline, plasma TxB 2 levels were almost undetectable (1.5 Ϯ 0.5 ng/ml), whereas serum levels were high (180 Ϯ 20 ng/ml), consistent with the intense platelet activation during serum formation. Three hours after aspirin intake we found that plasma TxB 2 levels were unchanged (1.2 Ϯ 0.4 ng/ml), while serum TxB 2 levels were reduced by 95% (10.1 Ϯ 1.8 ng/ml, p Ͻ 0.001), suggesting that ex vivo platelet COX activity was effectively suppressed. Aspirin intake did not significantly affect APP levels (data not shown) supporting our previous in vitro observation that APP secretion is largely COX-independent. Similar results were observed for A␤ secretion (not shown).
Role of Protein Kinase C-Although APP formation and release from a variety of cell lines is enhanced by protein kinase C (PKC) activation (20,21), to the best of our knowledge, the involvement of PKC in human platelets has not been examined. To investigate the role of this enzyme we first treated platelets with the PKC-activator phorbol 12-myristate 13-acetate (PMA). PMA activated platelets irreversibly and caused increased formation of TxB 2 (Table II). The addition of the PKC inhibitor, GF109203X (5 M), fully blocked PMA-induced platelet activation (Table II). By contrast, GF109203X did not significantly reduce platelet aggregation and activation induced by collagen (Table II). Platelets activated by PMA released A␤ and APP in the supernatant, addition of GF109203X did not significantly influence the release of A␤ but significantly reduced APP secretion (Fig. 5, A and C). Interestingly, even though GF109203X did not affect collagen-induced platelet activation, it prevented APP but not A␤ release in collagenactivated human platelets (Fig. 5, A and C). Similar results were observed when A23187 and thrombin were used as agonists (not shown). Furthermore, platelet preincubation with PMA, which is known to down-regulate PKC activity, and subsequent stimulation with collagen resulted in reduction of APP but not A␤ secretion (not shown). Taken together these results strongly suggest an involvement of PKC activity in APP but not A␤ formation and release from activated human platelets. LDH levels in all the samples incubated with GF109203X were not different from controls (not shown).
Role of Other Intracellular Signaling Events-To elucidate further the molecular mechanisms involved in PKC activation and subsequent cleavage and release of APP and A␤ release from activated platelets, we investigated two intracellular signaling molecules, Ca 2ϩ and phosphatidylinositol 3-kinase (PI3K). We found that the Ca 2ϩ chelator BAPTA inhibited APP secretion following platelet stimulation with A23187 or thrombin (Fig. 5B). Similar results were observed with EGTA (not shown). Furthermore, treatment of platelets with the PI3K inhibitors wortmannin or LY294002 was sufficient to inhibit APP secretion following platelet stimulation with A23187 or thrombin (Fig. 5B). Similar results were obtained when PMA was used to activate platelets (not shown). Interestingly, pharmacological inhibition of these two molecules did not inhibit A23187 and thrombin-induced platelet aggregation, by contrast it suppressed PMA-induced platelet aggregation (not shown). Taken together these findings suggest that both Ca 2ϩ and PI3K are important intracellular mediators of PKC-dependent APP cleavage and secretion in human platelets. BAPTA and EGTA both inhibited A␤ release from platelets activated with A23187 and thrombin, suggesting that calcium is an important intracellular mediator of A␤ secretion in these cells. In contrast, wortmannin and LY294002 did not inhibit significantly A␤ secretion from platelets activated by A23187 or thrombin (Fig. 5D). Finally, because APP processing (22) and A␤ production (23) can be modulated by the activation of the neutral cysteine proteases calpains, we investigated the effects of two membrane-permeable calpain inhibitors. We found that calpain inhibitors I and II were both effective in reducing APP secretion and A␤ release secondary to thrombin and A23187-induced platelet activation (Fig. 6, A and B). No increase in LDH was detected in the supernatant, suggesting that these compounds did not have any toxic effect on platelets. DISCUSSION In the present study we demonstrate that human platelets release A␤ and APP upon their activation by several agonists.
We found that resting platelets contain high levels of fulllength APP, which is cleaved and released into the supernatant following platelet activation. Moreover, we provide novel evidence suggesting that APP release from platelets is only partially regulated by COX activation. In contrast, we found that A␤ secretion by human platelets is totally insensitive to inhibitors of this pathway.
Previous reports have shown that, in other cellular systems, arachidonic acid metabolism may play a role in regulating the secretion of APP (24,25) and that prostaglandin E 2 , a product of this pathway, increases expression of APP by astrocytes (26). Because COX is the key enzyme of the arachidonic acid metabolism in platelets, we investigated its role in platelet A␤ and APP formation by a variety of in vitro and ex vivo experiments. We found that this pathway does not play a crucial role in regulating APP and A␤ release from human platelets. In vitro we observed that inhibition of platelet COX did not affect A␤  and partially reduced APP cleavage and release only when platelets were challenged with agonists that use mainly this metabolic pathway to activate them, such as collagen and arachidonic acid (14). By contrast, no difference was noted in the presence or absence of indomethacin or aspirin when platelets were challenged with agonists that do not use primarily the arachidonic acid metabolic pathway, such as thrombin and calcium ionophore (Fig. 7). This was corroborated by the observation that, even in the presence of indomethacin, very high concentrations of collagen, which are known in this setting to use a COX-independent mechanism to activate platelets, fully restored APP release. Ex vivo we found that serum had higher levels of APP and A␤ than plasma. Our study supports the notion that platelets are a source of both peptides in the circulation (5,8,11). However, the pharmacological suppression of platelet COX activity by aspirin, as assessed by the dramatic reduction of serum TxB 2 , did not interfere with platelet capacity to release APP or A␤, confirming our in vitro observations.
The inability of COX inhibitors to modulate the formation and release of A␤ and APP by human platelets could have some clinical relevance. Epidemiological studies have suggested that non-steroidal anti-inflammatory drugs, which mainly block COX activity, may delay the onset or even reduce the incidence of AD (27,28). Our data suggest that these drugs do not likely exert their effect by altering A␤-or APP-circulating levels derived from platelets. Rather, these compounds may act downstream of A␤ and APP production.
We found that the major species of A␤ released by activated human platelets is A␤ 1-40 , as determined by immunoreactivity with mAb BA27 (specific for A␤  ). This is consistent with the fact that the major component of vascular amyloid is A␤ 1-40 (29). Our finding is in agreement with that of others (11) who identified this peptide as the main component of A␤ from platelet releasates. The same authors were not able to detect any A␤  in platelet supernatants, perhaps due to the limit of detection of their methods.
Several reports have shown that APP secretion is increased by the activation of PKC (20,21). The majority of these studies to date have been conducted using cultured cells, and no data are available on the involvement of PKC in APP and A␤ for-mation and release by human platelets. We found that direct activation of PKC by PMA resulted in release of A␤ and APP from human platelets. On the other hand, addition of a PKC inhibitor and down-regulation of PKC activity, both decreased APP but not A␤ release. Although some caution must be taken when interpreting the action of this potentially nonspecific compound, our findings strongly suggest that PKC is involved in APP formation and secretion by human platelets. Parenthetically, we have previously shown that the same concentration of GF109203X used in these experiments is specific to prevent PKC activation and translocation in human platelets (30). The effect of PKC inhibition on APP release was further confirmed when collagen was used as platelet agonist, suggesting that APP secretory cleavage induced by a more physiological agonist also requires PKC activity. Furthermore, we have provided evidence that, along with PKC activation, two other intracellular signaling molecules, Ca 2ϩ and phosphatidylinositol 2-phosphate, are involved in regulating APP secretory cleavage in human platelets. Interestingly, pharmacological inhibition of both signals coincided with selective inhibition of PMAinduced platelet aggregation and APP secretion, suggesting that these molecules are necessary for PKC activation and APP cleavage and release. This confirms our previous studies in which we have shown that PKC activation in human platelets can be in part modulated by Ca 2ϩ and PI3K activity (31).
We also investigated the effects of two membrane-permeable calpain inhibitors and found that calpain inhibitors I and II were both effective in reducing APP secretion and A␤ release. Although we did not directly examine how calpains function to increase secretion of APP and A␤, we hypothesized that calpains act downstream of PKC and may be required at a late step of secretion, because they were required for both APP and A␤ release (Fig. 7). This hypothesis will be tested by further experiments, which will examine in detail the role of calpains in APP and A␤ release from human platelets.
Because the ADAM (a disintegrin and metalloprotease) family members tumor necrosis factor ␣ cleaving enzyme (TACE) and ADAM 10 have been implicated in the PKC-regulated cleavage and secretion of APP in cultured cells (32,33), it will be interesting to determine if they are also involved in PKCregulated APP formation in platelets.
In contrast with the secretory cleavage of APP by human platelets, we hypothesize that A␤ secretion may represent calcium-dependent release of preformed A␤ rather than increased ␤and ␥-secretase processing of APP. In support of this hypothesis, we found that two different calcium chelators significantly reduced A␤ release from human platelets, whereas a known inhibitor of ␥-secretase (MG132) did not. 2 This finding was further supported by the observation that stimulation of ␣-secretase cleavage of APP (e.g. by PMA treatment) did not result in a corresponding decrease in A␤ production. Indeed, any stimulus of platelets, regardless of its effect on APP processing, increased A␤ secretion. For example, treatment of platelets with PMA in the presence of GF109203X resulted in a reduced platelet aggregation (Table II), but this was still sufficient to stimulate A␤ release.
In summary, human platelets release A␤ and APP when activated by a wide range of agonists. This mainly occurs independent of the COX metabolic pathway and involves multiple signal transduction mechanisms that lead to differential regulation of APP and A␤ secretion. Our data suggest that PKC activation is crucially involved in human platelets, along with PI3K activity and Ca 2ϩ , in APP cleavage and secretion (Fig. 7). On the other hand, calcium plays an important role also in the release of A␤. Finally, activation of cysteine proteases calpains I and II seems to have a relevant function in both events. Interestingly, regulatory kinases, such as PKC, can be modulated by these proteases during platelet activation (34). However, future studies investigating this modulation and interaction are warranted.
Because abnormal secretion of A␤ and APP may contribute to the accumulation of amyloid peripherally in amyloidogenic diseases and in the amyloid angiopathy that occurs in AD, platelets could be a valuable tool for further exploration of these mechanisms.