Aspirin inhibits the production of proangiogenic 15(S)‐HETE by platelet cyclooxygenase‐1

Regular consumption of low‐dose aspirin reduces the occurrence of colorectal, esophageal, stomach, and gastrointestinal cancers. The underlying mechanism is unknown but may be linked to inhibition of angiogenesis. Because the effective doses of aspirin are consistent with the inhibition of cyclooxygenase‐1 in platelets, we used liquid chromatography with tandem mass spectrometry analyses and immunoassays of human platelet releasates coupled with angiogenesis assays to search for the mediators of these effects. Blood or platelet‐rich plasma from healthy volunteers stimulated with platelet activators produced a broad range of eicosanoids. Notably, preincubation of platelets with aspirin, but not with a P2Y12 receptor antagonist, caused a marked reduction in the production of 11‐ hydroxyeicosatetraenoic acid (HETE) and 15(S)‐HETE, in addition to prostanoids such as thromboxane A2. Releasates from activated platelets caused cellmigration and tube formation in cultured human endothelial cells and stimulated the sprouting of rat aortic rings in culture. These proangiogenic effects were absent when platelets were treated with aspirin but returned by coincubation with exogenous 15(S)‐HETE. These results reveal 15(S)‐HETE as a major platelet cyclooxygenase‐1 product with strong proangiogenic effects. Thus, 15(S)‐HETE represents a potential target for the development of novel antiangiogenic therapeutics, and blockade of its production may provide a mechanism for the anticancer effects of aspirin.—Rauzi, F., Kirkby, N. S., Edin, M. L., Whiteford, J. Zeldin, D. C., Mitchell, J. A., Warner, T. D. Aspirin inhibits the production of proangiogenic 15(S)‐HETE by platelet cyclooxygenase‐1. FASEB J. 30, 4256–4266 (2016). www.fasebj.org

Epidemiologic data have established that regular consumption of aspirin reduces the occurrence of some cancers, notably colorectal, esophageal, stomach, and gastrointestinal (1)(2)(3)(4). The mechanism underlying this remarkably significant effect is unknown, although it may be in part linked to reductions in angiogenesis.
Aspirin acts by irreversibly inhibiting the cyclooxygenase (COX) enzyme, and this explains its anti-inflammatory, analgesic, antipyretic, and antiplatelet effects (5)(6)(7). These effects of aspirin are produced at notably different dosages; the anti-inflammatory effects are seen only at very high dosage (many grams), the analgesic effects require high to medium dosage (around 600 mg), the antipyretic effects are exerted at medium dosage (300-600 mg), and the antiplatelet effects are fully achieved with a low dosage (75-100 mg) (7). Notably, the anticancer effects of aspirin are displayed at low dosages (4), consistent with those that inhibit platelets (8) and markedly below dosages required for analgesic or antiinflammatory effects.
Platelets are blood components central to hemostasis. Upon activation, platelets release large amounts of arachidonic acid (AA) from the membrane phospholipids largely through the activity of group IV A cytosolic phospholipase A 2 (9,10). The AA released is rapidly metabolized by multiple enzymatic pathways (9,10). The best characterized of these pathways within platelets is the conversion of AA to thromboxane (TX)A 2 , which is dependent upon the actions of COX-1 and TX synthase. TXA 2 is a potent prothrombotic hormone that drives platelet aggregation and thus the formation of blood thrombi. Regular consumption of low-dose aspirin irreversibly inhibits COX-1. This in turn ABBREVIATIONS: AA, arachidonic acid; AUC, area under the curve; COX, cyclooxygenase; EIA, enzymatic immune assay; FBS, fetal bovine serum; HETE, hydroxyeicosatetraenoic acid; LC-MS/MS, liquid chromatography with tandem mass spectrometry; LT, leukotriene; PAM, prasugrel active metabolite; PG, prostaglandin; PRP, platelet-rich plasma; TX, thromboxane blocks the formation of TXA 2 , which leads to reduced platelet activation and diminished thrombus formation. In contrast to the lack of insight regarding the ability of lowdose aspirin to reduce cancer, blockade of the production of TXA 2 provides a clear mechanism to explain low-dose aspirin's antithrombotic effectiveness (5)(6)(7)(8).
Building upon the evidence that platelet COX-1 is the therapeutic target of low-dose aspirin, we characterized the eicosanoids produced by activated platelets and their sensitivities to clinically relevant antiplatelet levels of aspirin. The eicosanoids identified as major AA-derived platelet products were then investigated for their influence on cell targets within the vasculature (i.e., platelets, leukocytes, and endothelial cells that could be relevant to the influences of aspirin upon cancer progression). From these studies we propose that 15(S)-hydroxyeicosatetraenoic acid (HETE) is a previously unidentified major COX-1 product of platelets whose inhibition could provide a mechanistic explanation for some of the anticancer effects of aspirin.

Blood collection and ethics
Blood from healthy volunteers was collected by venepuncture into lepirudin (250 mg/ml). All experiments were subject to written informed consent and appropriate local ethical approval (healthy volunteer samples: St. Thomas's Hospital Research Ethics Committee, reference 07/Q0702/24; patient samples: South East National Health Service Research Ethics Committee) and in accordance with Declaration of Helsinki principles. As reported previously, platelet counts were made to confirm that all samples were within the normal range (9,(11)(12)(13).
Preparation of whole-blood and platelet-rich plasma samples for liquid chromatography with tandem mass spectrometry analysis Blood was incubated with vehicle, aspirin (100 mM), the P2Y 12 receptor antagonist prasugrel active metabolite (PAM) (3 mM; a gift from AstraZeneca), or aspirin + PAM. Part of the blood was centrifuged (175 g; 15 min) to obtain platelet-rich plasma (PRP). Platelets in both whole blood and PRP were then activated under static conditions (in initial experiments) or under stirring conditions in a light transmission aggregometer by the addition of collagen (30 mg/ml) (Takeda Pharmaceuticals, Deerfield, IL, USA), TRAP-6 (30 mM) (Sigma-Aldrich, St. Louis, MO, USA), or A23187 (50 mM) and incubation for 5 min. Plasma was then quickly separated from the samples by centrifugation (2000 g, 5 min, 4°C) and stored at 280°C for eicosanomic analysis, as previously described (14,15). In brief, HyperSep Retain PEP SPE cartridges (Thermo Fisher Scientific, Waltham, MA, USA) were preconditioned with a solution of 0.1% acetic acid/5% methanol and spiked with 30 ng each of internal standards. Plasma (0.25 ml) was diluted in 0.1% acetic acid/5% methanol containing 0.009 mM butylated hydroxytoluene and added to the column. Samples were then washed with two volumes of 0.1% acetic acid/5% methanol, eluted in 1 ml of ethyl acetate and 1 ml methanol, dried by vacuum centrifugation at 37°C, and reconstituted in 30% ethanol. AA-derived metabolites were then separated by reverse-phase HPLC on a 1 3 150 mm, 5 mm Luna C18 (2) column (Phenomenex, Torrance, CA, USA) and quantified using a MDS Sciex API 3000 triple quadrupole mass spectrometer (Applied Biosystems, Foster City, CA, USA) with negative-mode electrospray ionization and multiple reaction monitoring. Data were captured and analyzed using Analyst 1.5.1 software. Relative response ratios of each analyte were used to calculate concentrations, and extraction efficiency for each sample was calculated based on recovery of the internal standards.

Enzymatic immune assay for 15(S)-HETE
Enzymatic immune assay (EIA) assay was performed in accordance with the kit insert (Cayman Chemical Co., Ann Arbor, MI, USA). In brief, PRP and whole blood samples were diluted 1:10 and assayed in parallel to known 15(S)-HETE standards, a maximum binding control, nonspecific binding control, and blank on a 96-well plate coated with goat anti-IgG antibodies.

Tube formation assay
Immortalized human microvascular endothelial cells (line HMEC-1; Centers for Disease Control and Prevention, Atlanta, GA, USA) were cultured in RPMI 1640 medium with 10% fetal bovine serum (FBS) onto 48-well plates coated with 100 ml/well basement membrane extract in the presence of vehicle or 15(S)-HETE. Tube formation was visualized by time-lapse microscopy (magnitude, 34) over 16 h using an inverted microscope (IX81; Olympus, Tokyo, Japan) with a digital camera (Orca-ER; Hamamatsu, Hamamatsu City, Japan). Image acquisition was performed using Cell M software (Olympus). After the 16 h incubation, images were taken in phase (original magnification, 310) and in fluorescence (original magnification, 310) by labeling cells with calcein AM. The number of branching points was counted after 4, 8, and 16 h incubation using Image J software (National Institutes of Health, Bethesda, MD, USA). In experiments investigating the effects of platelet incubates on HMEC-1 cells, platelets in PRP were activated under stirring conditions in a light transmission aggregometer by the addition of collagen (30 mg/ml) (Takeda) or TRAP-6 (30 mM) (Sigma-Aldrich) and incubation for 5 min [as for preparation for liquid chromatography with tandem mass spectrometry (LC-MS/MS) analysis] before being added directly to the cells together with an equal volume of RPMI 1640 medium + 2% FBS. For experiments investigating the effects of platelet releasates, plasma was quickly separated from the samples by centrifugation (2000 g, 5 min, 4°C) before combination with RPMI 1640 medium + 2% FBS and addition to the HMEC-1 cells.

Cell migration assay
Confluent HMEC-1 cells on 6-well plates were scraped in a straight line using a p200 tip to create a "scratch" and incubated (37°C; 5% CO 2 ) in RPMI 1640 medium with 10% FBS in the presence of vehicle or 15(S)-HETE. HMEC-1 migration was then visualized by time-lapse microscopy (magnitude, 310) over 20 h using an Olympus IX81 inverted microscope with a Hamamatsu Orca-ER digital camera, and image acquisition was performed using Cell M software (Olympus). The percentage of area covered by cells throughout the 20 h incubation was calculated. Platelet releasates were prepared as previously described.

Rat aortic ring assay
Thoracic aortas from 180-200 g Wistar rats (Charles River Laboratories, Wilmington, MA, USA) were sliced into equal sections and incubated (37°C; 5% CO 2 ) overnight in serum-free Opti-MEM (Invitrogen, Carlsbad, CA, USA) before being embedded in type I collagen (1 mg/ml) in E4 medium (Invitrogen). Rings were then incubated (37°C; 5% CO 2 ) in Opti-MEM (Thermo Fisher Scientific) plus 1% FBS (PAA Laboratories/GE Healthcare, Little Chalfont, United Kingdom) in the presence of platelet releasates and/or VEGF (10 nM), 15(S)-HETE (1 mM), or vehicle. Emergent angiogenic sprouts from rat aortas were counted after 4 d in culture. All animal experiments were conducted in accordance with the British Home Office regulation (Scientific Procedures) Act. Images were captured using an Olympus IX81 inverted microscope with a Hamamatsu Orca-ER digital camera, and image acquisition was performed using Cell M software (Olympus). Platelet releasates were prepared as described above.

Eicosanoid formation in whole blood and PRP
To test the source of eicosanoids produced in whole blood, incubations were made of blood or PRP from healthy volunteers together with A23187 (50 mM), collagen (30 mg/ml), TRAP-6 (30 mM), or PBS. Although a broad range of eicosanoids was detected ( Table 1), those produced in the highest amounts to collagen and TRAP-6 were 12-HETE, TXA 2 , 15-HETE, and 11-HETE, with similar levels in PRP and whole blood (Fig. 1). 5-HETE was a very prominent product of whole blood stimulated with A23187 but was absent in PRP.

Eicosanoid formation in whole blood and PRP in the presence of antiplatelet treatment
Focusing on the major eicosanoid products identified above, aspirin alone or in combination with prasugrel blocked the production of TXA 2 , PGE 2 , and PGD 2 , along with 11-HETE and 15-HETE, in whole blood and in PRP. EIA indicated that the 15(S)-HETE enantiomer was a major product and that aspirin inhibited its formation ( Table 2). The addition of 15(R)-HETE standards to the EIA verified that the assay was highly selective for 15(S)-HETE, providing further confirmation that 15(S)-HETE was produced by the activated platelets. The productions of 12-HETE in blood and PRP were not significantly affected by aspirin or PAM but were reduced by the combination of aspirin and PAM (Fig. 2). Levels of 5-HETE did not increase after stimulation of blood or PRP with collagen or TRAP-6 and were not affected by any of the treatments tested.

Effect of platelet COX-1-derived prostanoids and 11(R)-, 15(S,R)-HETE on platelet aggregation
Having established 11-HETE and 15(S)-HETE as platelet COX-1 products, the effects of these on platelet aggregation were tested alongside the other COX-1 products at the proportions identified by LC-MS/MS analysis (i.e., these experiments were aimed to clarify the net effect of the combination of COX-1-dependent products upon platelet aggregation). Aggregation induced by collagen (3 mg/ml) in the presence of vehicle (light transmission aggregometry; AUC, 503 6 118) was greatly reduced by aspirin (AUC, 92 6 18) and largely restored by the addition of U46619 (2 mM) as a surrogate for TXA 2 (AUC, 352 6 86). This effect was blunted by further addition of PGE 2 (1 mM) and PGD 2 (1 mM) (TXA 2 /PGE 2 /PGD 2 , 2:1:1; AUC, 176 6 42) but was unaffected by further addition of the HETE enantiomers
Effect of incubates of platelets activated by collagen or TRAP-6 in the presence or absence of aspirin on HMEC-1 tube formation Incubates of platelets stimulated with TRAP-6 (30 mM) or collagen (30 mg/ml) robustly promoted tube formation of HMEC-1 cells (Fig. 4), but this effect was absent when platelets were pretreated with aspirin (Fig. 4).
Effect of 15(S)-HETE on HMEC-1 tube formation in the presence of releasates from aspirin-or vehicle-treated platelets stimulated with TRAP-6 or collagen Releasates from platelets stimulated with TRAP-6 (30 mM) or collagen (30 mg/ml) robustly promoted tube formation of HMEC-1 cells (Fig. 5). This effect was strongly impaired in releasates from aspirin-treated platelets but was fully restored by the addition of 15(S)-HETE (1 mM). 15(S)-HETE (1 mM) did not increase the proangiogenic effects of releasates from non-aspirintreated platelets.
Effect of 15(S)-HETE on formation of sprouts from rat aortic rings in the presence of releasates from aspirin-or vehicle-treated platelets stimulated with TRAP-6 or collagen Incubation of rat aortic rings with releasates from platelets stimulated with collagen or TRAP-6 led to robust increases Whole blood was incubated under static conditions for 30 min at 37°C in the presence of vehicle control, collagen (30 mg/ml), TRAP-6 (30 mM), or Ca 2+ ionophore (A23187; 50 mM) and then centrifuged (1300 g; 24°C) to produce plasma samples. Data are shown as means 6 SEM. Data are ordered by amounts of eicosanoids (pg/ml) produced in response to collagen (n = 4 for each).
in the number of sprouts comparable to those caused by VEGF, a well-characterized proangiogenic factor (Fig. 6). However, the number of sprouts was strongly reduced in aortic rings incubated with releasates from platelets that had been treated with aspirin. The loss of sproutstimulating activity was restored by the addition of 15(S)-HETE (1 mM) to the aspirin-treated releasates.

DISCUSSION
Although the use of low-dose aspirin for the secondary prevention of vascular thrombotic events has been established for many years (8), it is only recently that epidemiologic and observational analyses have revealed that long-term use of aspirin at low doses confers protection against particular types of cancers, notably colorectal, esophageal, stomach, and gastrointestinal (1-4). As outlined above, aspirin at low dose exerts its strongest effect on platelets, with higher doses being required for systemic analgesic and anti-inflammatory effects. Although the mechanism of action of aspirin in reducing cancer progression is not clear, the rational conclusion to be drawn from available epidemiologic and pharmacokinetic data is that much of this beneficial effect is due to inhibition of platelet COX-1. Therefore, we hypothesized that exposure of platelets to aspirin limited the production of procancerous and/or proangiogenic factors associated with COX-1. To test this concept, we conducted an unbiased LC-MS/MS analysis of the products of blood and plateletrich plasma exposed to known platelet activators in control conditions in the presence of aspirin and in the presence of the platelet inhibitor PAM, which does not inhibit COX-1. We conducted incubations under stirring conditions because this better predicts platelet reactivity in the body than static conditions (17). As others and we have reported previously (9,18,19), activated platelets produce a broad range of AA-derived eicosanoids. Notably, in our studies we found 11-HETE and 15-HETE to be major products formed at similar levels to TXA 2 through the action of group IV A cytosolic phospholipase A 2 and COX-1. Despite not having been previously reported as platelet COX-1 products, this finding is consistent with an earlier report that isolated COX-1 enzyme can directly produce 11(R)-HETE, 15(R)-HETE, and 15(S)-HETE when AA is present within the active site of the enzyme in different catalytically competent arrangements (20). We hypothesize that this effect may be seen in strongly activated platelets where very high levels of AA can be released. Because our LC-MS/MS analyses could not discriminate 15(R)-HETE from 15(S)-HETE, we used a discriminatory immunoassay, which demonstrated 15(S)-HETE as a major product.
In our studies we also noted that in blood, but not PRP, the formation of 5-HETE increased robustly in response to  Ca 2+ ionophore (A23187) but not to the selective platelet activators collagen or TRAP-6. This finding is consistent with 5-HETE being a leukocyte product dependent upon the activity of the 5-LOX enzyme and provides a useful control for the selectivity and relevance of our assays.
Notably, and consistent with the discriminatory abilities of our assays, the levels of 12-HETE in both blood and PRP were not affected by the presence of aspirin or prasugrel alone but were reduced when the drugs were used in combination. This result is consistent with dual antiplatelet therapy providing a general inhibition of platelet activation with the consequent reduction of AA release from the membrane phospholipids and thus decreasing the levels of 12-HETE. The lack of effect of aspirin against the production of 12-HETE provides further evidence for the drug acting selectively on COX-1 in our assays. Likewise, the lack of effect of PAM against the production of 11-HETE and 15-HETE supports the conclusion that these are COX-1-dependent products (i.e., the inhibition of their production caused by aspirin is not explained by a reduction in platelet reactivity).
Having established that 11-HETE and 15(S)-HETE are major COX-1 and aspirin-sensitive products of platelets, we next sought to determine their functional roles. We reasoned that the prime cellular targets of these hormones were within the vasculature; thus, we focused on platelets, leukocytes, and vascular endothelial cells. We studied the effects of exogenous 11(R)-HETE, 15(S)-HETE, and 15(R)-HETE alongside the other COX-1 products we had identified on platelet aggregation in vitro using the proportions found by our LC-MS/MS analysis and concentrations previously reported as having relevant biologic activity (12,13). Particularly, we tested the TXA 2 -mimetic (U46619), PGE 2 , and PGD 2 together with 11(R)-HETE, 15(S)-HETE, and 15(R)-HETE on platelets treated with aspirin to block the effects of endogenous COX-1 products before stimulation with aspirin-sensitive concentrations of collagen. We found that U46619 restored aggregation, which was then partially inhibited by further addition of PGE 2 and PGD 2 . This confirmed the already well-known strong proaggregatory effects of TXA 2 , and the inhibitory effects of PGD 2 and high concentrations of PGE 2 , on platelet aggregation. Addition of 11(R)-HETE, 15(R)-HETE, and 15(S)-HETE in proportions matching those detected by LC-MS/ MS did not alter these aggregation responses. We noted in these studies that the levels of TXB 2 detected by LC-MS/ MS was of the order of 10-fold lower than the concentrations of U46619 required to produce platelet activation (0.1 vs. 2 mM). This is consistent with previous studies and may be explained by the differences in sampling from the entire incubate volume compared with the amounts of mediators present in the small spaces between interacting platelets. Therefore, we focused upon the relative proportions of the mediators to be added to the platelets. Overall, these studies characterized the individual and net effects of platelet COX-1 products on platelet activity and demonstrated that these are not responses regulated by 11(R)-HETE, 15(R)-HETE, or 15(S)-HETE when considered in a proportionate manner to COX-1 products known to be active, particularly TXA 2 , PGD 2 , and PGE 2 .
Having established a lack of effect of 11(R)-HETE, 15(R)-HETE, and 15(S)-HETE on acute platelet responses, we next assessed their effects on inflammatory cells because HETEs are known to modulate neutrophil trafficking and recruitment (21). We found that although 5-HETE and 12(S)-HETE produced strong chemoattractant effects on human neutrophils, as did the potent neutrophilderived chemoattractant LTB 4 , 11(R)-HETE, 15(R)-HETE, and 15(S)-HETE were without effect.
Based on the evidence that 15(S)-HETE promotes angiogenic responses in both human dermal microvascular endothelial cells and human umbilical vein endothelial cells by up-regulating VEGF through the PIK3-Akt and p38 MAPK signaling pathways, we next assessed whether this enantiomer could be involved in angiogenic responses using both in vitro and ex vivo models of angiogenesis (22)(23)(24)(25). Etulain et al. (26) have previously reported an aspirin-sensitive angiogenic effect of platelet releasates upon HMEC-1 cells. Using the same cell line, we similarly found that incubates of activated platelets strongly promoted the formation of angiogenic tubules and that this was significantly reduced when platelets had been pretreated with aspirin. To explore these interactions further, we centrifuged the activated platelet preparations before addition to the HMEC-1 cells to focus particularly on the platelet releasates produced during platelet activation and not on other products or platelet components that could interact with the cells over the much longer periods of cell culture required to study indicators of angiogenesis. With regard to the platelet incubates, releasates obtained from agonist-stimulated platelets strongly promoted the formation of angiogenic tubules by HMEC-1 cells in an aspirinsensitive manner, indicating that activated platelets acutely produce a proangiogenic factor through the activity of COX-1. We could then fully restore the response by the addition of exogenous 15(S)-HETE at the concentrations tested above, suggesting that 15(S)-HETE was the COX-1-dependent proangiogenic product. A similar outcome was seen in HMEC-1 migration. In direct accord with these results using cultured cells, our ex vivo studies clearly demonstrated that sprout formation from rat aortic rings was promoted by releasates from stimulated platelets but not by releasates prepared from platelets treated with aspirin. Addition of exogenous 15(S)-HETE rescued this process from aspirin inhibition.
In summary, our results suggest that 15(S)-HETE, formed as a direct COX-1 product, in combination with other proangiogenic factors released by activated platelets (e.g., VEGF and PDGF), is central to regulating endothelial angiogenesis, particularly the formation of tubule structures that represent the core of new vessels. It has been established for many years that inhibition of platelet COX-1 and consequent reduction in TXA 2 production explains the ability of low-dose aspirin to reduce thrombotic events. We suggest that the inhibition of 15(S)-HETE production by platelet COX-1 reducing angiogenesis could provide a similarly clear explanation for some of the cancer protective effects of low-dose aspirin.