Glutamate mediates platelet activation through the AMPA receptor

Platelets store glutamate, express glutamate transporters, and release glutamate during activation, but vascular targets of glutamate have not been characterized (13, 14, 16, 24, 25, 33). To define the role of platelet-derived glutamate in the vasculature, we first quantified platelet glutamate release locally within a developing thrombus using a glutamate-sensitive enzymatic probe. The probe provides continuous real-time glutamate concentration measurements by glutamate oxidase–catalyzed conversion of glutamate to α-ketoglutaric acid and hydrogen peroxide. Hydrogen peroxide is then ampomerically detected (34). We diluted whole blood in Tyrode's buffer (1:1), and glutamate concentration was measured during thrombin-initiated (0.5 U/ml) platelet aggregation. Glutamate increases rapidly during platelet aggregation to concentrations exceeding 400 μM (Fig. 1, A and B).

We next determined whether glutamate is a mediator of platelet activation. Platelet-rich plasma (PRP) was diluted 1:20 in Tyrode's buffer and incubated with glutamate (0–250 μM) for 5 min before activating platelets with a moderate concentration of thrombin receptor–activating peptide (TRAP-6; 5 μM) or a high concentration of TRAP (20 μM). Platelet activation was measured by FACS analysis for active conformation of surface GP IIb/IIIa using PAC-1 antibody (Fig. 2 A; or by surface P-selectin expression, Fig. S1 B, available). Glutamate does not increase resting platelet activation (Fig. 2 A), but in the presence of glutamate, 5 μM of TRAP significantly increases platelet activation (Fig. 2 A). Glutamate has no effect at high TRAP concentrations (Fig. 2 A). A dose–response curve with glutamate and moderate TRAP concentration (5 μM) was also performed, and PAC-1 binding was expressed as the percentage of increase in platelet mean fluorescence versus resting control platelets. Glutamate dose dependently increases platelet activation in response to TRAP (Fig. 2 B). As little as 150 μM of glutamate increases TRAP-induced platelet activation by >50% compared with control-treated platelets (Fig. 2 B). These are physiologically relevant plasma glutamate concentrations; control patient plasma glutamate is ∼100 μM at baseline, and in stroke patients it can rise to >200 μM on admission. Our data in Fig. 1 (A and B) demonstrate that local glutamate concentrations in platelet aggregation can exceed 400 μM (27, 35). Glutamate also dose dependently increases platelet activation in response to the thromboxane mimetic U46619 (Fig. 2 C).

Because glutamate appears to have a modulatory role in platelet activation, we determined its affects at concentrations that are most relevant, low concentrations of glutamate (150 μM), and low agonist concentrations (1 μM TRAP and U46619). Glutamate-mediated increase in platelet activation after low concentration thromboxane mimetic is approximately three times greater than control, and approximately two times greater in response to TRAP (Fig. 2 D).

We next sought to identify the glutamate receptor that augments platelet activation. Others have demonstrated the presence of the NMDA type glutamate receptor (NMDAR) on platelets, but this appears to have a different role in platelet function (20, 22, 23). To determine whether NMDAR mediates glutamate activity, we blocked the NMDAR with (±)-3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP) and treated platelets with glutamate before stimulating with TRAP. NMDAR inhibition does not inhibit glutamate-mediated increase in platelet activation (Fig. 3 A). We therefore explored whether platelets express another type of ionotropic glutamate receptor. We demonstrated the presence of AMPAR subunit proteins GluR1-4 by immunoblot (Fig. 3 B). We also localized expression of the AMPAR subunit GluR1 to the platelet surface using flow cytometry (FACS) with an antibody specific for an external region of GluR1 (Fig. 3 C). These data demonstrate that platelets express the AMPAR.

To define AMPAR's potential role as a mediator of platelet activation, we incubated PRPs with control or AMPA (0.1 – 1 mM) and activated the platelets with TRAP. Similar to glutamate, pretreatment of platelets with AMPA has no affect on resting platelets, but dose dependently increases agonist-induced platelet activation (Fig. 4 A). AMPA also increases platelet activation in response to U46619 (Fig. S1 C).

We next used additional pharmacological approaches to demonstrate signaling through the AMPAR and its potential as a target for antithrombotic therapy. First, we incubated human platelets with control, the AMPAR inhibitor CNQX (100 μM), AMPA, or both, and then we measured platelet activation in response to TRAP. CNQX is an AMPAR inhibitor with some kainate receptor blocking, but has a 20–150-fold selectivity for AMPAR subunits versus kainate receptor subunits. As before, AMPA increases TRAP-induced PAC-1 antibody binding (Fig. 4 B). CNQX inhibits TRAP-induced platelet activation and inhibits the effects of AMPA upon platelet stimulation (Fig. 4 B). Furthermore, as little as 10 mM of CNQX significantly decreases platelet activation, likely by inhibition of released endogenous glutamate activities, indicating an autocrine role of glutamate in platelet activation (Fig. 4 C and Fig. S1 D). Other AMPAR inhibitors, such as NBQX, have similar effects (Fig. S1 E).

The functional relevance of platelet AMPAR was explored by isolating platelets from WT and GluR1 subunit knockout mice (GluR1^−/−). Platelets from each type of mouse were incubated with vehicle or AMPA (1 mM) and activated or not activated with thrombin (0.25 U/ml). Unstimulated platelets from WT and GluR1^−/− mice have similar levels of activation, as measured by P-selectin localized to the platelet surface membrane (Fig. 4 D). However, WT platelets stimulated with thrombin have a greater increase in P-selectin externalization than GluR1^−/− platelets. Furthermore, AMPA increases activation of WT platelets, but not GluR1^−/− platelets (Fig. 4 D).

Collectively, these data demonstrate that AMPAR signaling increases platelet activation.

In the open state, AMPARs are permeable to Na⁺, and depending on its subunit isoform composition, can also be selectively permeable to Ca²⁺. To determine if platelet AMPAR increases Na⁺ influx into platelets, we preincubated platelets with the Na⁺-sensitive dye sodium-binding benzofuran isophthalate (SBFI) and added control, AMPA (1 mM), CNQX (100 μM), or both. Platelets were then activated or not activated with TRAP. AMPA increases intracellular Na⁺ concentration by approximately twofold upon activation, as compared with control (Fig. 5 A). CNQX not only decreases Na⁺ concentrations in platelets treated with both AMPA and TRAP, but CNQX also decreases Na⁺ levels in platelets treated with TRAP alone (Fig. 5 A). These data demonstrate that the AMPAR mediates an increase in intracellular Na⁺.

To confirm that the AMPAR mediates Na⁺ influx independent of the NHE, we inhibited the NHE with 100 μM amiloride (Fig. S1 F), and activated platelets or not with TRAP. Inhibition of the NHE reduces Na⁺ influx and platelet activation in response to TRAP (Fig. 5 B and Fig. S1 G). However, adding AMPA after amiloride restores Na⁺ influx and platelet activation, again suggesting that the AMPAR can act as a Na⁺ channel in platelet activation (Fig. 5 B).

To further characterize the signaling induced by platelet AMPAR activation, we performed electrophysiological recordings from bone marrow–isolated megakaryocytes. Whole-cell recordings showed that rapid application of AMPA (0.1 or 1 μM; n = 7; not depicted) or kainate (kainate; 0.1 or 1 mM; n = 3) did not induce a measurable current. In contrast, when the desensitization blocker cyclothiazide was included in the recording solution, a small but reproducible inward current was readily observed after rapid perfusion of AMPA (100 μM; 12.9 ± the SEM 7.1 pA; n = 9; Fig. 5 C) and kainate (100 μM; 4.8 ± SEM 1.4 pA; n = 7; Fig. 5 D). Whereas the inward current induced by kainate was of sustained amplitude throughout the duration of the application, that induced by AMPA was typically of slightly greater amplitude at the beginning of the application and subsided to a steady-state value (Fig. 5 C). To further demonstrate that the current induced by AMPA was mediated by activation of AMPARs, bone marrow megakaryocytes were isolated and electrophysiologic recordings performed in parallel using cells from a single megakaryocyte isolation. Administration of the AMPAR antagonist CNQX (30 μM) blocks the AMPA-induced current (n = 6; Fig. 5 E and Fig. S1 J).

We next sought to determine the reversal potential of the current induced by AMPA and kainate by determining the current–voltage relationship of the agonist-induced inward current. In the case of AMPA application, it was somewhat difficult to reliably and reproducibly match the voltage steps with the short-lived peak current obtained at the very beginning of the application (Fig. 5 C). As a consequence, I–V curves were constructed from the delayed, steady-state part of the AMPA-induced current. In these conditions, the current induced by both AMPA (n = 5) and kainate (n = 5) reversed at a potential close to 0 mV and did not show any obvious rectification (Fig. 5, C and D). Together, the results from these electrophysiological recordings support our previous finding in showing that megakaryocytes express functional AMPARs on their surface. In addition, the linear current–voltage relationship suggests that these receptors are GluR2 subunit–containing AMPARs.

In addition to conferring a linear I–V relationship to the AMPAR complex, the incorporation of GluR2 subunit renders the AMPAR impermeable to calcium. Joro spider toxin (JST) allows pharmacological discrimination between these subtypes of AMPARs by selectively inhibiting GluR2 Ca²⁺-permeable AMPARs (36). We next incubated platelets with 1 μM JST before AMPA. JST preincubation had no effect on AMPA-mediated increase in platelet activation (Fig. 5 F). Thus, these results suggest that platelet AMPARs are of the GluR2-containing, Ca²⁺-impermeable AMPAR subtype.

Collectively, these data suggest that the AMPAR modulates platelet activation in part by regulating Na⁺ influx.

Others have demonstrated an increase in GPCR ligand affinity and activity subsequent to membrane depolarization (6–8). Glutamate increases both thrombin and thromboxane receptor (both GPCR) agonist-induced platelet activation. To determine if it similarly increases platelet activation in response to other GPCR agonists, platelets were incubated with glutamate (250 and 500 μM) before activation with epinephrine, which is a GPCR ligand, or collagen, which is a non-GPCR ligand. Glutamate significantly increases epinephrine-induced platelet activation (Fig. 6 A), but has no effect on collagen-induced platelet activation (Fig. 6 B). As further proof, we activated platelets with ATP in the presence of glutamate. ATP stimulates platelet P2X₁, which is a purine-dependent, ligand-gated ion channel for Ca²⁺ (37). ATP-mediated platelet activation is not increased by glutamate (Fig. 6 C). Platelet P2Y receptors are GPCR purine-dependent receptors, and activation of platelets by the P2Y-specific agonist 2-Me-ADP is increased by glutamate (Fig. 6 D). These data support the idea that platelet AMPAR activation increases GPCR ligand signaling activity.

The functional relevance of targeting platelet AMPAR signaling was assayed by first measuring aggregation in response to TRAP (10 μM) after incubating platelets with control or CNQX (1–100 μM). CNQX dose dependently reduces platelet aggregation (Fig. 7, A and B). Exogenous AMPA (1 mM) was also added to platelets before aggregation. AMPA increases the rate of early aggregation. At 1 min after TRAP, control platelets are ∼55% aggregated and AMPA-pretreated platelets are ∼75% aggregated (Fig. 7 C). At later time points, the effect is less apparent (Fig. 7 C), which is likely a reflection of the large concentration of endogenous glutamate released into the growing platelet aggregate obscuring exogenous AMPA effects with time.

These data demonstrate that platelet AMPAR presents a potential novel in vivo therapeutic target. We therefore performed in vivo studies to support this idea. We treated mice i.v. with control PBS or CNQX (0.1 mg/kg), and 20 min later, tail bleeding times were determined. Control mice have a mean bleeding time of ∼200 s, with a cluster of mice that have very similar bleeding times (Fig. 8 A). AMPAR inhibitor–treated mice have a prolonged mean bleeding time (∼600 s) and a greater range of bleeding times (Fig. 8 A).

To confirm these data, we used intravital microscopy to examine the role of platelet AMPAR signaling in thrombus formation. Platelets were isolated from control mice, fluorescently labeled, and injected i.v. into recipient mice. Mice were then treated with CNQX or control PBS intravenously 20 min before mesentery externalization and FeCl₃ application. Images were collected to quantify in vivo platelet function (Fig. 8 B). Compared with PBS-treated mice, mice treated with CNQX have prolonged time to platelet adhesion (>10 platelets adherent for >10 s), thrombus formation (first stable thrombus >20 μm), and total vessel occlusion (Fig. 8 C).

To determine if mice lacking the AMPAR subunit GluR1 have in vivo changes in platelet function, isolated platelets from WT control or GluR1^−/− mice were fluorescently labeled and injected into mice of the same genotype. Mesenteric arterioles were then damaged with FeCl₃, and thrombus formation was again imaged. Similar to mice treated with the AMPAR inhibitor, mice lacking GluR1 have prolonged time to vessel occlusion compared with WT control mice (Fig. 8 D).

The major findings of our study are that AMPARs modulate platelet activation and thrombosis. These results have important implications for thrombotic diseases such as stroke, where plasma glutamate levels can rise significantly (25, 27). After compromise of the blood–brain barrier, glutamate released by neurons or platelets may contribute to the pathophysiology of stroke, not only as an excitatory amino acid but also by acting as a prothrombotic messenger. The importance of these results is not limited to cerebral vascular disease. We show that large concentrations of glutamate are released locally in platelet aggregation, making glutamate signaling important in disease processes such as myocardial infarction and ischemia-reperfusion injury. AMPAR antagonists may therefore be useful in the treatment and prevention of stroke, myocardial infarction, and other thrombotic diseases.

We propose that glutamate contained within platelet-dense granules is released in high concentrations locally into “the platelet synapse” as platelets come into close apposition during thrombus formation. This glutamate released into the formed platelet synapse early in platelet activation acts in an autocrine manner to promote complete platelet activation and in a paracrine fashion to rapidly increase the agonist sensitivity of recruited platelets. This helps drive and accelerate further thrombus formation. This model is supported by our data demonstrating that glutamate increases platelet sensitivity to low-dose agonist stimulation, that AMPAR antagonists inhibit platelet function, and that mice lacking the AMPAR subunit GluR1 have prolonged time to vessel occlusion. Our data does not rule out potential contributions of AMPARs expressed by other vascular cells, or the potential contribution of other platelet-expressed glutamate receptors, such as kainate receptors or metabotropic glutamate receptors. Taken as a whole, however, our ex vivo and in vivo data demonstrate an important role for platelet glutamate and AMPAR in platelet activation and thrombus formation.

Glutamate function in thrombosis is very similar to that of another neurotransmitter, serotonin (5-HT). At basal conditions, plasma 5-HT is relatively low and tightly regulated, but is rapidly released from dense granules upon platelet activation. Serotonin promotes continued platelet aggregation, as indicated by mice lacking the enzyme necessary for peripheral 5-HT production having prolonged time to vessel occlusion (38, 39). Similar to glutamate, 5-HT is a weak platelet agonist on its own and its effect is enhanced by the addition of other agonists (40). Serotonin reuptake inhibitors may provide myocardial infarction protection, but have not been extensively studied or used clinically as a preventative therapy (39).

We demonstrate that glutamate mediates an increase in platelet activation in response to the GPCR agonists thrombin, thromboxane, epinephrine, and ADP. However, glutamate has no effect on non-GPCR signaling pathways, such as through the collagen or P2X receptors. We also show that AMPAR signaling increases platelet Na⁺ influx and induces an inward current that depolarizes megakaryocyte membranes. Membrane depolarization can lead to an increase in GPCR ligand affinity by creating a charge movement that increases G protein coupling with the GPCR (8, 37). Platelet G protein receptors contain positively charged internal amino acid regions that are possible voltage-sensing domains, providing further support to our hypothesis. Alternatively, independent of a potential depolarization function, Na⁺ is an important ion in platelet activation. A rapid increase in intracellular Na⁺ is necessary to drive the Na⁺/Ca²⁺ exchanger and promote calcium-mediated events in platelet activation. The primary function of platelet AMPAR may be simply as a rapid sodium channel. Future studies are needed to determine the key regions of platelet GPCRs that AMPAR signaling and membrane depolarization may act upon to increase platelet activation.

Our electrophysiological studies support our data by showing that megakaryocytes express functional, most likely GluR2-containing, Ca²⁺-impermeable AMPARs. Notably, the inward current induced by AMPA application in megakaryocytes was of modest amplitude and only observed in the presence of a desensitization blocker, cyclothiazide. This contrasts with neurons where robust AMPAR-mediated inward currents are readily observed in the absence of cyclothiazide. Microglial cells, however, are similar to megakaryocytes in that AMPAR-mediated inward currents are only observed in the presence of cyclothiazide (41). Several mechanisms may account for these cell type–specific biophysical properties of AMPARs. For instance, transmembrane AMPAR-regulating proteins are important modulators of AMPAR functions, and their differential expression among cell types is expected to translate into different functional properties of AMPARs (35, 42). Future studies will be required to directly address this possibility.

There have been few reports of the NMDA receptor on platelets and megakaryocytes (20–23). Signaling through the NMDA receptor has been demonstrated to increase platelet production from megakaryocytes (20, 23), and interestingly, NMDA receptor signaling in platelets has been shown to inhibit platelet aggregation (21, 22). We now show that glutamate-mediated increase in platelet activation is independent of the NMDA receptor (Fig. 3 A). Prior NMDA receptor studies primarily used aggregation as the measure of platelet function, but endogenous platelet glutamate must be taken into account. We demonstrate that a substantial concentration of glutamate is released in the developing thrombus that reaches peak levels very rapidly (Fig. 1), and in our studies, exogenous glutamate or AMPA addition has no effect on final aggregation, likely because the rapid release of endogenous glutamate obscures any effect. However, the NMDAR studies do raise important questions for future research. NMDAR signaling in neurons is a major means to regulate AMPAR trafficking by inducing an influx of Ca²⁺ that initiates the delivery of AMPARs from the recycling endosome to the surface membrane. Prior NMDAR studies in platelets demonstrate that NMDA induces an increase in Ca²⁺ influx, and as a result may mediate an increase in AMPAR surface localization (21, 22).

Other glutamate regulatory elements in platelets, such as the glutamate transporters (EAATs), have primarily been studied as peripheral markers of CNS disease. However, EAATs and other glutamate receptors may also have an important role in normal platelet function by controlling the plasma glutamate concentration. Individuals with polymorphisms in EAAT-2 have increased plasma glutamate concentrations with a higher frequency of early neurological worsening in stroke (27). It is interesting to consider that these individuals may also be at increased risk of stroke, or that increased neurological decline may be secondary to continued platelet activation and increased infarct size. Further study is needed to fully appreciate glutamate in the periphery and its relevance to thrombosis.

L-glutamate, AMPA, CPP, ADP, CNQX:HBC, mouse thrombin, choline chloride, amiloride, U46619, and JST were purchased from Sigma-Aldrich. Antibody to GluR1 was purchased from Calbiochem; antibodies to GluR2, GluR3, and GluR4 were purchased from Santa Cruz Biotechnology. PAC-1 and P-Selectin antibodies were purchased from BD Biosciences. FITC-IgG secondary antibody for FACS, SBFI, and BCEF were purchased from Invitrogen. RT-PCR reagents were purchased from Invitrogen. TRAP-6 was purchased from Bachem.

Human platelets were isolated from healthy volunteers who had not taken aspirin or nonsteroidal antiinflammatory drugs for 10 d, under a protocol approved by the Johns Hopkins University School of Medicine Institutional Review Board. Blood was collected into citrate anticoagulant, and platelets were isolated as PRPs by centrifugation at 180 g for 15 min and diluted in Tyrode's buffer at a dilution of 1:20 for activation flow cytometry studies. To determine surface GluR1, a Fc blocking antibody was added to reduce nonspecific background antibody binding. Mouse platelets were isolated by collection into heparinized Tyrode's buffer and isolated by centrifugation. Washed platelets were resuspended in Tyrode's.

Intracellular ion concentrations were performed after incubating platelets diluted 1:20 in Tyrode's buffer with SBFI for 45 min before experiments.

Whole blood was isolated as described in the previous section and diluted 1:1 in Tyrode's buffer. A 1-mM glutamate-sensitive probe (Pinnacle Technology, Inc.) was used to determine real-time glutamate concentrations. The glutamate biosensor relies on the glutamate oxidase–catalyzed conversion of glutamate to α-ketoglutaric acid and hydrogen peroxide. The enzymatically produced hydrogen peroxide is then ampomerically detected by its oxidation at the probe's platinum-iridium electrode with an applied potential of 600 mV, with high sensitivity to glutamate at low (micromolar) concentrations, producing a linear response over a wide concentration range. The rapid response (1–4 s) is recorded using probe-specific software to monitor the time course of a physiological glutamate release. Probe calibration was performed by incubating the probe in a gradient of glutamate concentrations. Voltage output and glutamate concentration were then correlated. While stirring, the biosensor probe was placed in stirring blood and voltage recordings were initiated. Thrombin was added to induce platelet activation and aggregation.

Platelets purchased from HemaCare and rat brain cerebellar lysate were lysed with NP-40 lysis buffer, and supernatants were fractionated on a 4–15% gel. Membrane-transferred proteins were immunoblotted with antibodies against GluR1-4.

Mouse experiments were performed as approved by the Johns Hopkins University School of Medicine Animal Care and Use Committee.

6-wk-old male mice were anesthetized with ketamine and xylazine (80/13 mg/kg) and injected i.v. with either PBS or CNQX:HBC, with 0.1 mg/kg of active CNQX. 20 min later, the distal 3 mm of the tail was amputated and immersed in 37°C saline, and the time to visual cessation of bleeding was recorded.

Platelets were isolated from mice as above and resuspended in Tyrode's buffer at a concentration of 10⁸/100 μl, fluorescently labeled with 10 μM calcein-AM, and 100 μl was injected intravenously into a mouse anesthetized with ketamine and xylazine. The mesentery was externalized, thrombosis was initiated by the addition of a 5-mm² piece of Whatmann's paper soaked in 10% FeCl3, and thrombosis was recorded using a digital imaging camera and software (QCapture Pro; Retiga). GluR1^−/− mice, on a C57Bl6/J background, were a gift from R. Sprengel (University of Heidelberg, Germany).

All procedures using animals were approved by the University of Miyazaki Institutional Animal Care and Use Committee. Megakaryocyte preparation was performed as previously reported (43). In brief, mouse bone marrow was flushed with Na⁺-rich external solution (Na⁺-ES, 150 mM NaCl, 1 mM CaCl₂, and 10 mM Hepes, pH adjusted to 7.2 using NMDG-OH). Cells were dispersed by repetitive pipetting, washed three times by gentle centrifugation, resuspended in Na⁺-ES, and used within 90 min of isolation.

Whole-cell recordings were obtained from freshly isolated megakaryocytes using an intracellular solution of the following composition: 150 mM CsCl, 1 mM EGTA, and 10 mM Hepes. The pH was adjusted to 7.3 using NMDG-OH. The patch-clamp procedures were similar to those previously described (43). In brief, the experiments were performed in the tight-seal whole-cell configuration at room temperature, and electrical signals were acquired with an EPC-7 (HEKA). The pipette resistances ranged between 2 and 3 MΩ.

Current–voltage relationships (I–V curves) were obtained by clamping the cell to different potentials in 30-mV increments (−30, 0, and + 30 mV; each for 35 ms) from a holding potential of −60 mV. I–V curves obtained before the application of the agonists were subtracted from those (typically two) obtained in the presence of the agonists. The current responses were monitored on a chart recorder (WR7700; Graphtec) through a low-pass filter (200 Hz). Chemicals and reagents used were purchased from either Sigma-Aldrich or Wako Pure Chemicals.

Rapid perfusion of AMPA and Kainate was achieved by puff application from a nearby pipette of 30–35 mm diam to an isolated single megakaryocyte. The perfusion pipette was located within 30 mm of the cell, and the puff pressure was adjusted to obtain rapid and effective agonist application (< 10 ms).

Fig. 5 E was performed using two bathing chambers (test and control) placed on the stage of an inverted microscope. One chamber contained a megakaryocytes immersed in Na⁺-ES–containing CNQX (test chamber), and the other contained the same batch of megakaryocytes immersed in Na⁺-ES alone (control chamber). Traces used in the I–V curve were those which had good signal-to-noise ratio during the voltage steps.

Data are expressed as the mean ± the SD, unless otherwise stated. Statistical comparisons between two groups were performed using the Student's t test.

Fig. S1 shows glutamate probe calibration (A), data related to glutamate stimulation of platelet activation (B and C), AMPAR antagonist data (D and E), amiloride blocking of NHE (F and G), and inward current tracings (H–J). The online version of this article is available.

This work was supported by a National Institutes of Health (NIH) grant to C.N. Morrell (5K08HL074945) and grants to C.J. Lowenstein from NIH (P01 HL56091, R01 HL074061, R01 HL78635, and P01 HL65608), the American Heart Association (EIG 0140210N), the Ciccarone Center, the John and Cora H. Davis Foundation, and the Clarence P. Doodeman Professorship in Cardiology.

All authors have no conflicting financial interests.