Platelet-derived exosomes induce endothelial cell apoptosis through peroxynitrite generation: experimental evidence for a novel mechanism of septic vascular dysfunction

The established endothelial cell line derived from rabbit aorta characterized by Venter and Buonassisi [19] was a gift from Jose Eduardo Krieger (Heart Institute, University of São Paulo School of Medicine, São Paulo, Brazil). Cells were maintained in Ham's F12 medium supplemented with 10% (v/v) heat-inactivated fetal bovine serum (Invitrogen Brasil Ltda, São Paulo, Brazil) and allowed to grow to about 80% confluence. For 24 hours before use, cells were kept with 1% serum-supplemented medium to cause phase arrest.

Blood samples (40 ml) were collected from 12 patients admitted to the intensive care unit of the Hospital Israelita Albert Einstein (São Paulo, Brazil), with early (24 hours) diagnosis of septic shock, as defined in accordance with the criteria of the American College of Chest Physicians and the Society of Critical Care Medicine [20]. Patients were not on any antiplatelet or anti-inflammatory drug. The study was approved by the Institutional Ethics Board. Clinical data about septic patients and control subjects are given in Table 1.

Blood was collected in centrifuge tubes containing 10.5 mM trisodium citrate and was processed immediately. Initial procedures were performed at room temperature (between 20–25°C) to avoid artifactual platelet activation. Cells, platelets, and large debris were pelleted by centrifugation at 3,000 g for 10 minutes. Phenylmethanesulfonyl fluoride (3 mM), aprotinin (1 g/ml), and pepstatin (1 g/ml) as protease inhibitors were added to the supernatant, which was then sequentially filtered through 1.0, 0.45, and 0.22 μm nylon filters to remove platelets, cellular fragments, and apoptotic bodies. The remaining cell-free plasma was collected over ice and ultracentrifuged at 100,000 g for 90 minutes at 4°C. The pellet, containing exosomes, was first washed with PBS containing 0.1 mM EDTA to avoid contamination with plasma proteins, and then resuspended in 250 μl of PBS. The total exosome mass obtained was 9.6 ± 3.9 mg protein per sample. In previous work we have shown that this exosome population displayed almost exclusively platelet markers [18].

Blood (40 to 50 ml) was collected from healthy volunteers who had not taken any medication known to interfere with platelet function within the previous 2 weeks. The blood was drawn into tubes containing acid citrate dextrose anti-coagulant (3.8 mM citric acid, 7.5 mM trisodium citrate, 125 mM dextrose, 1.8 ml anti-coagulant per 8.1 ml of whole blood). Platelet-rich plasma was first obtained by centrifugation at 800 g for 5 minutes at 20°C, and subsequently leukocytes were removed through a commercial filter system (Pall Corporation, East Hills, NY, USA). Plasma-free platelet suspensions were obtained by centrifugation of platelet-rich plasma at 800 g for 15 minutes at 20°C, and the resultant pellet was resuspended in 5 ml of Krebs-HEPES buffer (in mM: NaCl 99, KCl 4.7, MgSO₄ 1.2, KH₂PO₄ 1, CaCl₂ 1.9, NaHCO₃ 25, glucose 11.1, and sodium HEPES 20).

Plasma-free platelet suspensions were incubated with agonist or with saline control (154 mM NaCl in water) for 1 hour as indicated, and the reaction was slowed down by placing samples on ice. Samples were centrifuged (800 g for 15 minutes) to obtain the platelet pellet fraction. The supernatant was further centrifuged (17,500 g at 30 minutes) to obtain the microvesicle fraction, and the supernatant from that microvesicle fraction was filtered sequentially through 0.45 and 0.22 μm low-protein-binding nylon membranes. The filtered product was further centrifuged (100,000 g for 90 minutes) to obtain the exosome pellet. All pellets were resuspended in 250 μl of PBS. The total exosome mass obtained was 10.6 ± 4.5 mg of protein per sample.

Sepsis and septic shock can be viewed as a state of immuno-inflammatory imbalance in response to an infection. Different models have been validated to simulate sepsis under in vivo or in vitro conditions, such as exposure to LPS or TNF-α. LPS is a component of the bacterial cellular wall known to stimulate the innate immuno-inflammatory response through Toll-like receptors present in leukocytes, dendritic cells, and endothelial cells [21]. TNF-α is a cytokine released in the early phases of the septic response and is believed to have a central role in its initial steps, promoting the further release of other inflammatory and anti-inflammatory cytokines and altering the vascular wall, leading to increased endothelial stickiness and permeability [22]. It is also well known that part of the vascular dysfunction arising during the clinical course of septic shock is due to an enhanced production of nitric oxide (NO) [23]. We therefore decided to stimulate platelets with those agents to create a suitable model of platelet exosome generation, similar to those found in septic patients. Platelets were incubated for 1 hour at room temperature with 100 ng/ml LPS, or 40 ng/ml human TNF-α, or with the NO donor diethylamine-NONOate (0.5 μM). Platelets incubated with 250 μl of saline or with 5 IU/ml thrombin were used as controls.

To generate apoptotic bodies, which served as controls for phosphatidylserine-exposing particles, apoptosis was induced in rabbit endothelial cells by treatment with ultraviolet radiation [18, 24]. In brief, after cells reached about 80% confluence on Petri dishes, culture medium was replaced with PBS and cells were irradiated for 30 minutes with ultraviolet radiation with a TUV 15 W/G15 T8 lamp (Philips, The Netherlands). After irradiation, fresh medium was added and cells were cultured for a further 24 hours. Supernatant medium was collected and centrifuged successively at 1,200 g and 10,000 g to pellet cells and large debris and finally at 100,000 g to collect apoptotic bodies.

Measurements of the generation of reactive species were all performed in a FARCyte plate reader (Amersham Biotech, Buckinghamshire, UK). Exosomes were resuspended in 100 μl of Krebs-HEPES buffer at a constant 100 μg/ml concentration. Luminescent or fluorescent probes were added 15 minutes before measurements started, and samples were equilibrated while being protected from light.

The luminescent probes lucigenin and coelenterazine were first used to detect the generation of ROS. The concentration of lucigenin and coelenterazine used (5 μM each) minimized the generation of artifactual readings, as shown previously [25]. Reactions were started by adding NADPH (0.1 mM) for the lucigenin assay and NADPH (0.1 mM) plus L-arginine (1 μM) for coelenterazine. Luminescence signals were measured in solid white plates, with the integration time set to 1,000 ms, without attenuation; background was automatically subtracted from all measurements. To compare the generation of ROS in exosomes with that in whole platelets, lucigenin and coelenterazine assays were performed with 10⁸ platelets/ml and results were corrected to protein content. Luminescent counts are presented as relative luminescence units (RLU)/min per mg of protein.

To better characterize the generation of reactive species, 2',7'-dihydrodichlorofluorescein diacetate (DCHF; 10 mM) for ROS [25] and 4,5-diaminofluorescein diacetate (DAF; 10 mM) for RNS [26] were used. Measurements were performed in the presence of NADPH (0.1 mM) with or without L-arginine (1 μM) for DCHF, and in the presence of L-arginine for DAF.

Further studies to characterize the source or type of reactive species were performed in the presence of specific inhibitors or quenchers such as L-NMA (N^G-methyl-L-arginine acetate; 5 mM), L-NAME (N^ω-nitro-L-arginine methyl ester; 1 μM) and D-NAME (N^ω-nitro-D-arginine methyl ester; 1 μM), urate (1 μM), the membrane-permeable superoxide dismutase mimetic Mn(III) tetrakis (4-benzoic acid) porphyrin chloride (SOD mimetic; 10 μM; Oxys Research, Portland, OR, USA), and the specific NADPH oxidase inhibitory peptide gp91ds-tat (10 μM) [27].

For flow cytometry analysis, we used aliquots of exosome or apoptotic body suspensions with 200 μg of particle protein/ml. To identify specific epitopes, aliquots were incubated with fluorescein 5(6)-isothiocyanate (FITC) or R-phycoerythrin-conjugated antibodies directed to specific membrane antigens at 1 μg/ml final concentration (BD Biosciences, San Jose, CA, USA), namely CD9, CD63, and CD81 (molecules from the tetraspan co-activator family, which characterize exosomes) [4, 8], and with annexin V-FITC conjugate in a calcium-containing binding buffer. Binding of annexin V indicates the exposure of phosphatidylserine on the particle surface. In contrast to signaling exosomes, apoptotic bodies are known to expose large amounts of phosphatidylserine [24]. Samples were acquired in a FACScan flow cytometer and analyzed with CellQuest software (Becton Dickinson, San Jose, CA, USA). Non-specific signals were inhibited by the addition of normal species serum. Binding of specific antibodies was corrected with identical concentrations of control IgG antibodies. Thresholds were set to correct for nonspecific antibody binding or fluorescence.

Because exosomes are, on average, too small for cytometry analysis, we believe that our data correspond to aggregates formed after ultracentrifugation. For this reason we did not attempt to perform any specific quantification.

Pellets of exosomes obtained from platelets were fixed under 2.0% glutaraldehyde in 0.1 M sodium cacodylate for at least 2 hours and postfixed with 2% osmium tetroxide in 10.56% sucrose for 2 hours and finally incubated with 0.5% uranyl acetate and 10.56% sucrose overnight. Pellets were then dehydrated and embedded in Spurr resin. Ultrathin sections 70 to 80 nm thick were cut on an ultramicrotome (Leica Ultracut R, Leica Microsystems GmbH, Wetzlar, Germany), picked up on copper grids and stained for contrast with 1% uranyl acetate and 1% lead citrate. Specimens were examined with a transmission electron microscope (Jeol Electric 1010; Jeol Ltd, Tokyo, Japan), operated at 80 kV.

Annexin V was used to detect apoptosis [28]. In brief, rabbit endothelial cells were grown on six-well plates as described. For 24 hours before use, cells were kept with 1% serum to cause phase arrest. A volume of exosome suspension equivalent to 100 μg of protein was added to each well (final protein concentration per well 400 μg/ml) and left to incubate for 30 minutes. Some experiments were performed after incubation with the membrane-permeable SOD mimetic (10 μM), with urate (1 μM), or with L-NAME (1 μM). After incubation, cells were washed, fresh medium was added. After 1 hour, cells were washed with ice-cold PBS and removed from the plates with 1% trypsin, followed by a short centrifugation and resuspension in calcium-containing binding buffer at a 10⁶ cells/ml into Eppendorf vials. Annexin V-FITC was added to a final concentration of 100 ng/ml, and the cells were incubated in the dark for 10 minutes and then washed again with PBS. Propidium iodide (30 μl) was added before analysis. Cells were spread on clean slides, covered with glass coverslips, and immediately examined under fluorescence microscopy. From three high-power fields per sample, a minimum of 200 cells were counted. Cells were considered apoptotic when membrane-bound annexin-FITC fluorescence was positive and nuclear staining with propidium iodide (evidence of late apoptosis or necrosis) was negative. Results are expressed as apoptotic cells per 100 cells.

Rabbit endothelial cells were cultured on six-well plates to 80 to 90% confluence as described. Cells were kept in 1% serum for 24 hours before use. A volume of microparticle suspension equivalent to 100 μg of protein was added to each well (final protein concentration per well 400 μg/ml) and incubated for 30 minutes. Some experiments were performed after incubation with the membrane-permeable SOD mimetic (10 μM) or with L-NAME (1 μM). Exposure to TNF-α (50 ng/ml) was used as a positive control for caspase-3 activation. After incubation, plates were kept on ice. Cells were washed with ice-cold PBS and lysed with Nonidet lysis buffer containing Tris/HCl (20 mM, pH 7.4), NaCl (150 mM), Na₄P₂O₇ (10 mM), leupeptin (1 μg/ml), pepstatin (1 μg/ml), phenylmethylsulfonyl fluoride (3 mM), and Nonidet P40 (1% v/v), placed on ice for 10 minutes, and centrifuged at 10,000 g for 10 minutes. The activity of caspase-3 was measured at 405 nm with a Caspase-3 Colorimetric Detection Kit (Assay Designs, Ann Arbor, MI, USA) in accordance with the manufacturer's instructions.

Exosome protein (40 μg), leukocyte and endothelial cell lysate (used as a positive control) were subjected to separation by SDS-PAGE and transferred to nitrocellulose. Equal separation and transference of the samples were confirmed by Ponceau staining during the preparation of membranes. Membranes were incubated with antibodies directed to the NADPH oxidase cytochrome b₅₅₈ components p22^phox, Nox1, and Nox2 (gp91^phox) (1:1,000 dilution; Santa Cruz Biotechnology, Santa Cruz, CA, USA) or to inducible nitric oxide synthase (NOS), endothelial NOS or neuronal NOS (1:1,000 dilution; Chalbiochem, EMD Chemicals, San Diego, CA, USA) followed by horseradish peroxidase-conjugated secondary antibody (1:5,000 dilution; Santa Cruz Biotechnology) and developed with the Chemiluminescence-Phototope-HRP (horseradish peroxidase)-conjugated Detection Kit (New England Biolabs, Beverly, MA, USA) as specified. Results are representative of at least three similar experiments.

Exosomes are known to expose several different markers related to their cellular origin and putative functions. Phosphatidylserine is typically not exposed, differentiating exosomes from apoptotic bodies or cellular debris. In contrast, proteins of the tetraspan family are considered to be specifically sorted during exosome generation. As shown in Figure 1, flow cytometry analysis clearly divided the exosomes in two groups: those obtained from platelets stimulated with either the NO donor diethylamine-NONOate or LPS, and those recovered from platelets exposed to saline (control), thrombin, or TNF-α (not shown).

Exosomes in the former group, which are similar to those recovered from septic patients, exposed large amounts of the tetraspan family members CD9, CD63, and CD81 and exhibiting low binding of annexin V. Exosomes in the latter group were similar to the apoptotic bodies with lower tetraspan exposure and higher annexin V binding capability.

Electron microscopy (Figure 2) revealed typical saucer-like structures with a diameter of 100 to 200 nm, which correspond to exosomes. Figure 2a shows exosomes derived from platelets exposed to diethylamine-NONOate, and Figure 2b exosomes from platelets exposed to thrombin.

Preliminary measurements of ROS-generating activity, performed with lucigenin, revealed that the redox activity of exosomes paralleled the surface characteristics disclosed by flow cytometry analysis. As seen in Figure 3, exosomes obtained from platelets exposed to LPS or to the NO donor generated ROS in a similar manner to that of exosomes from septic patients, whereas exosomes obtained from platelets exposed to saline (control), thrombin, or TNF-α (not shown) generated very small amounts of ROS. Intact platelets generated substantially higher luminescent signals than exosomes. Platelets from septic patients also displayed higher ROS generation than controls. The SOD mimetic and L-NAME had a similar inhibitory effect on whole-platelet ROS generation.

To characterize the exosome redox profile better, measurements with the luminescent probe coelenterazine were also performed (Figure 4). Results were similar to those obtained with lucigenin. Furthermore, coelenterazine is known to react with both superoxide and peroxynitrite. The SOD mimetic and the NO synthase inhibitors L-NAME and L-NMA significantly inhibited the luminescent signals, suggesting that platelet exosomes are able to generate both superoxide and NO. Controls with D-NAME did not show any significant decrease in signal.

The fluorescent probes DCHF and DAF were used for further clarification of the nature of ROS generated by the exosomes. DCHF is believed to react with hydrogen peroxide, whereas detection of superoxide radical with DCHF is still not clear. DCFH can also be oxidized by peroxinitrite [25]. In contrast, DAF is considered a specific probe for RNS, such as NO or peroxynitrite.

Figure 5 shows clearly that exosomes obtained from septic patients and from platelets stimulated with the NO donor diethylamine-NONOate or LPS generate large amounts of ROS, whereas exosomes from non-stimulated platelets (control) or from platelets exposed to thrombin do not possess this activity. DCHF signals were inhibited by the SOD mimetic, suggesting that superoxide could be involved. In a previous study we showed that exosomes from septic patients possess a superoxide-generating activity that is not inhibited by catalase (a hydrogen peroxide scavenger), fluconazol (a cytochrome P450 inhibitor) or by oxypurinol (a xanthine oxidase inhibitor), but sensitive to phenylarsine oxide and diphenylene iodonium, two well-known inhibitors of NADPH oxidase. To characterize the source of superoxide better, we performed experiments with the specific NADPH oxidase inhibitor peptide gp91ds-tat [27], which greatly decreased the DCHF fluorescence of exosomes compared with the scrambled peptide. These results indicate the participation of a Nox-based NADPH oxidase. Recent studies suggest that uncoupling of the NO synthase could also be a significant source of superoxide in the vascular milieu [11]. In fact, the addition of L-NAME, known to block not only the NO generation but also superoxide generation from uncoupled NO synthases, caused a 40% inhibition of DCHF fluorescence. In addition, supplementation with L-arginine (Figure 6), which may favor recoupling of the NO synthase, resulted in a similar decrease in DCHF signals, suggesting that electron transfer was redirected to NO synthesis. Finally, considering the coexistence of active NADPH oxidase and NO synthase, we postulate a role for peroxynitrite as a major oxidating species in this system, because the addition of urate abolished these effects (Figure 6).

Figure 7 shows the results obtained with DAF. Exosomes from platelets exposed to the NO donor or to LPS had a similar activity profile to those from platelets obtained from septic patients, whereas exosomes from platelets exposed to thrombin or saline had low redox activity. Furthermore, DAF signals could be significantly decreased by the NO synthase inhibitor L-NAME and by the peroxynitrite scavenger urate, but not by the SOD mimetic. To investigate the source of NO, preliminary experiments were performed with addition of the dication chelator EDTA (1 mM) in probe buffer. Although no specific signal inhibition was noted, DAF signals became highly variable. Interference with intermediate reactions involved in signal generation was hypothesized. To clarify the situation, experiments were performed with calcium-containing or calcium-free Krebs-HEPES buffer. Under these conditions, DAF signals were not affected at all, indicating the existence of an active calcium-independent (inducible) NO synthase.

Figure 8 summarizes the results of a Western blot analysis of the exosomes. As expected from the functional results, we were able to identify the presence of type II NO synthase but not that of types I or III. Furthermore, we were able to identify the subunits p22^phox, Nox1, and Nox2 of the NADPH oxidase, as well as its regulatory protein protein disulfide isomerase (PDI). A non-specific protein staining can also be seen, which confirms equal gel loading between samples.

To verify a physiological or pathophysiological role for platelet-derived exosomes, we exposed cultured endothelial cells to the different types of exosome. As seen in Figure 9, exosomes obtained from platelets exposed to thrombin had no effect on basal endothelial cell apoptotic rates (baseline). In addition, exosomes from platelets exposed to saline did not show any effect on endothelial apoptosis rate (data not shown). In contrast, exosomes from septic patients and exosomes from platelets exposed to an NO donor showed a twofold to threefold increase in apoptotic rates. This effect was heat sensitive and was fully inhibited by the SOD mimetic, the NO synthase inhibitor, and urate. These results suggest, in fact, a role for the ROS and RNS generated by enzymatic sources in the exosomes.

Caspase-3 is one central step in the apoptotic cascade, and it is well known to be redox sensitive [29‐31]. To verify whether exosome-induced apoptosis could be related to caspase-3 activation, we exposed endothelial cells to various exosome preparations and measured caspase-3 activation colorimetrically. Figure 10 summarizes the results, revealing that exosome-triggered caspase-3 activation paralleled increased apoptosis rates in endothelial cells. In addition, we demonstrated that caspase-3 activation is clearly dependent on the generation of superoxide or NO. Exosomes obtained from platelets exposed to saline did not show any significant effect (data not shown).