Sterile inflammation of endothelial cell-derived apoptotic bodies is mediated by interleukin-1α

Induction of HUVEC Apoptosis.

To re-create in vitro the inflammatory events of prothrombotic and hypoxic conditions on endothelial cells in vivo, we cultured HUVEC using low serum and growth factor concentrations combined with stimulation by TNFα. When cultured in the absence of serum and growth factors, HUVEC undergo several morphological changes of swelling and detachment (Fig. 1A, a and b). These changes become more prominent with TNFα stimulation (Fig. 1 A, c). In addition, with DAPI labeling, it appears that starvation-induced nuclear modifications such as fragmentation, increased after TNFα stimulation (Fig. 1A, a′– c′). These morphological changes are consistent with apoptosis. To confirm this observation, we first used annexin V (AnnV) and propidium iodide (PI) labeling and observed that double-labeled cells were significantly more numerous in HUVEC subjected to starvation and TNFα stimulation (22 ± 2.3%) compared with cells grown in normal medium (10.3 ± 1.4%) or in starved conditions alone (14.5 ± 2.9%) (Fig. 1B). We next assessed the presence of the cleaved form of caspase 3 and observed that both the 17- to 19-kDa bands representing the large fragment of the cleaved caspase-3 were clearly present in TNFα-stimulated starved HUVEC, but not in cells grown in normal medium (Fig. 1C). Similarly, using a specific monoclonal antibody, the presence of the cleaved form of caspase-3 was confirmed by immunofluorescence microscopy in the cytoplasm of TNFα-stimulated starved HUVEC (Fig. 1D). These data are consistent with the concept that serum starvation combined with TNFα stimulation induces HUVEC apoptosis.

Supernatants of Apoptotic HUVEC Contain a Heterogeneous Particulate Population.

Differential centrifugation allowed us to define 2 different populations of particles in the supernatants of TNFα-stimulated starved HUVEC. After a 300 × g centrifugation to remove dead cells and debris, a 4,500 × g centrifugation was performed on the supernatant (Fig. S1). This step resulted in the identification of two particulate populations in the pellet by flow cytometry analysis, each of which was labeled with AnnV: one predominant population with a size <1 μm and a population of larger particles between 1–3 μm (Fig. 2A, Left). We then subjected the 4,500 × g supernatant to ultracentrifugation (75,000 × g) to purify membrane MP. In agreement with previous reports (7), after ultracentrifugation, only the smaller AnnV positive particles were present in the pellet, whereas the large particles were nearly absent (Fig. 2A, Right). When using both DAPI and AnnV labeling of the 4,500 × g and 75,000 × g pellets, we observed that only the population of large particles was labeled, whereas no DAPI labeling was found in the population of small particles, notably those present in the 75,000 × g ultracentrifugation pellet (Fig. 2B). Similar data were obtained when combining AnnV and 7-AAD labeling (Fig. S2). Thus, sequential 4,500 × g centrifugation followed by ultracentrifugation allowed the separation of a population of large (1–3 μm) AnnV⁺/DAPI⁺ or 7-AAD⁺ particles from the dominant population of small (<1 μm) AnnV⁺/DAPI⁻ or 7-AAD⁻ particles released by the apoptotic HUVEC.

To better characterize the particulate population, we asked whether histones can be detected in the large particles. As shown in Fig. 2C, Western blot analysis showed that these particles contained ample amounts of H3 and p-H2B histones, whereas these histones were weakly detected in the 75,000 × g fraction. Together, these data are consistent with the concept that the large particle AnnV⁺/DAPI⁺/histone⁺ population represents apoptotic body-like particles, therefore designed as apoptotic bodies (AptB), whereas the small AnnV⁺/DAPI⁻/histone⁻ particles found after ultracentrifugation represent membrane MP.

IL-1α Is Present in AptB but Not in MP.

Using an IL-1α immunoassay (Fig. 3A), we observed that IL-1α was significantly elevated in the 4,500 × g pellet from starved and TNFα-activated HUVEC compared with starved HUVEC. On the contrary, no significant concentration of IL-1α was detected in the 75,000 × g pellet from both starved and TNFα-activated HUVEC. Using flow cytometry analysis (Fig. 3B), we observed that IL-1α was detected in the AnnV⁺ 7-AAD⁺ AptB from the 4,500 × g pellet but not in the AnnV⁺ MP from the 75,000 × g pellet. To determine the molecular mass of IL-1α in AptB, we performed a Western blot analysis and observed that the 31-kDa precursor form, the 17-kDa mature form, and also an intermediate 23-kDa form of IL-1α were present in the 4,500 × g pellet (Fig. 3C). Very faint bands can be detected in 4,500 × g pellet from starved unstimulated HUVEC, whereas in agreement with the ELISA and cytometry data, no band was observed in the 75,000 × g pellet. All together, these data showed that IL-1α was present in AptB but not in MP. In addition, we submitted the 4,500 × g pellet to immunolabeling with anti–IL-1α mAb or control isotype followed by gold labeling and observed specific labeling with anti–IL-1α mAb but not with the control isotype using electron microscopy (Fig. 3D).

AptB but Not MP Induce Chemokine Production from HUVEC via an IL-1α–Mediated Mechanism.

Endothelial-derived AptB and MP were compared for their proinflammatory properties in vitro. As shown in Fig. 4, when added to monolayer cultures of HUVEC, AptB induced the secretion of IL-8 (Fig. 4A) and MCP-1 (Fig. 4B). AptB from starved HUVEC induced low levels of IL-8, whereas AptB from starved and TNFα-activated HUVEC induced significant IL-8 production, which was nearly completely inhibited by addition of IL-1Ra or anti-IL-1α mAb (Fig. 4A). However, anti–IL-1β did not affect IL-8 induced by AptB, whereas anti–IL-1β completely inhibited recombinant IL-1β–induced IL-8 production (Fig. S3). Similar data were obtained when measuring MCP-1 production from HUVEC in response to AptB (Fig. 4B). In parallel experiments, we studied the capacity of MP obtained from starved and TNFα-stimulated HUVEC to induce chemokine production. In contrast to AptB, MP did not induce significant production of chemokines (Fig. 4 C and D). These results identified IL-1α as the primary mediator of the proinflammatory effect of AptB generated from HUVEC.

i.p. Injection of AptB into Mice Induces Serum KC and a Local Neutrophilic infiltration.

To determine whether the chemokine-inducing property of AptB is active in vivo, we injected these particles into the peritoneal cavity of mice. Fifteen hours after the injection, there was an increase in the number of peritoneal cavity leukocytes (Fig. 5A). As shown in Fig. 5B, these leukocytes consisted mostly of neutrophils (2.1-fold increase). Moreover, injection of IL-1Ra a few minutes before AptB completely prevented the neutrophil influx (Fig. 5 A and B). In the same experiments, no significant increase of monocyte count was observed after the injection of AptB (Fig. S4). In agreement with these data, the concentration of KC (mouse form of human IL-8) was significantly increased in the sera of mice injected with the particles (threefold increase) but not in animals cotreated with IL-1Ra (Fig. 5C).

Reagents and Antibodies.

Earle's modified 199 medium, Dulbecco's PBS (PBS) without CaCl₂ and MgCl₂, donkey Alexa Fluor 594-conjugated anti-rabbit IgG antibody (Ab) were obtained from Invitrogen Life Technologies and EGM2 medium from Clonetics (Lonza). 4',6'-diamidino-2-phénylindole (DAPI), paraformaldehyde (PFA) and polymyxin B were purchased from Sigma Aldrich Chimie. ECL Plus Western blotting detection system kit was obtained from GE Healthcare Europe. Fluorescein isothiocyanate (FITC)-conjugated annexin V (AnnV-FITC), 7-AAD, PI, Phycoerythrin (PE)-conjugated goat anti-mouse IgG (H+L) Ab were purchased from Beckman Coulter. Recombinant human TNFα was obtained from Active Bioscience. Recombinant human IL-1α (200-LA), anti-human IL-1α mAb (MAB200, IgG2a, clone 4414.141), goat anti-human TNFα Ab (AF-210-NA) were purchased from R&D Systems Europe. Rabbit anti-human IL-1β polyclonal Ab and IL-1Ra were from Charles Dinarello. Rabbit anti-human cleaved caspase-3 mAb (Asp175) (5A1E) was obtained from Cell Signaling (Ozyme). Mouse irrelevant IgG2a, rat anti-mouse Ly6G-FITC (Gr1, RB6-8C5), rat anti-mouse CD11b-Allophycocyanin (APC) (M1/70) Ab were purchased from eBioscience. Rat anti-mouse CD45-PE Ab (30-F11) was from BD Bioscience. Rabbit control IgG, rabbit anti-human histone H3 (FL-136), goat anti-human IL-1α (C-18) Ab were purchased from Santa Cruz Biotechnology. Rabbit anti-phospho-human histone H2B (Ser-14) Ab was obtained from Millipore. Fifteen-nanometer beads gold-labeled Ab was obtained from BB International.

Eight- to 10-week-old male C57BL/6J mice were purchased from Charles River Laboratories.

Cells.

Human umbilical vein endothelial cells (HUVEC) were isolated as previously reported (6), grown in EGM2 and used between passages 2 and 4. When required, HUVEC were starved of serum and growth factors by culturing in M199 with 5% 0.22-μm filtered MP-free FCS (FCS) during 16 h before 50 ng/mL TNFα stimulation for 24 h to induce particles generation.

In some experiments HUVEC monolayers were incubated for 16 h with freshly obtained particles (5 particles per 1 HUVEC ratio) in the presence of 7 μg/mL polymyxin B and various inhibitors such as 200 ng/mL IL-1Ra, 1 μg/mL anti-IL-1α, 1 μg/mL anti–IL-1β, or 25 ng/mL anti-TNFα Ab, before supernatants were discarded, centrifuged, and assayed for chemokine concentrations.

Apoptosis Assays.

Nonadherent and trypsinized HUVEC were pooled then labeled with AnnV-FITC, and later PI and analyzed by flow cytometry. In other experiments, HUVEC were grown in eight-well culture slides, fixed with 1% PFA, then incubated with PBS containing 0.1% Triton X-100 for 1 h. Cells were incubated with rabbit anti-human cleaved caspase-3 mAb or with 4 μg/mL control rabbit IgG for 1 h at 4 °C, before the addition of donkey Alexa Fluor 594-conjugated 2 μg/mL anti-rabbit IgG for 45 min. Cells were then colabeled with DAPI and analyzed on TE2000-U epi-fluorescence microscope (Nikon Instruments Europe BV).

Generation, Numeration, and Characterization of Particulate Populations.

Endothelial particles were obtained from the supernatants of starved HUVECs, either unstimulated or stimulated with TNFα after a three-step centrifugation method, as described in Fig. S1. Particles were analyzed by flow cytometry using a Coulter FC500 with CXP software (Beckman Coulter) as reported (42). Two different gates were defined, the first one included particles between 1 and 3 μm whereas the second one included particles between 0.5 and 1 μm. Protocol setting and gates were adjusted by using 0.5-, 0.9-, and 3-μm Megamix beads (Biocytex). The presence of nucleic acid was determined by using 4,500 × g and 75,000 × g pellets colabeling with AnnV-FITC and DAPI or 7-AAD. To detect the presence of IL-1α, 10⁶ particles of either 4,500 × g and 75,000 × g pellets were incubated at 4 °C for 1 h with 8 μg/mL anti–IL-1α mAb or 8 μg/mL control IgG2a Ab in 0.1% saponin containing PBS then revealed by addition of goat PE-conjugated anti-mouse IgG. After washing, 4,500 × g and 75,000 × g pellets were costained with AnnV-FITC and 7-AAD.

Protein Extraction.

Trypsinized HUVEC, 4,500 × g and 75,000 × g pellets were lysed in a buffer containing 1% Triton X-100, 25mM Hepes at pH 7.5, 150 mM NaCl, 1 mM EDTA, 100 μM leupeptin, 10 μg/mL aprotinin, 10 μM pepstatin A, and 1 mM phenylmethylsulfonyl fluoride. Respective lysates were centrifuged at 20,000 × g for 15 min and the supernatants recovered. Protein concentrations were measured by micro BCA protein assay kit (Pierce, Perbio Science), whereas the remaining supernatants were stored at –80 °C until assayed IL-1α by Western blot (WB) or ELISA.

For WB, samples of respective lysates containing equal amounts of protein (30–50 μg) were separated by continuous SDS/PAGE on 12% or 15% slab gels then transferred to nitrocellulose membranes (Schleicher and Schuell). Blots were blocked with TBST (150 mM NaCl, 50 mM Tris⋅HCl at pH 7.5, 0.1% Tween 20) with 3% BSA for 30 min then probed with rabbit anti-human histone H3 (1:1,000), rabbit anti-human phospho-histone H2B (1:10,000), rabbit anti-human cleaved caspase-3 (1:1,000), or goat anti-human IL-1α (1:1,000) Ab. Membranes were then incubated with respective secondary antibodies conjugated to horseradish peroxidase and specific bands revealed by using ECL Plus WB detection system.

Electronic Microscopy.

Ten-microliters of the 4,500 × g pellet isolated from TNFα-stimulated starved HUVECs were applied to 100 mesh nickel formvar-carbon–coated grids (Agar Scientific) for 10 min. Samples were fixed with PBS containing 2% PFA and 0.2% glutaraldehyde, incubated with 5 μg/mL anti-human IL-1α mAb or irrelevant mouse IgG2a, then incubated with beads gold-labeled Ab for 1 h. Grids were negatively stained with 0.3% phosphotungstic acid before observation with JEOL 1220 electron microscope.

Cytokine Assays.

Measurements of IL-1α, IL-8, and MCP-1 were performed with specific ELISA (R&D Systems).

Animal Studies.

Animal studies were approved by the local ethical committee. C57BL/6J mice received i.p. injections of PBS or IL-1Ra (30mg/kg), then injections of either 30 × 10⁶ particles from the 4,500 × g pellet in PBS or PBS alone. Fifteen hours after the injections, animals were killed and peritoneal cavities were washed with 2 mL of PBS. Peritoneal fluid was harvested in EDTA tubes, and blood was collected by intracardiac puncture for serum collection. In each sample, the total number of leukocytes, neutrophils, or monocytes in the peritoneal exudate cells was determined by the count of CD45⁺, CD45⁺/Ly-6G⁺/CD11b⁻ or CD45⁺/Ly-6G⁻/CD11b⁺ cells, respectively, using a Gallios flow cytometer and Fluorescent beads Flowcount (Beckman Coulter).

Statistical Analyses.

Data represent the cumulative results of all experiments and are expressed as mean ± SE [or median [25th – 75th) for Fig. 4] for the indicated number of experiments (n). Statistical analysis was performed with Graph Pad Prism version 4 software (GraphPad Software). Significant differences were determined by using nonparametric Mann–Whitney test (or ANOVA test for Fig. 5C). P values <0.05 were considered statistically significant.