Exosomes account for vesicle-mediated transcellular transport of activatable phospholipases and prostaglandins.

Exosomes are bioactive vesicles released from multivesicular bodies (MVB) by intact cells and participate in intercellular signaling. We investigated the presence of lipid-related proteins and bioactive lipids in RBL-2H3 exosomes. Besides a phospholipid scramblase and a fatty acid binding protein, the exosomes contained the whole set of phospholipases (A2, C, and D) together with interacting proteins such as aldolase A and Hsp 70. They also contained the phospholipase D (PLD) / phosphatidate phosphatase 1 (PAP1) pathway leading to the formation of diglycerides. RBL-2H3 exosomes also carried members of the three phospholipase A2 classes: the calcium-dependent cPLA2-IVA, the calcium-independent iPLA2-VIA, and the secreted sPLA2-IIA and V. Remarkably, almost all members of the Ras GTPase superfamily were present, and incubation of exosomes with GTPγS triggered activation of phospholipase A2 (PLA2)and PLD2. A large panel of free fatty acids, including arachidonic acid (AA) and derivatives such as prostaglandin E2 (PGE2) and 15-deoxy-Δ12,14-prostaglandinJ2 (15-d PGJ2), were detected. We observed that the exosomes were internalized by resting and activated RBL cells and that they accumulated in an endosomal compartment. Endosomal concentrations were in the micromolar range for prostaglandins; i.e., concentrations able to trigger prostaglandin-dependent biological responses. Therefore exosomes are carriers of GTP-activatable phospholipases and lipid mediators from cell to cell.

Exosome preparation. For exosome preparation 1.5 × 10 7 adherent cells were harvested with PBS-EDTA and added into 250 ml complete RPMI medium in a spinner bottle for cell culture in suspension. Culture volume was doubled every day in the spinner bottles to maintain a cell density of around 0.25 × 10 6 cells/ml for good cell viability until about 10 9 cells were produced overall. The cells were spun down, washed with DMEM medium, and concentrated to 10 8 cells in 10 ml of DMEM without FCS to avoid contamination by any microvesicles that might be present in the fetal calf serum. Exosomes were recovered following 20 min cell stimulation by ionomycin (1 µM fi nal) and purifi ed by differential centrifugations as reported previously ( 16 ). Correct exosome preparation required viable cells, which were checked by trypan blue exclusion. Briefl y, viable activated cells were eliminated by centrifugation at 300 g for 5 min. To get rid of possible cell debris, the supernatant underwent two consecutive centrifugations at 2000 g for 20 min at 4°C and 10,000 g for 30 min at 4°C. Exosomes were isolated from the 10,000 g supernatant by ultracentrifugation at 110,000 g for 70 min at 4°C. The pellet was resuspended in PBS and centrifuged again at 110,000 g for 70 min at 4°C. The fi nal pellet referred to as exosomes was resuspended in PBS for analysis. The quality of the preparations was checked by D 2 O/sucrose discontinuous gradient ( 1 ) and by electron tion, then as part of an essential process in the immune response, and recently as an enabler of the mechanism that modulates the translational activity of target cells by transferring selected micro RNA between cells ( 5 ). Whether exosomes participate dynamically in lipid metabolism is not known.
Exosomes appear to be involved in additional intercellular signaling beside soluble agonists. They interact with cell peripheral receptors, such as CD91 ( 6 ), a member of the LDL receptor-related proteins (LRP) receptors, and Tim4 ( 7 ), the phosphatidylserine receptor ( 8 ), a G protein coupled receptor (GPCR) member. Other nanovesicles similar to exosomes trigger the Notch signaling pathway ( 9 ).
The exosome biogenesis pathway can be "hijacked" by pathogens like the human immunodefi ciency virus (HIV), by proteins like prions involved in Creutzfeld-Jacob disease ( 10 ), and by the amyloid precursor protein (APP) of Alzheimer's disease ( 11,12 ).
Mast cell-derived exosomes trigger functional maturation of dendritic cells (DC) ( 6 ). The DC maturation process has been shown to involve lysophosphatidylcholine and secreted PLA 2 ( 13 ), and prostaglandins ( 14 ). We previously reported that exosomes from RBL-2H3 cells contain a high amount of lysophosphatidylcholine (LPC) ( 15 ) and that phospholipase D was involved in exosome release ( 16 ).
We undertook a large analysis of RBL-2H3 exosome content by proteomic high-throughput analysis together with immunodetection and determination of lipolytic activities. We showed that exosomes can behave as "signalosomes" not only by transporting GTP-activatable phospholipases D 2 (PLD 2 ) and phospholipase A 2 (PLA 2 ), but also by carrying the whole set of prostaglandins, including prostaglandin E2 (PGE 2 ) and the peroxysome proliferator activated receptor ␥ (PPAR ␥ ) agonist 15-deoxy-⌬ 12,14 -prostaglandin J 2 (15d-PGJ 2 ). We observed that exosomes could traffi c between resting or activated RBL-2H3 cells, thereby modulating RBL-2H3 cell activation by means of the lipid messengers they carry. In addition, exosomes could constitute a mechanism of entry for 15d-PGJ 2 , as the way this prostaglandin enters cells is as yet unknown (17)(18)(19).

Materials
For cell cultures, RPMI 1640, PBS, penicillin, streptomycin, L-glutamine, and FCS were purchased from Invitrogen. 4,4-Difl uoro-4-bora-3a,4a-diaza-s-incadene (BODIPY)-PC as phospholipase substrate and BODIPY-ceramide for exosome labeling and uptake detection by immunofl uorescence were obtained from Invitrogen Molecular Probes and stored in ethanol at Ϫ 20°C after dilution. GTP ␥ S was from Sigma. Methyl arachidonyl fl uorophosphonate (MAFP) was from Calbiochem. Bromoenol lactone (BEL, or HaloEnol Lactone Suicide Substrate) was from Biomol International. Pyrrolidine-1 and Me-indoxam were generous gifts from Prof. M. H. Gelb (University of Washington, Seattle, WA). The cPLA 2 monoclonal antibody (recognizing type IVA) and iPLA 2 polyclonal antibody (recognizing type VIA) were from Santa Cruz Biotechnology. Mouse sPLA 2-IIA and -V recombinant Plasma membrane labeling was fi rst performed on live RBLpld2 cells with fl uorescent cholera toxin added in PBS containing 10% BSA at 4°C for 30 min. The cells were washed with PBS, fi xed with 3% PFA for 20 min at 4°C, and permeabilized for 15 min at room temperature with 0.05% saponin in PBS-BSA. HA-PLD 2 location was then detected by incubation with anti-HA antibody diluted at 1/50, followed by incubation with a secondary antibody (45 min each antibody) at room temperature. Acquisition was performed with a Zeiss LSM 510 confocal microscope.
Measurement of PA phosphatase activity. Fluorescent phosphatidic acid (BODIPY-PA) was prepared from BODIPY-PC by in vitro hydrolysis with commercial phospholipase D. For enzymatic PAP activity measurement, 50 µg of exosomes were preincubated for 10 min at room temperature in a total volume of 500 µl of PBS with 2 mM Ca 2+ /Mg 2+ , with or without 100 µM GTP ␥ S, in the presence of 5 µl protease inhibitor cocktail. The reaction was started by addition of 1 µl BODIPY-PA (1 µM fi nal) supplied in ethanol and incubation proceeded at 37°C. Fluorescent lipids were extracted with 2 × 500 µl 1-n-butanol and resolved by HPLC.
HPLC analysis. HPLC separation and quantifi cation of BODIPY-PC-derived products were performed as already reported ( 26 ) on a silica diol column with a solvent fl ow rate of 0.4 ml/ min. Fluorescent standards of lysophosphatidylcholine (LPC), phosphatidic acid (PA), diglycerides (DG) and phosphatidylethanol (PEt) were prepared by in vitro incubations of BODIPY-PC with appropriate lipolytic enzymes.
Calibration curves for quantifi cation were plotted with BO-DIPY-PC as standard. HPLC peaks across chromatograms were identifi ed by fl uorescent standards injected in the middle of each series of samples to overcome variations in retention times.
Phospholipase immunodetection. For phospholipases A2, cells and exosomes were lysed in Laemmli sample buffer at 95°C for 10 min and sonicated. 10 mM EDTA was added for phospholipase D detection. 40 µg of proteins were run on 7.5% SDS-PAGE and transferred onto PVDF membrane. The membranes were saturated with 5% nonfat milk in TBS 0.1% Tween 20 for 1 h at room temperature and blotted at 4°C overnight with mouse or rabbit primary antibodies supplied in blotting buffer. Membranes were then washed and incubated in TBS 0.1% Tween 20 with HRP-labeled anti-mouse IgG or anti-rabbit IgG secondary antibodies for 1 h at room temperature.
For sPLA 2 detection, 40 µg of exosome proteins were separated on a 15% SDS-polyacrylamide gel, compared to 50 ng of microscopy (performed by D. Lankar, Institut Curie Paris; B. Payré, CMEAB, UPS Toulouse III, France). We also checked the size homogeneity of vesicles obtained using a Zetasizer Nano ZS90 (see below). Protein concentration was determined by the Lowry method ( 22 ) in the presence of 0.1% w/v SDS fi nal.
Size distribution and zeta potential analysis of RBL-2H3-derived exosomes. The Zetasizer Nano ZS 90 (Malvern Instruments, Orsay, France), allowed the analysis of particles with sizes ranging from 1 nm to 3 µm. Exosomes (50 µg from two pooled preparations) derived from RBLwt or RBLpld 2 cells were diluted in 1 ml PBS, and parameters such as zeta potential (electronegativity) and size distribution were analyzed at 37°C according to the manufacturer's instructions (see supplemental Fig. II).
Quantifi cation of exosome vesicles. The correlation between exosome protein content and the number of vesicles was established by FACS analysis on the basis of the method used to quantify the number of circulating microparticles ( 4 ). Exosomes were diluted in PBS-EDTA and the number of vesicles was taken as the number of events in the SSC/FSC quadrant.
Quantifi cation of exosome internalization. Exosomes were labeled with the fl uorescent lipid probe BODIPY-ceramide so that fl uorescence monitored the amount of vesicles directly ( 16 ). Fluorescent exosomes (25 µg proteins) were incubated with 10 6 adherent cells. At appropriate times, the excess of added exosomes removed, the cells washed, and cell-associated fl uorescence monitoring internalized exosomes were extracted with butanol and quantifi ed. The fl uorescence was converted into µg exosome protein using a calibration curve as previously reported ( 16 ).
Confocal microscopy. Internalization of fl uorescent exosomes was monitored under a Zeiss LSM 510 confocal microscope on live cells using LSM 510 software. Cells (3 × 10 4 in RPMI medium buffered with 25 mM Hepes) were seeded in LabTek chambers and kept overnight in an incubator. Then medium was removed, and 0.5 ml of the same fresh medium was added. The LabTek chambers were placed into a microscope chamber adaptor warmed to 37°C and with CO 2 fl ow. Exosomes (20 g), previously made fl uorescent by a 1 h incubation at 37°C with 1.2 M BODIPY-ceramide ( 23 ) and washed, were added in a small volume (20 l) into the cell medium and data acquisition started.
The compartment of exosome internalization in target cells was characterized by antibodies directed against late endosome markers. 2 × 10 5 cells were seeded on coverglass in 1 ml RPMI culture medium and incubated for 24 h with 75 µl anti-LBPA antibody (hybridoma supernatant) or 50 µl (10 µg) anti-CD63 antibody. Cells were washed with PBS, then overlaid with 0.5 ml culture medium, and 10 µg fl uorescent (BODIPY-ceramide labeled) exosomes were added. Incubation proceeded for 4 h at 37°C. Cells were washed with PBS and fi xed with 3.7% PFA for 20 min and washed again. The remaining PFA was quenched with 50 mM NH 4 Cl for 10 min. The cells were washed with PBS, then maintained for 30 min in PBS 3% BSA. Permeabilization was performed with 0.05% saponin in PBS 3% BSA for 10 min. The cells were washed and incubated 30 min with appropriate secondary antibodies (anti-mouse PE for LBPA and anti-goat FITC for CD63). Coverslips were mounted with Mowiol, and samples were examined under a LSM 510 confocal microscope.
To label the late endosome compartment with Rhodamine-PE, cells were incubated in suspension at 4°C with 3 µM fi nal of the probe, washed with PBS 3% BSA, and incubated for an additional 3 h at 37°C. Cells were seeded on coverslips and pulsed for 4 h with fl uorescent exosomes. After washing, cells were fi xed with PFA and examined with the LSM 510. found in the present study are reported in Table 1 . Typical exosome markers, such as the transferring receptor, tetraspanins (CD63, CD81, CD82), and heat shock proteins, were detected ( Table 1A ), assessing the quality of the preparation. Exosomes were also characterized by their size and their electronegativity (supplemental Fig. II).
Regarding lipid-related proteins, we found a phospholipid scramblase, a protein that transports phospholipids between the two membrane leafl ets, in both directions. The presence of this protein was consistent with the lack of membrane phospholipid asymmetry we reported earlier in RBL-derived exosomes ( 16 ). Also a member of the fatty acid binding proteins (E-FABP) was detected. FABPs constitute a multigene family of structurally homologous cytosolic proteins that bind and transport polyunsaturated fatty acids, such as arachidonic acid (AA) ( 31 ). Another type of protein was a prostaglandin F 2 receptor negative regulator, also called FPRP ( 32 ). FPRP associates with the PGF 2 ␣ receptor, thereby reducing ligand binding ( 33 ). However, the PGF 2 receptor was not found in exosomes, and the presence of the FPRP protein might be better related to its ability to form a tight complex with the tetraspan molecule CD81 ( 32 ).
Among the phospholipases, only phospholipase C ε hydrolyzing phosphoinositides was detected ( Table 1B ). Note that proteins known to interact with phospholipases D and A2 were present ( Table 1C ). Fructose bisphosphate aldolase interacts directly with phospholipase D isoform PLD 2 and inhibits its activity ( 34 ). The exosomes contained casein kinase II (cK2) that can phosphorylate PLD 2 ( 35 ) and can also interact with sPLA 2 -IIA ( 36 ), precisely one of the sPLA 2 isoforms we detected in the present work. Hsp 70, one of the typical exosome markers ( Table 1A ), has also been shown to interact with iPLA 2 ( 37 ).
Phospholipase C ε has been shown to be regulated by G proteins, either the subunits of heterotrimeric G proteins or monomeric GTPases ( 18 ), both being recovered in the exosomes ( Table 1D ). This prompted us to consider that GTPases could participate in the regulation of exosome lipolytic enzymes. Note that exosomes contained almost all members of the Ras superfamily GTPases [ARF, Rho, Rap, Rab, p21Ras, and Ran ( Table 1D )] except Cdc42 ( 38 ). Possible pathways connecting the Ras superfamily GTPases and phospholipases have been reported. The GTPases RhoA and Arf 6 ( Table 1D ) are direct activators of PLD 2 ( 39, 40 ) from rat or human origin ( 41 ). Fig. 1A reports the presence of DG, PEt, and LPC when RBLwt exosomes were incubated with the fl uorescent and membrane-diffusible phosphatidylcholine. We investigated whether an autonomous regulation of the lipolytic enzymes involved in the production of these lipid mediators could occur in exosomes. Addition of GTP ␥ S up to 300 µM in RBLwt exosomes had no effect on the PLD activity ( Fig. 1B, curve a). The activity was not increased by exosome sonication. Immunodetection showed the selective sorting of the PLD 2 isoform in exosomes ( Fig. 1C, lane 2) compared with the parental cells, containing mainly group IIa and V recombinant sPLA 2 proteins as standards, and transferred onto PVDF membrane. Membranes were saturated in NETG buffer (150 mM NaCl, 5 mM EDTA, 50 mM Tris pH 7.4, 0.05% Triton X-100, 0.25% gelatin), washed in PBS 0.05% Tween 20, and incubated in NETG buffer with HRP-labeled anti-rabbit IgG secondary antibodies.

Exosomes contain the PLD/PAP pathway
In all cases, the signal was detected by the enhanced chemiluminescence system from GE Healthcare/Amersham.
Cyclooxygenase detection by fl ow cytometry. Exosomes (5 µg proteins) were bound on 4 µm beads (5 µl aldehyde sulfate latex beads; Invitrogen) for 1 h at room temperature under gentle shaking. Unoccupied sites were saturated with vesicle-free FCS for 1 h at room temperature. Beads were spun down, washed with PBS, and resuspended in FACS buffer. Bound exosomes were incubated with anti-COX-1 or anti-COX-2 primary antibodies, or control isotype for 1 h at room temperature and spun down, then labeled for 30 min with secondary FITC-labeled antibody. COX expression was analyzed by fl ow cytometry on the FITC channel (FL-1) of a FACSCalibur analyzer (Becton-Dickinson) using settings previously reported ( 16 ).
Protein analysis. High-throughput protein analysis was performed on 100 µg protein of purifi ed exosome, fi rst separated by one-dimensional SDS-PAGE. The protein gel lane was cut into 16 pieces, which were digested by trypsin. Tryptic peptides were analyzed by nanoLC-MS-MS with a Qstar XL spectrometer (Applied Biosystems). Data were searched against mouse entries in Sprot-Trembl with the Mascot software. Analyses were performed at the IPBS, CNRS, Toulouse, France ( 27 ).
Lipid analysis. Determination of free fatty acids in exosomes was performed at the lipidomics facility of IFR-BMT (Institut Fédératif de Recherche Bio-Medicale de Toulouse), Toulouse, France. RBL-2H3 derived exosomes (100 µg) were incubated in PBS with 2 mM Ca 2+ /Mg 2+ at 37°C for 4 h. The lipids were extracted by the Bligh and Dyer method ( 28 ) in the presence of EGTA in water. Fatty acids were methylated and further analyzed by gas chromatography with an HP5890 instrument ( 29 ).
Prostaglandins were quantifi ed by GC-MS at the lipidomics facility of IMBL/INSA-Lyon, Villeurbanne, France. Lipids from RBL-2H3 derived exosomes (70 µg) were extracted with ethylacetate, derivatized into pentafl uorobenzyl esters, purifi ed by silica gel TLC using chloroform/ethanol (93:7, v/v), and then modifi ed into trimethylsilyl ethers before analysis by GC-MS. Samples were spiked with 10 ng of deuterated prostaglandin standards (Cayman) and GC-MS was carried out with a Hewlett Packard quadrupole mass spectrometer interfaced with a Hewlett Packard gas chromatograph ( 30 ).
Data presentation. HPLC profi les and phospholipase determinations representative of at least two independent experiments were plotted. Pooled exosome preparations from two to three experiments were used for protein and lipid analysis, which fulfi lled the quality control stipulations of the respective facilities (IPBS and IFR-BMT, Toulouse, France; IMBL/INSA-Lyon, Villeurbanne, France). Confocal pictures were representative of at least two experiments performed by distinct operators. Errors bars corresponded to SEM from three determinations.

Lipid-related proteins in exosomes
To determine which of the diverse lipid-related proteins were present on the exosomes, we fi rst performed an exhaustive protein analysis of the vesicles. The proteins strong conversion into diglycerides was observed ( Fig. 1E ). Kinetics analysis demonstrated that 60% of the PA was hydrolyzed into diglycerides within 15 min ( Fig. 1F ), indicating the presence of a very active PAP1 in intact exosomes. The kinetics of PA hydrolysis were similar in the presence or absence of GTP ␥ S ( Fig. 1F ).

Exosomes from RBL cells carry GTP-activatable PLA 2 and contain the three classes of PLA 2
Fig . 1A reported the presence of PLA 2 activity as evidenced by the high LPC content. During the course of the studies on PLD activation ( Fig. 1 ) we noticed a GTPdependent enhancement of the LPC peak, both on RBLwt and RBLpld2 exosomes.
We then investigated whether a dynamic regulation of PLA 2 activity might occur in RBLwt exosomes ( Fig. 2 ). GTP ␥ S was able to reveal PLA 2 activity on intact exosomes incubated in calcium-free PBS and in the presence of the inhibitor MAFP ( Fig. 2A ). GTP ␥ S dose-dependent PLA 2 activation ( Fig. 2B ) fi ts a hyperbolic curve as shown by the linearity of the double-reciprocal plot ( Fig. 2B , insert).
Exosomes exhibited a higher PLA 2 activity following sonication, indicating that the PLA 2 were partly located in the exosome lumen. The relative parts played by each PLA 2 class (cytosolic calcium-dependent cPLA 2 , cytosolic calcium-independent iPLA 2 , and secreted sPLA 2 ) were next investigated in the presence of GTP ␥ S on sonicated PLD1 ( Fig. 1C, lane 1). PLD 2 activation in RBLwt exosomes could be repressed by aldolase A, which was reported by the protein analysis ( Table 1 [P05064] fructose bis-phosphate aldolase) and has been established as a direct inhibitor of PLD 2 by acting on its PH domain ( 34 ). Therefore, the occupation of the PH domain might prevent activation of the phospholipase.
We expected to modify the natural stoichiometry between the putative inhibitor aldolase A and PLD 2 by overexpressing the human HA-PLD 2 in RBL cells (supplemental Fig. II), the hHA-PLD 2 being targeted to exosomes ( Fig.  1C , lanes 3, 4). Indeed, the basal PLD activity in RBLpld2 exosomes was twice as high as that of RBLwt exosomes ( Fig. 1B ; GTP ␥ S = 0). When increasing amounts of GTP ␥ S were added, a clear GTP dependency of the PLD activity in RBLpld2 exosomes was then observed ( Fig. 1B, curve b). Phospholipase D activity generates the transphosphatidylation product PEt in the presence of ethanol ( Fig. 1A ), which competes with the water required to form PA. However, even in the absence of ethanol, PA was not detected across the chromatograms, suggesting the presence of a phosphatidate phosphatase (PAP1) on exosomes. When purifi ed BODIPY-PA was injected into the HPLC system, the resulting peak exhibited a typical asymmetrical shape ( Fig. 1D ), which was not observed in any chromatograms obtained from exosome incubations with BODIPY-PC. Indeed, upon incubation of BODIPY-PA with exosomes, a Ras-related prorein Rab Q9QUI0 Transforming protein RhoA Interacts and activates phospholipase D (PLD2) P63835 Ras-related protein Rap-1A Q61411 Transforming protein P21/H-Ras-1 (c-H-Ras) Q61820 GTP-binding nuclear protein Ran P62331 ADP-ribosylation factor 6 Interacts and activates phospholipase D (PLD2) Q8BGX0 GTP-binding protein ARD-1 (ADP-ribosylation factor domain protein 1) Only proteins related to exosome markers, lipid metabolism, and G proteins are reported. The overall analysis identifi ed 382 different proteins. Observations are detailed in the text.
whereas bromoenolactone (BEL), the specifi c iPLA 2 inhibitor, abolished 39% of overall PLA 2 activity ( Fig. 2C ). The concentration of inhibitors was 50 µM; i.e., above that used to inhibit the various PLA 2 in cells ( 24,25 ). The sum of the inhibitions triggered by pyrrolidine-1 and BEL in exosomes. We observed that MAFP decreased the total PLA 2 activity by 60% ( Fig. 2C ). MAFP inhibits both cPLA 2 and, to a lesser extent, iPLA 2 ( 25 ). We next checked specifi c inhibitors. The specifi c cPLA 2 inhibitor (pyrrolidine-1) reduced total PLA 2 activity by 37% ( Fig. 2C ),   Only one member of the omega-3 series, namely, docosahexaenoic acid (DHA), was detected. Interestingly, AA accounted for about 30% of the total polyunsaturated fatty acids. We next investigated whether bioactive lipids derived from AA could be found in exosomes. Quantifi cation of prostaglandins was performed by GC-MS and demonstrated the presence of mainly PGE 2 and 15-deoxy-⌬ 12,14 -PGJ 2 [15d-PGJ 2 ] ( Fig. 3E ). The 15d-PGJ 2 was slightly enriched in the vesicles compared with the parent cells. The respective amounts of the various prostaglandins (PGF 2 ␣ , PGE 2 , PGD 2 , and 15d-PGJ 2 ) in exosomes was not enhanced by incubation with GTP ␥ S ( Fig. 3E ), indicating that the prostaglandins originated either from the basal exosome PLA 2 activities or were loaded in the exosome membrane at the time of their biogenesis in parent cells. Note that COX-1 and COX-2 involved in the early steps of prostaglandin biosynthesis were expressed in exosomes ( Fig. 3F ), indicating that exosomes could be autonomous biological structures for the biosynthesis of the various prostaglandins. In that respect, the AA concentration in exosome membrane (45 nmoles/mg protein; Fig.  3D ) was in excessive compared with the total membrane pros taglandin concentration (0.31 nmoles/mg protein; Fig. 3E ).
The number of exosome vesicles per unit protein was established and used to calculate the amount of prostaglandin associated with a defi ned number of vesicles. Compared with the same number of parental cells, exosomes carried from 12 to 15 times more prostaglandins ( Fig. 3G ). To our knowledge, this is the fi rst report of vesicle-associated release of prostaglandins from cells.
To evaluate the potential of exosomes as vehicles of bioactive lipids, we investigated whether exosomes could traffi c between RBL-2H3 cells.

Exosomes are internalized by resting and activated RBL-2H3 cells and concentrate into endosomes
Confocal microscopy performed on living cells showed an accumulation of exosomes on the cell periphery detectable as soon as 5 min, with subsequent internalization leading to the formation of intracellular aggregates indicating storage in an endosomal compartment, as observed after 1 h and 4 h ( Fig. 4A ). Exosome uptake was an active process, as cross-linking of peripheral protein on target cells by paraformaldehyde impaired intracellular exosome accumulation ( Fig. 4A , bottom panels). Only faint, diffuse cell labeling was observed in this case and was attributed to some exchange of the lipidic fl uorescent probe between exosomes and the target cell during the step of exosome interaction with the peripheral cell membrane. We investigated whether activated cells, which release exosomes, were also able to internalize them. RBL-2H3 cells activated exosomes ( Fig. 2C ) led to the reduction of global PLA 2 activity by 76%, indicating that another type of PLA 2 activity was present. Me-indoxam, a specifi c inhibitor of secreted phospholipases ( 42 ) was checked, and it decreased total PLA 2 activity by 19% ( Fig. 2C ). Together, the cumulative effect of the three inhibitors diminished total PLA 2 activity by 95 ± 6.5%. The residual 5% activity might be related to PLA 2 insensitive to the inhibitors, such as some secreted sPLA 2 ( 43 ). Therefore, the three classes of PLA 2 contributed to the global PLA 2 activity detected in the exosomes.
We next assessed the presence of members of the three PLA 2 classes by using specifi c antibodies ( Fig. 2D ). cPLA 2 -IVA was detected as a single band, whereas iPLA 2 -VIA was present as processed forms ( 44 ) The iPLA 2 -VIA can be cleaved at three different sites by caspase 3, which generates various processed forms depending upon the combination of the sites effectively cleaved ( 45 ). Fragmented forms of iPLA 2 , similar to those we reported in Fig. 2D, were observed in erythrocyte-derived exosomes ( 44 ). Proteolytic processing has been shown to enhance the iPLA 2 -VIA activity by removing part or complete ankyrin repeats suggested to function as a negative regulator ( 45 ). The form of 66 kDa we observed in Fig. 2D could correspond to the residual protein after caspase 3-mediated cleavage at the DVTD site of the iPLA 2 , leading to the release of the fi rst ankyrin repeat ( 45 ) and making likely that the iPLA 2 -VIA is highly active in exosomes. Among the third class of PLA 2 , namely, secreted sPLA 2 , the presence of sPLA 2 -IIA and sPLA 2 -V groups was observed ( Fig. 2D ). Therefore, RBL exosomes concentrated members of each of the three classes of PLA 2 and are thus a unique cell compartment.

Exosomes as carriers of bioactive lipids
A large panel of free fatty acids was recovered from exosomes ( Fig. 3A ). Fatty acids could be carried from the parental cells or directly generated within the exosomes by the phospholipase A 2 activities. We previously established that 1 mg of exosome protein contained 230 nmoles phospholipid [see Ref. ( 16 )]. Therefore 230 nmoles free fatty acid could be potentially released by the respective PLA 2 activities considering they displayed 100% effi ciency, with an additional amount of free fatty acid originating from the lysophospholipase activity borne by cPLA 2 and iPLA 2 . However, a total amount of 1,420 nmoles free fatty acid/ mg exosome protein was measured ( Fig. 3A ), indicating that most of the exosome free fatty acid content was already present at the time of exosome membrane biogenesis. The chain length of the saturated fatty acids ranged from 14 to 24 carbons ( Fig. 3B ); the major ones were palmitic and stearic acids. Monounsaturated fatty acids were essentially from the omega-9 series ( Fig. 3C ); oleic acid was the most abundant. Polyunsaturated fatty acids almost exclusively contained members of the omega-6 series ( Fig. 3D ).
respectively. F: Exosomes contain cyclooxygenases 1 and 2. Analysis of exosome cyclooxygenase expression by fl ow cytometry, compared to control isotype (gray shaded curves). G: Comparative prostaglandin content of exosomes and parent cells. Prostaglandin content per mg protein plotted in Fig. 3E was converted into cell-equivalents or vesicle (exosome)-equivalents, and normalized to 10 6 cells or 10 6 exosome vesicles, respectively. 1 mg protein corresponded to (71.4 ± 0.54) × 10 5 cells and (5.96 ± 0.13) × 10 5 exosome vesicles. by Fc ε -RI cross-linking appeared even more effi cient at internalizing exosomes compared with resting ones ( Fig. 4B ). Uptake was linear at least up to 1 h of incubation as shown by fl ow cytometry monitoring of exosome internalization ( Fig. 4C ). Similar data were obtained by measuring the BODIPY-ceramide content of target cells following organic extraction of the probe after exosome internalization. This procedure allowed us to quantify the amount of internalized exosomes. It was found that 1 g of exosome was internalized in 1 h in 10 6 resting cells ( Fig. 4D ).
Late endosomes in RBL-2H3 cells feature a dual function: being able to release their contents upon stimulation and being recipients of an endocytosis activity ( 46 ). We investigated whether exosomes could be colocalized with endosome markers following their incubation with target cells. Three markers were checked: (i) CD63, a general marker of late endosomes in RBL cells ( 47 ); (ii) the lysolipid lysobisphospatidic acid (LBPA; also called BMP for bis[monoacylglycero]phosphate) that accumulates in MVB ( 21 ); i.e., late endosomes containing intralumenal entire endosomal compartment is about 500 times lower than the total cell volume ( Table 2 ). Therefore exosometransported lipid mediators accumulated inside endosomes ( Figs. 4, 5 ) were 500 times more concentrated than if they were diluted in the whole cell volume ( Table  2 ). As a consequence, the resulting PGJ 2 endosomal concentration reached 52 µM ( Table 2 ). Other endosomal concentrations were 33µM, 2.4 µM, and 1.7 µM for PGE 2 , PGD 2 , and PGF 2 ␣ respectively, whereas fatty acids, such as AA or DHA, reached millimolar concentrations (4 mM and 0.9 mM, respectively). Thus, target cell endosomes behave as a "concentrator compartment" of lipid mediators transported by exosomes, allowing micromolar concentrations of prostaglandins to be reached (i.e., concentrations able to trigger further biological responses, such as PGJ 2 -mediated PPAR ␥ activation). Note that the extent of exosome internalization by cells (1 µg exosomes/10 6 cells/hr; Fig. 4D ) was similar to the amount of exosomes released by 10 6 cells upon stimulation (1.44 ± vesicles (see supplemental Fig. IIa); and (iii) Rh-PE, a fl uorescent lipidic probe that accumulates specifi cally in late endosomes. Fig. 5A -C shows that exogenously added exosomes were labeled inside the cells by anti-CD63 antibody internalized by fl uid-phase endocytosis. Localization of exosomes inside the endocytic track was more precisely investigated by monitoring the MVB distribution with anti-LBPA labeling (Fig. 5D-F). Remarkably, colocalization of exosomes was observed inside MVBs located close to the nucleus ( Fig. 5F , white circles and arrows). With Rh-PE as the MVB-specifi c probe, supplementary evidence was obtained that exogenous exosomes joined the MVB compartment located close to the nucleus ( Fig. 5G-I ).
We estimated the concentration that could be reached by exosome-transported lipid mediators accumulated inside the endosomal compartment of a target cell. Considering an average diameter of 600 nm for an RBL-2H3 cell endosome ( 48 ), with an average number of 30 endosomes per RBL-2H3 cell ( 49 ), the resulting volume of the expression scores, whereas PLDs were not even detected by high-throughput analysis. Purifi ed aldolase dosedependently inhibits the PLD 2 activity ( 34 ); the inhibitor interaction occurs at the PH domain of PLD 2. This interaction might impair GTPases, such as RhoA or Arf6, to activate PLD 2 , or it might prevent another domain of PLD 2 , the phox homology domain (PX) that exhibits GAP activity, from activating GTPases ( 53 ).
Interaction domains on Arf 6 and RhoA with PLD 2 have been mapped ( 39,40 ); however, no direct interaction between any of the small G proteins reported in the present study and PLA 2 have been reported so far. In whole cells, the cPLA 2 and the sPLA 2 -IIA can be activated downstream of RhoA GTPases with subsequent effects on PGE 2 formation ( 54 ). Similarly, iPLA 2 activation downstream of RhoA has been suggested ( 55 ). However, in a cell-free system, such as exosomes, the signaling network between RhoA and the phospholipases might be different compared with whole cells, and exosomes might reveal a specifi c regulation of PLA 2 activities.
Regarding the functional role of exosome phospolipases A2, the calcium-independent iPLA 2 has been shown to allow the elimination of erythrocyte-derived exosomes by apoptotic cells ( 44 ). Concerning sPLA 2 , exosomes transporting sPLA 2 IIA and V ( Fig. 3 ) might account for the transcellular activity of these phospholipases reported to occur from activated RBL-2H3 cells ( 56 ). The group sPLA 2 -V was reported to be secreted from RBL-2H3 cells and to trigger eicosanoid biosynthesis in neighboring target granulocytic cells ( 56 ). The sPLA 2 -V activity appears related to IgE-dependent PGD 2 formation and to enhanced exocytosis in RBL-2H3 cells ( 57 ). Vesicular secretion of cPLA 2 has not been reported so far. 0.47 µg), suggesting an effi cient cell-to-cell communication process.

DISCUSSION
Exosomes are nanovesicles released from intact viable cells. They participate in cell-to-cell communication in various physiological and pathological situations, such as the immune response ( 50 ), infl ammation ( 51 ), or atherogenesis ( 7 ). Mast cell-derived exosomes trigger functional maturation of dendritic cells ( 6 ). This maturation process involves secreted PLA 2 ( 13 ) and prostaglandins ( 14 ). Therefore, we investigated the presence of lipid-related proteins and lipid mediators on exosomes derived from the mast cell line RBL-2H3.
High-throughput protein analysis reported the presence of only four proteins related to lipid metabolism. However, we revealed the presence of other lipolytic proteins by their activity and by immunodetection. The presence of a high content in monomeric G proteins led us to hypothesize specifi c regulation of these phospholipases in exosomes. The subfamilies of Ras GTPases reported in Table 1 are cytosolic proteins likely to be located inside the exosomes. Therefore, GTP must cross the exosome membrane to activate GTPases and, subsequently, phospholipases. GTP transporters might be present in exosomes, as they have been reported in synaptic vesicles ( 52 ), a type of vesicle very similar to exosomes.
The difference in activation of the PLD by GTP between RBLpld2 and RBLwt exosomes appeared related to the stoichiometry between PLD 2 and aldolase A. Among the proteins recovered by protein analysis and reported in Table 1 , aldolase (P05064) exhibited one of the highest The amount of PGJ2 (column D) transported by the amount of exosomes plotted in column C was converted into µM concentrations (column E) either by considering the total volume of the recipient cells (column E line 2), or the total volume of late endosomes present in cells (column E line 4). The resulting values indicated that PGJ2 was about 500 times more concentrated in the total endosome compartment than in the total cell volume (column F). Column A: Values were obtained from literature data (see "Results") and correspond to RBL-2H3 cells. zymes, such as COX-1 and COX-2 ( Fig. 3F ), for the early steps of prostaglandin biosynthesis, exosomes could account for transcellular metabolism of prostanoids reported to occur between normal and tumor cells ( 59 ). Arachidonic acid present in exosomes ( Fig. 3D ) would serve as transcellular biosynthetic precursor. Although eicosanoid transcellular metabolism has been reported to occur at in- The set of prostaglandins transported by exosomes ( Fig.  3 ) are derived from PGH2, and metabolic conversion of PGH2 has been shown to occur through a transcellular mechanism between two different types of cells, containing either COX-1 and COX-2 or the terminal prostaglandin synthases ( 58,59 ). Transporting from cell to cell metabolic precursors, such as PGD 2 ( Fig. 3G ), and en- Fig. 6. Exosomes as intercellular signalosomes carrying GTP-activatable phospholipases and prostaglandins A: Exosomes carry GTP-activatable phospholipases. Proteins detected in RBLwt-derived exosomes related to phospholipase activation and reported in Table 1 were represented together with the phospholipases detected in the present work. Arrows indicate possible activation pathways based on literature data (see "Discussion"). Note the presence of lipid mediators in the exosome membrane. B: Exosomes as intercellular "signalosomes." Exosomes released upon cell activation can traffi c between resting cells (white) and activated cells (color). Exosomes could trigger autocrine and paracrine-type signals. C and D: Possible mechanism of exosome-mediated bioactive lipid delivery from endosomes in target cells. Inside the intracellular compartment accumulating exosomes in target cells (endosomes, Fig. 5 ), phospholipases borne by exosomes could participate upon activation by GTP in the fusion between exosomal and endosomal membranes, allowing delivery of exosome content into the cytosol. Exosomes carry prostaglandins, such as the PPAR ␥ agonist 15d-PGJ 2 ( Fig. 3 ) and a FABP ( Table 1 ) which can bind arachidonic acid ( Fig. 3 ) and interact with PPAR ␥ (Fig. 6C). The 15d-PGJ 2 /PPAR ␥ and FABP/arachidonic acid/PPAR ␥ complexes would be further addressed to the nucleus (Fig. 6D).
Exosomes could also supply the 15d-PGJ 2 already bound to its receptor, as a recent report indicates the presence of the PPAR ␥ receptor among proteins found in exosomes isolated from human serum ( 71 ). Interestingly, the exosome FABP we report in Table1 could bind the AA present in exosomes ( Fig. 3 ) and then interact directly with the PPAR ␥ receptor, the resulting FABP-AA-PPAR ␥ complex being subsequently addressed to the nucleus of target cells to regulate transcription ( Fig. 6C ) ( 72 ). In line with the possible modulation of nuclear receptors by exosome-carried mediators, note that PAP1, the diglyceride-generating enzyme we reported in Fig. 1 , has recently been characterized as a transcriptional coactivator of the PPAR ␣ receptor ( 73 ).
Further experiments are required to support the functional role of exosomes in RBL-2H3 cells. As a fi rst step, we evaluated whether exosomes could carry suffi cient amounts of the prostaglandin 15d-PGJ 2 to possibly trigger PPAR ␥ activation in target cells. When added to cells, 15d-PGJ 2 has been reported to trigger biological effects in the 10-40 µM range ( 74 ). Exosome accumulation in endosomes were allowed to reach values > 50 µM ( Table 2 ); i.e., bioactive 15d-PGJ 2 concentrations.
Because of the dynamic regulation of their phospholipases by GTP, exosomes appear to behave as "signalosomes" ( Fig. 6A ). The "signalosomes" would circulate between cells and might regulate their functions whether cells are resting or activated ( Fig. 6B ). Stimulated RBL-2H3 cells feature enhanced endocytosis ( 46 ) and could internalize exosomes they had just released. Preliminary data we obtained indicate that exosomes inhibited Fc εmediated degranulation of RBL-2H3 cells. That this effect involves PGE 2 , which is known to inhibit Fc ε RI-mediated exocytosis of mast cells ( 70 ), appears conceivable on the basis of data reported here. Also, by possibly providing 15d-PGJ 2 to PPAR ␥ of target cells, exosomes can repress the transcription of proinfl ammatory mRNAs ( 75 ). Circulating simultaneously with allergens that activate cells via Fc ε RI receptors, exosomes appear as a signaling device able to modulate the Fc ε RI-mediated mast cell response by means of phospholipases and lipid mediators that can be activated.
fl ammation sites between different cell types ( 60 ), one can conceive that RBL-derived exosomes are a mixed population bearing either the COX-1 and COX-2 or the terminal prostaglandin synthases; therefore, exosome exchange between RBL-2H3 cells would be required to complete the entire prostanoid biosynthesis pathway. In this respect, we showed earlier that RBL-2H3 cells release three distinct subpopulations of exosomes ( 23 ). The present work opens further investigations to understand the mechanisms underlying the transcellular metabolism of eicosanoids.
This transcellular metabolism requires exosome traffi cking between cells. We have shown that exosomes added to target cells are rapidly internalized ( Fig. 4 ) into the endocytic track and join the MVB network located close to the nucleus ( Fig. 5 ). It is likely that the GTP-dependent activation of PLD and PLA 2 we observed in exosomes could occur inside the endocytic track of target cells. Many GTPases are present within the endocytosis track, some of them maintained in an active state even in unstimulated cells ( 61 ). GTP-activated phospholipases could participate in exosome fusion with the limiting membrane of the endosome, a process called "back-fusion" ( 62,63 ). This process allows the lumen content of the exosomes to be released into the cytosol. Back-fusion molecular mechanisms require the lipid LBPA, whose biosynthesis involves a cPLA 2 -type activity ( 64,65 ), as well as a combination of PLA 2 and PLD activities ( 66 ). A previous report describes GTP-dependent cPLA 2 -mediated fusion of secretory granules ( 67 ). Phosphatidic acid resulting from PLD activity is a fusogenic compound in presence of calcium ( 68 ). Diglycerides generated by the PI-PLCε ( Table 1 ) or the PLD/ PA phosphatase pathway ( Fig. 1 ) could participate in exosome-endosome fusion processes by lowering the surface pressure of the phospholipids ( 69 ). More DG can be expected in RBLpld2 exosomes and could account for the modifi cation of the biophysical parameters (size and electronegativity) shown in supplemental Fig. II. In addition, phospholipid mixing between exosome and endosome membranes triggered by the scramblase we reported in Table 1 would facilitate membrane fusion.
We established in this work that exosomes transport prostaglandins from the parent cells. RBL-2H3 cells feature a mast cell phenotype, and eicosanoids play an essential role in mast cell physiology by regulating their function in host defense and disease ( 70 ). PGE 2 can block Fc ε RImediated exocytosis of mast cells ( 70 ). Exosomes, during at least the fi rst 5-20 min ( Fig. 5 ), provide a vehicle for PGE 2 to interact with its respective GPCRs on the periphery of target cells. Thereafter, exosome internalization provides the fi rst mechanism described for 15deoxy ⌬ 12,14 -PGJ 2 to enter the cells and possibly reach its intracellular targets. Actually, no specifi c peripheral receptors or mechanisms of entry have been identifi ed to-date for this prostaglandin ( 17,19 ). The exosome as a vehicle would allow the plasma membrane to be bypassed and 15d-PGJ 2 to accumulate in the endosomes of target cells, from where the prostaglandin would be released into the cytosol after fusion between exosome and endosome membranes. A possible mechanism is summarized in Fig. 6 .