Activation of Macrophages by P2X₇-Induced Microvesicles from Myeloid Cells Is Mediated by Phospholipids and Is Partially Dependent on TLR4

Released microvesicles (MV) mediate intercellular communication between multiple cell types, and affect cytokine secretion, inflammatory processes, and tumor progression (1). MV can be released from intracellular storage sites or shed directly from the cell surface. Release of plasma membrane-derived MV is typically regulated by intracellular Ca²⁺-mediated processes (2, 3), or by protein kinase B (4) or protein kinase C (PKC) (5, 6), and may involve engagement of a number of cell-type–specific receptors. Shedding typically involves a budding process, in which surface blebs selectively accumulate cellular constituents that are then packaged into MV. In contrast, MV deriving from inside cells include secretory lysosomes (7–9), characterized by expression of lysosomal proteases and other markers, and exosomes (1, 10), which are stored in multivesicular bodies before being released by an active exocytic process.

In macrophages and other myeloid cells, engagement of P2X₇ purinergic receptors triggers the release of MV deriving from both inside the cell (8, 11, 12) and the plasma membrane (2, 3, 13–15). Secretion of IL-1β mediated by P2X₇ was suggested to occur through the release of cell surface-derived MV many years ago (14), but the content of these MV is still unclear. A more recent study suggests that the majority of IL-1β–containing MV may be exosomes as opposed to larger shedding vesicles (11). Members of the IL-1 family, including IL-1β, IL-1α, and IL-18, are stored as inactive precursors, and their release after processing can be MV mediated (2, 3, 11, 14), but this does not exclude other mechanisms of secretion (16–18). In any case, rapid secretion of mature IL-1β requires the Nod-like receptor family, pyrin domain containing 3 (NLRP3) inflammasome, which recruits caspase-1, allowing cleavage of the proform of the cytokine to its bioactive form (19). Some exosomes from macrophages contain class II MHC, and their exocytosis requires the NLRP3 inflammasome despite being caspase-1 independent (12). Less clear is the requirement for NLRP3 in controlling the P2X₇-dependent release of secretory lysosomes (8), although their exocytosis from human monocytes is caspase-1 independent.

Macrophages and other cells of the phagocytic lineage bind and internalize MV from various sources with diverse effects on their function (10). Human monocytes binding tumor-derived MV become activated, expressing increased HLA-DR, reactive oxygen species, and TNF-α through a CD44-dependent mechanism (20). In another study, human tumor-derived MV promote differentiation of CD14⁺ myeloid cells, resulting in an HLA-DR low phenotype lacking CD8 and CD86 expression, and secreting TNF-α, TGF-β, and IL-6, with suppressive effects on T cell proliferation (21). Less is known about potential immune stimulatory effects that macrophage-derived P2X₇-induced MV may exert, which could involve IL-1β.

To address this question, we induced MV shedding from nonprimed mouse primary macrophages or cell lines through P2X₇ activation and tested isolated MV for their ability to activate bone marrow-derived macrophages (BMDM). We found that MV were able to activate BMDM in a partially TLR4-dependent manner, and that the stimulatory component within the MV was found within phospholipid fractions. MV-derived phospholipids activated macrophage through TLR4. Furthermore, MV-induced activation is independent of observed loaded cargo, such as IL-1β, TNF-α, and high-mobility group box 1 (HMGB1).

J774A.1 (TIB-67; American Type Culture Collection, Manassas, VA), a murine macrophage cell line, and D2SC-1, a murine splenic derived immature dendritic cell (DC) line (a gift of L. Kane, University of Pittsburgh, Pittsburgh, PA), were maintained in DMEM (Mediatech, Manassas, VA) supplemented with 10% FBS (Gemini Bio-Products, West Sacramento, CA), 1% additional l-glutamine (Lonza, Basel, Switzerland), and 1% penicillin and streptomycin (Lonza; hereafter called DMEM complete). FSDC, a murine fetal skin-derived immature DC line (a gift of P. Ricciardi-Castagnoli, Singapore Immunology Network, Singapore), and THP-1, a human monocyte cell line (American Type Culture Collection), were maintained in IMDM (Lonza) supplemented with 10% FBS, 1% additional l-glutamine, and 1% penicillin and streptomycin (hereafter called IMDM complete). For experiments, THP-1 cells were treated with 20 μM PMA for 2 d to differentiate them to become more macrophage-like. RAW264.7 murine macrophage cell line transfected with NF-κB reporter plasmid pNF-κB-MetLuc Vector (BD Clontech, Mountain View, CA) encoding inducible Metridia luciferase protein expression and secretion was a gift of R. Binder (University of Pittsburgh) and was maintained in DMEM complete supplemented with 500 μg/ml G418. Murine BMDM were derived from C57BL/6 bone marrow (gift of L. Borghesi, University of Pittsburgh) and were differentiated with L cell-supplemented media, as described previously (22). TLR4^−/− mouse bone marrow (23) was a gift of R. O’Doherty (University of Pittsburgh). RAGE^/− mouse bone marrow (24) was a gift of T. Oury (University of Pittsburgh). All knockout mouse bone marrow described were of the C57BL/6 background.

Other reagents include ATP (Thermo Fisher Scientific, Waltham, MA), A23187 (Sigma-Aldrich, St. Louis, MO), A 438079 (Tocris Bioscience, Ellisville, MO), A 740003 (Tocris Bioscience), brefeldin A (Sigma-Aldrich), SB203580 (EMD Chemicals, Gibbstown, NJ), BATPA/AM (EMD Chemicals), bisindolylmaleimide I HCl (BIM; EMD Chemicals), MDL-12330A (Enzo Life Sciences, Farmingdale, NY), H-89 2HCl (EMD Chemicals), CAY10593 (Enzo Life Sciences), CAY10594 (Enzo Life Sciences), bromoenol lactone (BEL; Enzo Life Sciences), LPS from Escherichia coli 026:B6 (Sigma-Aldrich), poly(I:C) (Sigma-Aldrich), synthetic monophosphoryl lipid A (MPLA; InvivoGen, San Diego, CA), Limulus amebocyte lysate QCL-1000 (Lonza; used according to manufacturer’s protocol to assess endotoxin levels), and soluble RAGE (24) (gift of T. Oury).

To measure P2X₇-induced large pore formation, 1 × 10⁶ J774A.1 were pretreated with or without 100 μM A 740003 for 15 min before exposure to ATP for 30 min, and then were stained with 5 μM Yo-Pro-1 (Invitrogen, Carlsbad, CA). Cells were kept on ice until flow cytometric analysis, which was performed with a BD Biosciences LSR II, and results were analyzed using FlowJo software (Tree Star, Ashland, OR).

J774A.1 and other myeloid cell types were plated in T225 cm² flasks in duplicate per treatment. Cells were washed twice with PBS before adding inhibitors in serum-free DMEM with no additions. Cells were then treated with 3 mM ATP and 10 μM A23187 in a final volume of 20 ml serum-free DMEM with no additions for 30 min at 37°C. Supernatant was harvested and centrifuged at 309.1 × g for 10 min at 4°C to remove cells and larger debris. MV were collected by centrifugation at 100,000 × g ultracentrifugation for 1 h at 4°C. The pellet material from the 100,000 × g ultracentrifugation was resuspended in 500 μl PBS. MV were disrupted with 10 passes through a 27-gauge needle. Bradford assay (Thermo Fisher Scientific) was used to determine the protein concentration with each MV fraction, according to manufacturer’s specifications. MV were either used immediately or stored at −20°C for later use.

For biochemical fractionation of MV, proteins and lipids were separated through the Bligh and Dyer method of protein/lipid extraction with 1:2 of chloroform and methanol (25). The protein fraction was harvested at the biphasic interphase and reconstituted in PBS. The lipid fraction was either dried by speed vacuum centrifugation or further separated through a lipid polarity extraction technique (26). Lipid extract was passed through a silicic acid (Sigma-Aldrich) column (1 mg silicic acid/1 μl vol lipid extract). The pass through was collected as a sample. Bed layer volumes of chloroform, acetone, and then methanol were passed in sequential and separate fashion to elute off potential neutral lipids, glycolipids/sulpholipids, and phospholipids, respectively. Samples were harvested from each elute, dried with speed vacuum centrifugation, and stored at −20°C.

Ten million J774A.1 were washed twice in PBS, pelleted, and resuspended in homogenization buffer (100 mM KCl, 25 mM NaCl, 2 mM MgSO₄, 12 mM sodium-citrate, 10 mM glucose, 25 mM HEPES, 5 mM ATP, 0.35% BSA [pH 7.0]) supplemented with protease inhibitor mixture (Sigma-Aldrich). Cells were lysed through four cycles of freeze-thawing and homogenized with 30 strokes within a tight-fitting dounce homogenizer. The homogenate was then ultracentrifugated at 100,000 × g for 1 h at 4°C. Resulting pellets were resuspended in 500 μl PBS, and then homogenized with 10 passes through a 27-gauge needle. Samples were used immediately or stored at −20°C.

BMDM (wild type [WT] or knockout derived where indicated) were harvested following differentiation and plated at 1 × 10⁶ cells/ml in IMDM complete (unless described differently) in petri dishes. Inhibitors where indicated were applied for at least 30 min prior to exposure to MV or other compounds, and were maintained throughout the experiment. Supernatants were tested for TNF-α ELISA (eBioscience, San Diego, CA), IL-12p70 ELISA (eBioscience), or IL-23 ELISA (eBioscience), according to manufacturers’ protocols.

For flow cytometry studies to determine viability and expression of costimulatory receptors, cells were blocked with 1.5% normal goat serum diluted in 0.1% BSA in PBS for 20 min, and then stained with allophycocyanin anti-CD80 (BD Biosciences, San Jose, CA), PE anti-CD83 (BD Biosciences), FITC anti-CD86 (BD Biosciences), allophycocyanin anti-CD86 (BD Biosciences), or PE anti–I-A^b (BD Biosciences) Abs for 40 min. Cells were stained with 1 μg/ml DAPI (Sigma-Aldrich) viability dye. Flow cytometry was performed with a BD Biosciences LSR II, and results were analyzed using FlowJo software. Mean fluorescent intensity (MFI) and population percentages of FITC, PE, and/or allophycocyanin were calculated for DAPI-negative cell populations (i.e., living cells), which were >85% of the total cell population for all results and treatments shown in this work.

RAW264.7 macrophages stably transfected with pNF-κB-MetLuc Vector (BD Clontech) under G418 selection were plated as 1 × 10⁶ cells/ml DMEM complete for each treatment. Supernatant following treatment was centrifuged at 10,000 × g for 10 min and was then stored at −20°C until used to assay for luciferase activity, according to manufacturer’s protocol (BD Clontech). Luminescence readings from each sample were read with an Orion microplate luminometer (Berthold Detection Systems, Huntsville, AL) in duplicate, and the average was taken using Simplicity version 2.1 software (Berthold Detection Systems). Average readings and SEM were calculated according to the ratio of fold change over nontreated cells at each respective time point of the time course.

For whole-cell culture supernatant studies, 2 × 10⁶ J774A.1 were plated in 1 ml DMEM complete with or without 1 μg/ml LPS for 4 h. Cells were washed once with serum-free DMEM with no additions, and treated with or without 3 mM ATP and 10 μM A23187 for 30 min. Supernatants were centrifugated at 10,000 × g for 30 s, and proteins were precipitated through a TCA/cholic acid procedure (22) before loading onto a 9% SDS-PAGE gel. SDS-PAGE and protein transfer to 0.45-μm polyvinylidene difluoride transfer membrane were performed, as previously described (22).

For Western blot analysis of whole-cell lysates for phosphorylated and total p38 expression, 1 × 10⁶ BMDM were plated with or without 1 μg/ml LPS or 25 μg protein equivalents/ml J774A.1-derived MV for 2, 10, 30, or 60 min. Following the time course, cells were washed once with PBS and then given 100 μl 1% Triton X-100 lysis buffer in the presence of protease inhibition mixture (Sigma-Aldrich) and phosphatase inhibition (Sigma-Aldrich) for 15 min on ice. Lysates were collected and centrifuged at 10,000 × g for 10 min at 4°C before addition of sample buffer and loading onto 11% SDS-PAGE gels.

Western blot was performed using the SNAP-ID, according to manufacturer’s procedures (Millipore, Billerica, MA). Abs for Western blotting included 0.5 μg/ml mouse anti-HMGB1 Ab (Abcam, Cambridge, MA), 6 μg/ml mouse anti–IL-1β Ab (3ZD; National Cancer Institute Biological Resources Branch, Frederick, MD), 0.1 μg/ml rabbit anti–phospho-p38 (Thr¹⁸⁰/Try¹⁸² epitope; Millipore), 1/2000 diluted rabbit anti-p38 (Poly6224; BioLegend, San Diego, CA), 1/1666.7 diluted HRP-conjugated goat anti-mouse IgG (Santa Cruz Biotechnology, Santa Cruz, CA), and 1/1666.7 diluted HRP-conjugated donkey anti-rabbit IgG (BioLegend). Signals were developed using Western blotting luminol reagent (Santa Cruz Biotechnology). Imaging was performed with KODAK Image Station 4000MM and its accompanying KODAK MI SE Software Informer (Carestream Molecular Imaging, New Haven, CT). In some instances, the membrane was stripped for Western blot reprobing with Restore Western blot stripping buffer (Thermo Fisher Scientific), according to manufacturer’s protocols.

MV were labeled with biotin or Cy5 using EZ-Link Sulfo-NHS-SS-Biotin reagent (Thermo Scientific) or monoreactive Cy5 dye (GE Healthcare, Piscataway, NJ), respectively, for 1 h in the 500 μl PBS reconstitution using the protocols supplied by the manufacturers. An additional wash with PBS followed by ultracentrifugation at 100,000 × g for 1 h at 4°C was used to remove nonconjugated biotin or Cy5 reagent. Sizing of MV was performed with BD Biosciences FACSAria with various sized YG beads (Polysciences, Warrington, PA).

For assessing MV association with recipient BMDM, 1 × 10⁶ BMDM were plated on 12-mm poly(d-lysine)-coated coverslips in petri dishes. Cells were allowed to adhere for at least 4 h. Indicated amounts of Cy5-labeled or biotinylated MV were given to the BMDM for varying times. Cells not associated to the coverslips were harvested and processed for FACS analysis of CD86 and biotin/Cy5 label. For some experiments, coverslip-associated cells were further stained with 5-chloromethylfluorescein diacetate as a cytoplasmic counterstain (Invitrogen) and LysoTracker Red to label lysosomal compartments (Invitrogen), according to manufacturer’s protocols. Coverslips were then washed, 2% paraformaldehyde fixed, permeabilized, and blocked with 0.5% saponin, 1.5% normal goat serum, and 1% BSA in PBS for 30 min. Alexa647-conjugated streptavidin (Invitrogen) and FITC anti-CD86 (BD Biosciences) were used at 1/100 dilution for 1 h to visualize biotin and CD86 expression, respectively. After successive washes, the coverslips were stained for nuclei with 1 μg/ml DAPI and then mounted with gelvatol. Confocal microscopy images were taken with an Olympus Fluoview 1000 (Inverted) and accompanying software (Olympus America, Center Valley, PA). Laser excitations and emissions were performed sequentially for DAPI, Cy5/Alexa647, LysoTracker Red, and FITC/5-chloromethylfluorescein diacetate, and background noise was minimized. Differential interference contrast microscopy images for each field of view were also taken. Final images were then directly exported to Adobe Photoshop CS2 (Adobe, San Jose, CA).

Unpaired Student t test or one-way ANOVA analyses were performed using Graph Pad Prism (GraphPad, La Jolla, CA). Values of p were calculated where indicated, and for all statistical studies p < 0.05 was considered as significant.

J774A.1, a murine macrophage cell line that expresses P2X₇ (27), was used to produce MV in response to 3 mM ATP plus 10 μM A23187. Whereas ATP alone was sufficient for generating detectable levels of MV, the addition of the calcium ionophore A23187 produced greater quantities of MV as judged by Bradford assay of collected material (data not shown). A23187 alone did not induce MV release. These results suggested that MV release is P2X₇ dependent, and enhanced by A23187. To test this, J774A.1 cells were treated with P2X₇ inhibitor A 740003 or A 438079 prior to generating MV. To confirm that the inhibitors blocked P2X₇ activity, we measured Yo-Pro-1 uptake, a common measure of P2X₇-induced pore formation (28). Whereas treatment with 3 mM ATP alone was sufficient to induce Yo-Pro-1 uptake, cotreatment with 10 μM A23187 and 3 mM ATP greatly increased the signal (Fig. 1A). Furthermore, Yo-Pro-1 uptake by cells exposed either to 3 mM ATP or to 3 mM ATP plus 10 μM A23187 was blocked by pretreatment with A 740003 inhibitor. Both P2X₇ inhibitors either completely abolished or significantly decreased MV shedding, as determined through protein concentration determination by Bradford assay (Fig. 1B). These results demonstrate that MV shedding induced by ATP is dependent on P2X₇ activity, as others have shown (2, 3, 11, 12). MV isolated from culture supernatants by ultracentrifugation were analyzed by flow cytometry and range in size from 0.5 to 1 μm (Supplemental Fig. 1A).

To study the effect of MV on macrophage function, purified MV were incubated with BMDM for 18 h and supernatants were analyzed by ELISA for TNF-α. MV induced TNF-α secretion in a dose-dependent fashion, with the highest dose of MV (75 μg protein equivalents) stimulating more TNF-α secretion than LPS (Fig. 1C, 1D). Importantly, an equivalent amount of cell homogenate from J774A.1 cells did not induce TNF-α secretion at significant levels. MV-depleted ultracentrifugation supernatant from MV generation also did not induce TNF-α at significant levels.

CD86 surface expression was also upregulated in a dose-dependent fashion following exposure to MV (Fig. 1E). Importantly, cell homogenates and MV-depleted supernatants did not increase CD86 levels compared with nontreated control BMDM (Fig. 1E). To exclude the possibility that CD86 upregulation reflected a passive uptake of the protein from MV, BMDM were pretreated with brefeldin A. This significantly diminished CD86 upregulation induced by MV, suggesting that observed increase of CD86 was due to transport of endogenously synthesized CD86 to the cell surface (Fig. 1F). Similar results of CD86 upregulation were also observed for BMDM treated with MV generated from BMDM, human monocyte cell line THP-1, the murine splenic DC line D2SC-1, and the murine fetal skin-derived DC line FSDC (Fig. 1G).

Exposure of cells to MV also upregulated the expression of other markers of activation, specifically CD80, CD83, and I-A^b (Fig. 2A). We next characterized the potential for MV to elicit the production of cytokines known to have roles in polarizing Th1 and Th17 responses. In contrast to TNF-α, however, no increase in IL-12p70 or IL-23 was observed following treatment (Fig. 2B). These data demonstrate that P2X₇-induced MV contain a stimulatory activity that can activate macrophages and are not present in equivalent amounts of homogenates of the cells from which the MV derive. This also suggests that MV primarily stimulate innate immune responses in myeloid cells as opposed to directly influencing adaptive responses.

To examine their interaction with BMDM, MV were first biotinylated. After washing and repelleting, biotinylated MV were incubated for varying times with BMDM. Alexa647-labeled streptavidin was then added to visualize the interaction. When added to BMDM, MV bound to the cell surface within 30 min, with little increase seen over time (Fig. 3A, 3B). In contrast, CD86 expression increased steadily, reaching an observed maximum at 18 h. Similar kinetics were seen with TNF-α secretion (Fig. 3C). Confocal microscopy demonstrated that CD86 and MV were not colocalized at the cell surface; additionally, by 4 h some MV congregated within lysosomes, whereas most of the MV remained at the plasma membrane (Supplemental Fig. 2).

Damaged cells can release a number of compounds (such as damage-associated molecular patterns) that can activate immune cells through TLR engagement. In particular, HMGB1 (29), hyaluronic acid (30), and S100A8/S100A9 complex (31), among others, activate TLR4 for cellular activation. To examine whether MV activate through a TLR4-dependent pathway, BMDM were generated from WT and TLR4^−/− mice. TNF-α induction by MV was reduced in TLR4^−/− BMDM, which produced 35–81% less TNF-α than WT BMDM; however, the decrease was not statistically significant (Fig. 4A). TLR4^−/− BMDM were significantly impaired in CD86 upregulation when treated with MV compared with WT BMDM (Fig. 4B). In contrast, CD86 upregulation induced by the TLR4 agonist MPLA was completely dependent on TLR4, whereas poly(I:C) stimulation was shown to be TLR4 independent, as expected. These data suggest that MV contain a stimulatory component that activates BMDM via TLR4, but that additional means of activation that are not dependent on TLR4 may also be present.

Given the partial TLR4 response from macrophage-derived MV, HMGB1 could mediate MV-induced activation. HMGB1 can activate monocytes and macrophages (32) through TLR2, TLR4, and RAGE (29). Also, HMGB1 can be released from activated monocytes and macrophage (33) or necrotic cells (34) and is expressed within secretory lysosomes that are released from monocytes following stimulation with ATP (35). Furthermore, unlike hyaluronic acid and S100A8/S100A9 complexes, which require prior priming through agents like IFN-γ or LPS (36–38), HMGB1 can be passively released through cellular damage in nonprimed cells (39). Indeed, HMGB1 was detected in the supernatants of ATP/A23187-treated J774A.1 cells and also in purified MV, but not in supernatants from untreated or LPS-only–treated cells (Fig. 5A, 5B). LPS treatment before exposure to ATP/A23187 increased secretion of HMGB1 (Fig. 5A).

To assess whether MV-induced activation requires RAGE, BMDM from RAGE-deficient mice were generated. We determined that there were no significant differences in TNF-α release between MV-treated WT and RAGE^−/− BMDM (Fig. 5C) following stimulation with either LPS or MV. To determine more broadly whether HMGB1 is a stimulatory agent in MV-induced activation, BMDM were incubated with MV in the presence of soluble RAGE, which through competitive inhibition should block the binding of HMGB1 to all of its potential receptors. Soluble RAGE coincubated with MV did not diminish MV-mediated TNF-α release (Fig. 5C). Additionally, there were no significant differences in CD86 upregulation between WT and RAGE^−/− BMDM, however, demonstrating that RAGE does not participate in MV-induced activation (Fig. 5D). Furthermore, soluble RAGE did not block MV-induced CD86 upregulation even at high concentrations, suggesting that HMGB1 does not mediate MV activation (Fig. 5E).

NF-κB and p38 MAPK activation pathways are commonly initiated through TLR engagement (40). Indeed, in BMDM, p38 phosphorylation was induced within minutes of exposure to MV, and then declined over time, similar to the response to LPS (Fig. 6A). To address whether p38 MAPK blockade was sufficient to diminish MV-mediated CD86 upregulation, BMDM were preincubated with a titration of the phosphorylated p38 inhibitor SB203580 before exposure to MV. In the presence of the drug, CD86 upregulation was strongly inhibited, supporting a role for p38 MAPK in MV-mediated activation (Fig. 6B). In contrast, LPS-induced CD86 expression increased in the presence of 1 and 10 μM inhibitor, with significant inhibition only observed at 50 μM, suggesting differences in the two signaling pathways (data not shown).

We also tested whether MV were able to activate NF-κB, using a RAW264.7 reporter cell line (Fig. 6C). The kinetics of the response induced by MV were similar to LPS, with activation evident after 4 and 18 h, but not after 1 h. The magnitude of the response was less with MV than with LPS, however, and declined by 18 h posttreatment.

Protein kinase A (PKA) and PKC can influence p38 and NF-κB in immune cells (40). PKC is largely modulated with intracellular calcium or diacylglycerol (41), whereas PKA is cAMP controlled (42). To indirectly inhibit PKC, BMDM were treated with 30 μM BAPTA-AM to chelate intracellular calcium. Treatment did not impair CD86 upregulation in response to MV or LPS (Fig. 7A). Direct PKC inhibition through use of 50 μM BIM significantly decreased MV-mediated CD86 upregulation (Fig. 7B). In contrast, PKC inhibition enhanced LPS-mediated CD86 upregulation. MDL-12330A was used to inhibit adenylate cyclase, which generates cAMP, resulting in a significant decrease in expression from MV, but not LPS-treated BMDM (Fig. 7C). Inhibition of PKA using 10 μM H-89 HCl also resulted in a significant decrease (Fig. 7D). These results support that MV-mediated activation requires PKA and PKC pathways, which is in contrast to the observed activation response induced by LPS.

To characterize the stimulatory agent(s) from MV, biochemical fractionation was used to separate lipids and proteins, as described in 1Materials and Methods. Only the lipid fraction, not protein, significantly activated BMDM, as measured by CD86 upregulation (Fig. 8A). Lipids were further separated according to polarity. The phospholipid fraction provided significant upregulation of CD86 as compared with nontreated control. Only minimal activity was recovered from the flow through, the neutral lipids fraction, or the glycolipid/sulpholipid fraction (Fig. 8B). Mock elution from the column for a phospholipid fraction also did not recover any stimulatory material, demonstrating that the columns themselves did not contain contaminants that could activate BMDM.

To test the TLR4 dependence of the phospholipid-containing fraction from the MV, TLR4^−/− or WT BMDM were given equal amounts of phospholipid fraction. In agreement with the partial TLR4 dependence for MV-mediated activation, as observed in Fig. 4, the phospholipid fraction from MV had a significant difference in activation between the TLR4^−/− and WT BMDM. Additionally, phospholipid fraction activation of TLR4^−/− BMDM did not significantly differ from nontreated BMDM. This would at least partially confirm that the TLR4 agonist from the MV came in the form of a phospholipid.

These results led us to hypothesize that purinergic receptor activation in MV producer cells exposed to ATP might generate stimulatory phospholipids that are packaged into MV. This could explain why MV activated BMDM more strongly than the equivalent amount of cell lysate, as described earlier. P2X₇ activation, which is required for generating ATP-induced MV, can activate lipid-modifying enzymes, such as calcium-independent phospholipase A₂ (iPLA₂) (43), phospholipase D (PLD) (44), and calcium-dependent phospholipase A₂ (cPLA₂) (43). To test whether PLD, iPLA₂, or cPLA₂ was required in generating stimulatory MV, producer cells were treated with either PLD inhibitors (50 μM CAY10593 for PLD1 and 50 μM CAY10594 for PLD2), iPLA₂ inhibitor BEL (10 μM BEL), or cPLA₂ inhibitor (10 μM AACOF3) 30 min prior to and during 3 mM ATP and 10 μM A23187 administration for MV induction. When equivalent amounts of these respective MV populations were applied to recipient BMDM, MV from PLD1- and PLD2-inhibited J774A.1 producer cells were unable to stimulate CD86 upregulation (Fig. 9A). PLD inhibitors used to inhibit PLD during MV generation did not directly impact MV-mediated CD86 upregulation of recipient BMDM as inclusion of MV from PLD1- and PLD2-inhibited J774A.1 with MV from mock-treated J774A.1 did not result in any significant differences from BMDM treatment with MV from mock-treated J774A.1 (Fig. 9B). Furthermore, MV from mock-treated or from PLD1- or PLD2-inhibited cells did not decrease phosphatidic acid levels of treated J774A.1 recipient cells relative to nontreated J774A.1 (data not shown). In contrast to PLD, inhibition of iPLA₂ (Fig. 9C) or cPLA₂ (Fig. 9D) in MV producer cells did not decrease the stimulatory capacity of the MV generated, which were at least as potent as those from producer cells not pretreated with inhibitors. It should be noted that MV were obtained from producer cells treated with these inhibitors in amounts approximately equal to mock-treated producer cells, suggesting that the stimulatory phospholipid is not required for MV structural integrity or secretion from cells. Furthermore, PLD1, PLD2, iPLA₂, and cPLA₂ inhibitors were not toxic at the tried concentrations; cell viability was >85% (data not shown).

We lastly considered whether lysophosphatidic acid (LPA) could be identified as a potential stimulatory phospholipid from the MV. LPA is a product of P2X₇ activity (45), and it can activate macrophages to promote cytokine production (46) and cAMP synthesis (47). Pretreatment of BMDM with Ki16425, a lysophosphatidic acid receptor inhibitor selective for LPA₁ and LPA₃ receptors, did not diminish CD86 upregulation by MV (Fig. 9E). This suggests LPA does not participate in MV-induced activation of macrophages, although it does not exclude the possibility that LPA could mediate effects through other receptors.

MV released by cells can potently influence immune responses in a number of ways. Many cell types release exosomes constitutively, and, depending on the cell of origin, may transfer Ags or other cargo to DCs that can initiate immune responses. A specialized type of MV release exists for myeloid cells that express P2X₇ receptors, which, when exposed to receptor agonists, such as ATP, shed MV from the cell surface as well as release them from intracellular stores. Characterizing MV generation induced by P2X₇ activation on macrophages is important for understanding inflammatory processes, because tissue damage has been shown to release intracellular constituents like ATP into an environment containing large numbers of these cells (48–51).

When macrophages are primed by exposure to TLR agonists, cytokines including IL-1β and TNF-α are synthesized and may be released from the cell following appropriate stimulation by a secondary signal (19). For IL-1β release, P2X₇ engagement is followed by cleavage of pro–IL-1β into the bioactive form by caspase-1 in a NLRP3-dependent process. MV have been demonstrated to contain mature IL-1β and were first characterized as surface-derived vesicles (3, 14), but later as exosomes (11). TNF-α–containing exosomes have been reported to be secreted from human melanoma cells (52). We have detected TNF-α within P2X₇-induced MV from primed macrophages; however, TNF-α was also detected from cell culture supernatants devoid of MV (data not shown). Nevertheless, only primed macrophages would release proinflammatory cytokines, whose synthesis primarily depends on NF-κB activation (40), whether in MV or as soluble proteins.

Our study instead focuses on characterizing the biological effects of MV released by nonprimed, primary macrophages, a condition that may be seen during sterile inflammation. As reported by others, MV are released following P2X₇ engagement in nonprimed macrophages (12), yet, to the best of our knowledge, have not been studied for their ability to activate macrophages. Whereas 3 mM ATP alone is sufficient for P2X₇ activation, we observe enhanced P2X₇ activity when coadministering 10 μM A23187 with 3 mM ATP. A23187 with 0.3 mM ATP was not able to induce P2X₇ activity; thus, in this manner, A23187 is serving in some other function than for the autocrine release of endogenously stored ATP for P2X₇ activation. We found that P2X₇-induced MV from nonprimed myeloid cells can induce expression of CD86, CD80, CD83, and class II MHC (Fig. 2A), and also induce secretion of TNF-α from primary macrophages (Fig. 1A, 1B). We further characterized the contents of these MV and found that phospholipids were responsible for stimulating macrophage activation in a TLR4-dependent process (Fig. 8B). As separated phospholipids from MV activate macrophages, we can exclude a requirement for interaction of intact MV in this process. That being said, based on our observation that MV-mediated activation was only partially TLR4 dependent, we hypothesize that MV must contain components in addition to phospholipids that activate macrophages independently of TLR4, and that these activities may be contingent upon intact MV delivery into intracellular departments.

Stimulation of primed macrophages by P2X₇ yields MV that are heterogeneous, consisting of exosomes (11, 12) and surface-derived vesicles (3, 14), and possibly including secreted lysosomes (7–9). The MV described in our study are predominantly 0.5–1.0 μm in diameter (Supplemental Fig. 1A), distinguishing them from larger 1- to 4-μm apoptotic blebs (53). We observed relatively few smaller MV that would be characteristic of exosomes as analyzed by electron microscopy (data not shown), and believe that our preparations were devoid of them for several reasons. Whereas our MV were obtained from ultracentrifugation at 100,000 × g, we also found that material obtained from 10,000 × g centrifugation exerts equivalent ability to activate macrophage (data not shown). Exosomes do not pellet at the low speed as they are only 50–100 nm in diameter. Thus, MV-induced macrophage activation seems to be exosome independent. Furthermore, the release of class II MHC-containing exosomes from macrophages requires apoptosis-associated speck-like protein containing a CARD (ASC)/NLRP3 inflammasome (12). Nonprimed myeloid cells, such as we have used, typically do not express high levels of NLRP3 (54), and thus, would not be expected to release class II MHC⁺ exosomes efficiently. In addition, we have purified MV from D2SC-1 cells, a murine splenic DC-derived cell line that lacks ASC (data not shown), and found that these MV are potent stimulators of macrophage activation (Fig. 1G). In this way, D2SC-1 act similarly to RAW264.7, which also lack ASC (27), yet also shed MV in response to P2X₇ stimulation (55). These results support the conclusion that MV distinct from previously characterized class II MHC⁺ exosomes are shed by myeloid cells and stimulate primary macrophages through a TLR4-dependent process involving recognition of phospholipids contained within MV.

Endogenous phospholipids can activate macrophage through TLR4 activities. Recently, it was demonstrated that oxidized low density lipoprotein (oxLDL), which binds to the scavenger receptor B family member CD36, can promote sterile inflammation through activation of TLR4/6 heterodimer on macrophages (56). Both cell death (57, 58) and foam-cell formation (59) have also been shown to be induced by oxLDL through TLR4 in macrophages. Furthermore, oxidized phospholipids from minimally modified low density lipoprotein, which contain essentially the same phospholipids as oxLDL, stimulate macrophage reactive oxygen species generation (60), ERK activation (61), membrane spreading (62), and inhibition of phagocytic uptake of apoptotic bodies (62), through a partially TLR4-dependent pathway. Whether phospholipids in P2X₇-induced MV are structurally similar to those in oxLDL or minimally modified low density lipoprotein will be addressed in future studies. Our study suggests that the stimulatory phospholipid is not lysophosphatidic acid (Fig. 9E). We were also unable to induce BMDM activation with commercially available lysophosphatidic acid, phosphatidic acid, or phosphatidylserine (data not shown).

Importantly, P2X₇-induced MV from PLD1- and PLD2-inhibited MV producer cells were unable to activate macrophages (Fig. 9A); at the same time, MV yields were equivalent between drug-treated and nondrug-treated cells (data not shown). This dissociates MV formation from incorporation of the stimulatory phospholipid into vesicles, and suggests that generation of a bioactive phospholipid results from P2X₇ activation, leading to downstream PLD activation. Whereas PLD (44) and iPLA₂ (43) are activated following P2X₇ activation (data not shown), blocking PLD activity, but not iPLA₂, impaired the MV-activating capacity (Fig. 9). These results may explain why cell homogenates of producer cells were unable to stimulate macrophage activation (Fig. 1B, 1C), because PLD was not activated. Activated PLD produces phosphatidic acid to coordinate ADP-ribosylation factor-6, a known regulator of exocytosis, to sites of potential exocytosis (63). Future studies will be done to understand the specific PLD activities that enable stimulatory phospholipid loading into P2X₇-induced MV.

It has been suggested that host cell-derived stimulators of TLR activity might contain microbial contaminants introduced during biochemical purification, a hypothesis described in detail recently in a thought-provoking review (64). Based on this, we considered whether MV preparations might contain endotoxin. When tested by Limulus amebocyte lysate assay, endotoxin was present within stimulatory MV preparations at low levels, typically ∼0.25 EU/ml (data not shown). However, the same amount of endotoxin is observed in nonstimulatory MV from PLD-inhibited cells (data not shown), indicating that these low levels cannot explain the stimulation we observe. Furthermore, treatment of BMDM with equivalent amounts of LPS to that found in MV (∼50 pg/ml LPS for 25 μg protein equivalents of MV) did not induce significant TNF-α release or CD86 upregulation (data not shown). It should also be noted that in addition to testing nonstimulatory MV, we observed that cell homogenates and ultracentrifugation supernatants were also devoid of BMDM-stimulating activity.

Our study suggests that MV derived from macrophages in an environment where there is tissue damage without infection could have potent biological activities that may further drive inflammation. In tumors and other settings with significant necrosis, infiltrating macrophages expressing P2X₇ would be exposed to elevated levels of extracellular ATP, as previously shown in tumors (65–67). Macrophage-produced MV would then bind to adjacent cells, including macrophages and DCs, leading to their activation and resulting in secretion of TNF-α and potentially other proinflammatory mediators. The most novel aspect of this work, implicating phospholipids from MV as the stimulatory component, may be consistent with recent work implicating phospholipids in inflammation present in atherosclerotic lesions, as discussed above. In cancer and in atherosclerosis, as well as potentially in other pathological settings, there are strong indications that sterile inflammation plays a role in disease processes through largely unknown mechanisms. Innate responses by macrophages in inflammatory diseases may in part be explained by release of stimulatory phospholipids in MV at sites of tissue damage.

We thank numerous laboratories for reagents and bone marrow, as indicated in the text; BD Biosciences FACSAria help from Dewayne H. Falkner (University of Pittsburgh); luminometer assistance from Judong Lee from Lawrence P. Kane’s laboratory (University of Pittsburgh) and Sudesh Pawaria from Robert J. Binder’s laboratory. We also thank Michelle E. Heid, Jessica Chu, Peter A. Keyel, and Chengqun Sun from our laboratory for helpful discussions and cell culture assistance and Judson M. Englert from Tim D. Oury’s laboratory for discussions and reagents. We also thank Laura E. Kropp for critical reading of the manuscript.

Disclosures The authors have no financial conflicts of interest.