ATP redirects cytokine trafficking and promotes novel membrane TNF signaling via microvesicles

Abstract

Cellular stress or injury induces release of endogenous danger signals such as ATP, which plays a central role in activating immune cells. ATP is essential for the release of nonclassically secreted cytokines such as IL-1β but, paradoxically, has been reported to inhibit the release of classically secreted cytokines such as TNF. Here, we reveal that ATP does switch off soluble TNF (17 kDa) release from LPS-treated macrophages, but rather than inhibiting the entire TNF secretion, ATP packages membrane TNF (26 kDa) within microvesicles (MVs). Secretion of membrane TNF within MVs bypasses the conventional endoplasmic reticulum- and Golgi transport-dependent pathway and is mediated by acid sphingomyelinase. These membrane TNF-carrying MVs are biologically more potent than soluble TNF in vivo, producing significant lung inflammation in mice. Thus, ATP critically alters TNF trafficking and secretion from macrophages, inducing novel unconventional membrane TNF signaling via MVs without direct cell-to-cell contact. These data have crucial implications for this key cytokine, particularly when therapeutically targeting TNF in acute inflammatory diseases.-Soni, S., O'Dea, K. P., Tan, Y. Y., Cho, K., Abe, E., Romano, R., Cui, J., Ma, D., Sarathchandra, P., Wilson, M. R., Takata, M. ATP redirects cytokine trafficking and promotes novel membrane TNF signaling via microvesicles.

Keywords: cellular communication; danger signals; extracellular vesicles; protein signalling.

Conflict of interest statement

This work was supported by grants from the Medical Research Council and British Journal of Anaesthesia (P54008) and the Chelsea and Westminster Health Charity. The authors declare no conflicts of interest.

Figures

Figure 1

ATP inhibits soluble TNF release. A) RAW 264.7 macrophages exposed either LPS alone (1 µg/ml, 2 h) or 2-hit injury model with LPS (1 µg/ml, 1 h) followed by ATP (3 mM, 1 h). LPS induced soluble TNF secretion into cell- and MV-depleted supernatants, which was almost entirely abolished by ATP treatment (left, n = 9–10). ATP reduced both surface [middle, represented as mean fluorescence intensity (MFI), n = 5–6] and total cellular TNF expression (right, n = 5–6). B) Confocal images clearly showed disappearance, rather than accumulation, of cellular TNF from the RAW macrophages upon ATP exposure (n = 3). C) To confirm these results in primary cells, we exposed mouse BMDMs to LPS (1 µg/ml, 1 h) followed by ATP (3 mM, 15 min). Similarly, ATP inhibited soluble TNF secretion (left, n = 6) and reduced total cellular TNF expression (middle and right, n = 6). D) Confocal images of BMDMs illustrating the reduction of cellular TNF expression following ATP treatment (n = 3). Cells were also checked for TNF expression under nonstimulated conditions (i.e., in PBS or medium) and, as expected, TNF was not detected in this sterile or noninflamed environment (Supplemental Fig. S1G). Parametric or nonparametric data are displayed as means ± sd or box-whisker plots showing the median, IQR, and minimum or maximum values, respectively. Scale bars, 10 µm. **P < 0.01, ***P < 0.001.

Figure 2

ATP preferentially packages membrane TNF within shed MVs. A) Scanning (left; scale bar, 1 µm) and transmission (right; scale bar, 0.2 µm) EM of RAW cells illustrates membrane blebbing and MV formation in response to ATP. B) RAW cell–derived MVs, identified as CD45+ or CD11b+ particles by flow cytometry (left and middle), significantly increased in response to ATP (right; n = 6–10). C) These MVs contained TNF, accounting for the above missing TNF from the cells but in the form of 26 kDa transmembrane pro-TNF isoform rather than 17 kDa soluble mature TNF (left and middle; n = 4–8). ATP packaged substantial amounts of pro-TNF within MVs (right; n = 4–8). D) To confirm these results in primary cells, we exposed mouse BMDMs to LPS (1 µg/ml, 1 h) followed by ATP (3 mM, 15 min). Similarly, ATP induced significant release of MVs (left; n = 6), which contained substantive amounts of pro-TNF (right; n = 6). E) Confocal microscopy illustrates that these BMDM-derived MVs, identified as CD11b+ (green, top left) particles negative for nuclear materials (DAPI−, top right), actually contained TNF (red, bottom left, colocalization shown in the bottom right combined image). Scale bars, 1 µm. F) Immune EM demonstrates transfer of pro-TNF (black dots, red arrows) from BMDMs to MVs during their formation, and these pro-TNF molecules gradually localized to MV membrane surface (hence, membrane TNF) (right). Scale bars, 0.2 µm. Parametric or nonparametric data are displayed as means ± sd or box-whisker plots showing the median, IQR, and minimum or maximum values, respectively. **P < 0.01, ***P < 0.001.

Figure 3

In vivo, ATP selectively packages membrane within alveolar macrophage–derived MVs while simultaneously switching off soluble TNF release. A) Mice were exposed to an in vivo 2-hit injury model: intratracheal LPS (50 ng, in 25 µl saline) instilled into murine lungs for 30 min followed by 5 mM ATP (in saline, 25 µl) for 30 min. LPS control mice consisted of intratracheal LPS followed by intratracheal saline. LPS induced soluble TNF secretion into alveoli (measured in cell- or MV-depleted BALF samples), which was substantively decreased by ATP (left; n = 5–6). ATP also caused a reduction in total cellular TNF content (measured by flow cytometry) in LPS-stimulated alveolar macrophages (middle and right), confirming that TNF was not internalized or did not accumulate within cells but disappeared from the cells (n = 5). B) In addition to inhibiting soluble TNF release, ATP significantly stimulated the release of MVs from alveolar macrophages, identified as CD45+CD11c+ particles by flow cytometry (n = 5). C) As in vitro, in vivo–derived MVs contained almost exclusively membrane TNF. Consequently, in vivo ATP preferentially packaged membrane TNF within MVs compared with LPS control MVs (n = 4). Parametric and nonparametric data are displayed as means ± sd or box-whisker plots showing the median, IQR, and minimum or maximum values, respectively. **P < 0.01.

Figure 4

ATP secretes membrane TNF within MVs via an unconventional, ER- and Golgi-independent pathway. A) RAW macrophages were pretreated with BFA (5 µg/ml), an inhibitor of classic cytokine secretory pathway, in order to block TNF secretion in our in vitro 2-hit model. As expected, BFA completely abolished LPS-induced, soluble-TNF release. However, even in the presence of BFA (which blocked ER- and Golgi-dependent protein transport), ATP caused a dramatic reduction in total cellular TNF content, demonstrating that TNF is still secreted in some way. TNF expression in BFA-treated cells incubated in PBS was undetectable, confirming that the accumulation of TNF within macrophages was because of LPS rather than an effect of BFA itself. B) MVs released from BFA-treated RAW cells were identified via flow cytometry as CD45+ and CD11b+ (left), and ATP caused a significant increase in MV production in these cells (right; n = 4). C) These MVs contained a substantial amount of membrane TNF (left and right; n = 4). Data are displayed as means ± sd. **P < 0.01, ***P < 0.001.

Figure 5

Switch of TNF secretion to nonclassic pathway is mediated by ASM. A) ASM activity was increased in RAW macrophages following ATP treatment (n = 3). B) LPS-primed RAW cells were treated with recombinant sphingomyelinase, which caused a marked increase in MV release compared with LPS alone or LPS- or ATP-treated RAW cells (n = 4–9). C) These ASM-generated MVs contain substantial amounts of membrane TNF compared with ATP-generated MVs (n = 4–7). D) Knockdown of ASM in RAW macrophages using siRNA prevented packaging of membrane TNF within MVs (n = 4). E) To confirm these results in primary cells, BMDMs were pretreated with the ASM inhibitor desipramine (25 mM) prior to exposure to LPS or LPS and ATP. Desipramine prevented ATP-induced reduction in soluble TNF release (left; n = 4), markedly attenuated ATP-induced MV release (middle; n = 4), and significantly reduced packaging of membrane TNF within MVs (right; n = 4–5). Parametric and nonparametric data are displayed as means ± sd or box-whisker plots showing the median, IQR, and minimum or maximum values respectively. NS, not significant. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 6

Membrane TNF is stable within macrophage-derived MVs. A) Membrane TNF is harbored in a stable, protected environment within MVs and was not spontaneously cleaved into the mature 17 kDa isoform despite being incubated at 37°C for 1 h. (n = 4). B) ATP also packages TACE into RAW cell–derived MVs (top) and in vivo–derived MVs (bottom). Confocal images of BMDM-derived MVs demonstrate the presence of TACE within BMDM-derived MVs. Scale bars, 1 μm. Data are displayed as means ± sd.

Figure 7

MVs provide a vehicle for membrane TNF signaling, causing significant inflammation in vivo. Membrane TNF containing BMDM-MVs from WT mice (TNF+ve MVs) were instilled intratracheally into the lungs of mice. Their biologic effects were compared with intratracheal instillation of BMDM-MVs taken from TNF−/−mice (TNF−ve MVs) and intratracheal high-dose recombinant (i.e., soluble) TNF (50 µl 100 ng/ml). TNF+ve MVs caused significant increases in infiltrating inflammatory monocytes in lung tissue (flow cytometry) (A), ICAM1 expression on alveolar epithelial cells (flow cytometry) (B), BALF CXCL1 levels (by ELISA) (C), and BALF protein (by Qubit assay) (D) compared with TNF−ve MVs and intratracheal soluble TNF. Data are displayed as means ± sd (n = 4). *P < 0.05, **P < 0.01.