In acute inflammation, extracellular ATP activates P2X7 ion channel receptors (P2X7R) on M1 polarized macrophages to release pro‐inflammatory IL‐1β through activation of the caspase‐1/nucleotide‐binding domain and leucine‐rich repeat receptor containing pyrin domain 3 (NLRP3) inflammasome. In contrast, M2 polarized macrophages are critical to the resolution of inflammation but neither actions of P2X7R on these macrophages nor mechanisms by which macrophages switch from pro‐inflammatory to anti‐inflammatory phenotypes are known. Here, we investigated extracellular ATP signalling over a dynamic macrophage polarity gradient from M1 through M2 phenotypes. In macrophages polarized towards, but not at, M2 phenotype, in which intracellular IL‐1β remains high and the inflammasome is intact, P2X7R activation selectively uncouples to the NLRP3‐inflammasome activation but not to upstream ion channel activation. In these intermediate M1/M2 polarized macrophages, extracellular ATP now acts through its pyrophosphate chains, independently of other purine receptors, to inhibit IL‐1β release by other stimuli through two independent mechanisms: inhibition of ROS production and trapping of the inflammasome complex through intracellular clustering of actin filaments.
Macrophages are critical to both the innate and adaptive immune response; they must develop and respond to rapid changes in the microenvironment. Inflammation resulting from pathogens or tissue damage activates resident macrophages to initiate or increase production of pro‐inflammatory cytokines and other inflammatory mediators. However, macrophages are equally critical in the resolution of inflammation by producing anti‐inflammatory cytokines and chemokines and by increased phagocytic activity. On the basis of Th1/Th2 polarization concepts (Romagnani, 2000), phenotypically polarized macrophages are now generally termed pro‐inflammatory M1 or classically activated, and anti‐inflammatory M2 or alternatively activated (Gordon, 2003; Martinez et al, 2008). Experimentally, macrophages in vitro are polarized to the M1 state by treatment with IFN‐γ and inducers of TNF‐α, such as lipopolysaccharide (LPS) or other bacterial products (Ehrt et al, 2001; Gordon, 2003). M1 macrophages induce synthesis and upregulation of several pro‐inflammatory cytokines and chemokines, key among these are TNF‐α, IL‐12, IL‐6, CCL2 and IL‐1β, as well as increased production of reactive oxygen species (ROS) and nitrogen intermediates (Gordon, 2003; Martinez et al, 2008). At the other extreme, macrophages are polarized to the M2 state by stimuli such as IL‐4, IL‐13, IL‐10 or glucocorticoid hormones (Gordon, 2003; Martinez et al, 2008). M2 macrophages upregulate scavenger, mannose and galactose receptors, IL‐1 receptor antagonist and downregulate IL‐1β and other pro‐inflammatory cytokines (Gordon, 2003; Scotton et al, 2005; Martinez et al, 2006, 2008). Very high levels of IL‐1β are present in M1 polarized macrophages because of activation of the NF‐κB and MAPK cascades (Dinarello, 1996) but no IL‐1β protein is found in M2 polarized macrophages (Scotton et al, 2005; Martinez et al, 2006).
IL‐1β processing and release is a tightly regulated process involving a multiprotein complex, the capase‐1 inflammasome (Martinon et al, 2002). IL‐1β is synthesized as a 34 kD precursor protein that is cleaved to its biologically active 17 kD form by the protease caspase‐1. Inactive caspase‐1 is constitutively present in both M1 and M2 macrophages and is activated in M1 macrophages by self‐cleavage that occurs in the inflammasome complex. Distinct caspase‐1 inflammasomes have been distinguished based on their protein composition and means of activation (Mariathasan et al, 2004, 2006; Kanneganti et al, 2006; Martinon et al, 2006). The nucleotide‐binding domain and leucine‐rich repeat receptor containing pyrin domain 3 (NLRP3) inflammasome can be activated by specific microbial motifs (pathogen‐associated molecular patterns, PAMPs), by Escherichia coli and Staphylococcus aureus bacteria, by microbial toxins such as nigericin and maitotoxin (MTX), by uric acid crystals and by extracellular ATP acting through the ATP‐gated plasma membrane ion channel, the P2X7 receptor (P2X7R) (Mariathasan et al, 2004, 2006; Martinon et al, 2006; Pétrilli et al, 2007). P2X7R activation by ATP released at sites of inflammation is currently the only well‐established physiological stimulator of the NLRP3 inflammasome (Ferrari et al, 2006; Mariathasan et al, 2004, 2006) and, as such, it has become an attractive target for development of new anti‐inflammatory drugs (Pelegrin, 2008). P2X7R gene ‐deleted mice show anti‐inflammatory phenotypes in models of neuropathic and inflammatory pain (Labasi et al, 2002; Chessell et al, 2005) and a recently developed P2X7R antagonist has shown reduced joint swelling, tenderness and pain perception in Phase II clinical trails in patients with rheumatoid arthritis (McInnes et al, 2007). The P2X7R is a highly unusual ion channel predominantly expressed in macrophage and microglia; its activation opens a cationic channel with high permeability to calcium but it also forms within seconds a larger pore permeable to small molecules up to 900 Da (Ferrari et al, 2006; Pelegrin and Surprenant, 2006; Pelegrin et al, 2008). All downstream sequelae of P2X7R activation except for the initial cation channel opening require the presence of the intracellular C‐terminus, which is key to the formation of the P2X7R‐multiprotein complex (Ferrari et al, 2006).
This study was prompted by a surprising observation we made while comparing E. coli and P2X7R‐induced activation of the NLRP3 inflammasome and IL‐1β release in M1 polarized macrophages. We found that, though E. coli‐induced release of IL‐1β remained constant, ATP‐evoked IL‐1β release ceased over time in spite of continued high levels of intracellular IL‐1β and without decreased functional P2X7R as assayed by upstream ion channel signalling. We found that unexpected changes in the polarization state of the macrophages were causing a P2X7R‐specific uncoupling to the caspase‐1 cascade. We therefore developed an in vitro model of a likely macrophage polarization gradient from extreme M1 through to extreme M2 and examined the actions of different NLRP3‐inflammasome activators (ATP, E. coli, MTX and nigericin) over this polarization gradient, as well as in acute and chronically P2X7R ‐deleted M1 polarized macrophages. This study has identified a new mechanism by which extracellular ATP produces physiological inhibition of IL‐1β release that may be key to switching a macrophage from an inducer of inflammation towards its anti‐inflammatory phenotype in the resolution phase of infection or inflammation.
Characterization of a dynamic macrophage polarity gradient
We developed an in vitro polarization protocol designed to rapidly mimic (over 4–8 h) a continuum of signals that are likely to drive the macrophage from the classic inflammatory M1 polarity to the anti‐inflammatory M2 state (Supplementary Figure S1A available online). Macrophages from polarities close to M1 (states 1–3) were expanded and tightly fixed to substrate, formed numerous lamellar processes and elongated filopodia, with diffuse F‐actin staining throughout the cytoplasm but distinctly absent from the filopodia (Figure 1A and B). Macrophages from a polarity close to M2 (state 4) presented less lamellar processes but remained expanded and tightly fixed to the substrate, whereas the extreme M2 polarized (alternatively activated) macrophages had the same morphology as untreated cells with F‐actin being tightly compacted close to the nucleus (Figure 1A and B). We quantified these changes in cell morphology by measuring total surface area over all polarizations states and found a gradual progression from state 1 through state 5 (Figure 1C, n=50–70 cells for each polarization state). We also confirmed that our protocols did, indeed, establish a molecular polarization gradient by quantitative RT–PCR (qPCR) for the classical M1 and M2 genes, TNF‐α and mannose receptor C (MRC‐1) (Gordon, 2003; Martinez et al, 2008); TNF‐α gene expression was upregulated by 200‐fold in M1 macrophages and this gradually decreased to near zero in extreme M2 macrophages, whereas MRC‐1 gene expression gradually increased over the gradient to reach a 10‐fold upregulation at extreme M2 polarization (Figure 1D).
P2X7R, but not E. coli, uncouples to inflammasome activation during macrophage polarization
We found that intracellular levels of proIL‐1β remained high in all polarization states (1–4) except, as expected (Gordon, 2003; Martinez et al, 2008), the alternatively activated M2 macrophages (Figure 2A and B). ATP stimulation of P2X7R in M1 activated macrophage and E. coli infection are both well‐characterized pathways known to initiate NLRP3‐inflammasome‐dependent release of mature IL‐1β (Ferrari et al, 2006; Mariathasan et al, 2006; Pétrilli et al, 2007). However, we found striking differences in their ability to release IL‐1β over the polarization gradient. In agreement with numerous studies (Mariathasan et al, 2006; Pelegrin and Surprenant, 2006, 2007; Pétrilli et al, 2007), both P2X7R and E. coli stimulation induced the release of IL‐1β from M1 polarized macrophages but P2X7R‐induced release of IL‐1β was significantly reduced, or absent, in macrophages of intermediate polarity (states 3, 4; Figure 2A and B). Conversely, E. coli induced significantly greater IL‐1β release from these intermediately polarized macrophages (Figure 2B). Moreover, we observed a similar uncoupling of P2X7R from IL‐1β release during traditional long‐term activation of macrophages with LPS/IFN‐γ (Arend et al, 1989). That is, though P2X7R or E. coli‐induced release of IL‐1β peaked at similar times (5–10 h post activation with either LPS/IFN‐γ or E. coli), ATP failed to stimulate IL‐1β release subsequently in spite of maintained high levels of intracellular IL‐1β and in contrast to the continued release of IL‐1β by E. coli (Figure 2C). Initially, we hypothesized a decrease in P2X7R expression was the most likely explanation for the failure of ATP to stimulate IL‐1β release under these conditions, as clearly the NLRP3‐inflammasome cascade known to be activated by E. coli (Pétrilli et al, 2007) was not affected. However, we could not attribute this loss of P2X7R‐mediated release of IL‐1β to loss of functional receptors because (i) P2X7R gene expression was actually downregulated in M1 macrophages and did not change from states 1 to 3, then increased in states 4 and 5 (Figure 1D); (ii) P2X7R protein localization remained constant through all polarization states, in particular the P2X7R protein showed a similar punctate expression along the cell (Figure 2D); (iii) assays for upstream signalling of P2X7R (calcium flux and dye‐uptake) (Pelegrin and Surprenant, 2006) did not change over the polarity gradient (Figure 2E and F). Macrophages obtained from P2X7R gene‐deleted mice were subjected to the same polarization gradient protocols and served as controls; in these macrophages no anti‐P2X7R antibody staining was present (data not shown) and no ATP‐mediated calcium flux or dye‐uptake occurred in M1 macrophages (Figure 2E and F) or in M2 macrophages (data not shown).
From positive to negative regulation of IL‐1β release by ATP during macrophage polarization
We next compared inflammasome activation profiles induced by ATP and by two toxins, MTX and nigericin, which activate the NRLP3‐inflammasome with similar kinetics and mechanisms to P2X7R but which are not dependent on P2X7R (Mariathasan et al, 2006; Pelegrin and Surprenant, 2007) (Supplementary Figure S1B). IL‐1β ELISA assays (which recognize pro and mature forms of the cytokine) showed that ATP‐induced IL‐1β release was reduced by 50–80% in state 3 intermediate polarized macrophages compared with M1 polarized cells (Figures 2B and 3A), whereas MTX‐induced IL‐1β release was the same in macrophages from either polarity state or similar to levels of ATP‐induced release from M1 macrophages (Figure 3A). When ATP and MTX were co‐applied, IL‐1β release was increased in M1 macrophages to the same extent as in the presence of either compound alone (Figure 3A). However, co‐application of MTX and ATP to state 3 polarized macrophages resulted in a 50% reduction of IL‐1β release (Figure 3A). Western blots showed typical release of mature IL‐1β from M1 macrophage in response to ATP, MTX or a combination of MTX and ATP whereas in state 3 polarized macrophages; mature IL‐1β was detected in response to MTX but not ATP stimulation and co‐application significantly decreased release of mature IL‐1β (Supplementary Figure S2A). Strikingly, ATP completely prevented MTX‐induced release of IL‐1β in M1 polarized macrophages after acute blockade of P2X7R with the P2X7R‐selective antagonist, A740003 (Honore et al, 2006) (Figure 3B; Supplementary Figure S2B). ATP similarly inhibited both MTX and nigericin‐induced release of IL‐1β from M1 polarized macrophages obtained from P2X7R‐deleted mice (Figure 3C; Supplementary Figure S2C). ATP also inhibited caspase‐1 activation (Figure 3C, lower panels) and release of the inflammasome adaptor protein, Apoptosis‐associated Speck‐like protein containing a C‐terminal caspase‐activating recruiting domain (ASC, Figure 3D) induced by either MTX or nigericin in M1 polarized P2X7R−/− macrophages. Re‐introduction of P2X7R into P2X7R−/− macrophages through adenoviral delivery reconstituted all features of ATP‐mediated IL‐1β release and caspase‐1 activation from M1 polarized macrophages (Supplementary Figure S1D and E). ATP also failed to induce caspase‐1 activation in M2 wild‐type polarized macrophages (Figure 3E), although upstream signalling through P2X7R (calcium flux, dye‐uptake) was not significantly different than observed in M1 macrophages. Taken together, these results show that the inhibitory action of ATP on these macrophages was due solely to lack of P2X7R downstream signalling and not to alterations in the NLRP3 inflammasome itself. ATP was equally able to inhibit IL‐1β release from the in vitro E. coli model of infection and slow NLRP3 activation (e.g. Figure 2C; Supplementary Figure S1C) in M1 polarized macrophages from P2X7R−/− macrophages (Figure 3F; Supplementary Figure S2D). This inhibition was not because of alterations in binding or phagocytosis of the bacteria by these macrophages in the presence of ATP (Supplementary Figure S2E).
Negative regulation of IL‐1β by ATP is not through purinoreceptors but is due to pyrophosphate actions
We expected the inhibitory actions of ATP, which occurred during macrophage polarization from M1 towards M2 states or which were revealed by acute (selective P2X7R antagonist) or chronic (P2X7R−/− macrophage) blockade of P2X7R, to result from activation of one, or more, purinergic P2X, P2Y or adenosine receptors. However, results shown in Figure 4, Supplementary Figure S3A, and E quickly ruled out involvement of any of these receptors in the inhibition of IL‐1β release by ATP. We used M1 polarized macrophages from P2X7R−/− mice (Figure 4A–E) or from wild‐type mice in the presence of A740003 (data not shown) and examined the actions of ATP and other purines on MTX‐induced IL‐1β release and caspase‐1 activation. First, the ATP concentration–inhibition curve was 20–100‐fold to the right of all known purine and adenosine receptors except for P2X7R; indeed, the IC50 value for ATP inhibition of MTX‐induced IL‐1β release in P2X7R‐depleted macrophage (1.54±0.03 mM, n=3) was not significantly different from the ATP EC50 value to induce release of IL‐1β from wild‐type M1 macrophages (1.65±0.02 mM, n=3) (Figure 4A and E). Second, significant inhibition was observed with ATP>ADP≫AMP but not with adenosine (Figure 4B and C). Third, the non‐selective P2X/P2Y inhibitor, suramin and the non‐selective P2X antagonist, PPADS, did not alter the inhibition by ATP of the nigericin‐induced IL‐1β in P2X7R‐deficient M1 macrophages (Supplementary Figure S3A). Most directly, pyrophosphate (PPi, Figure 4B–E) and triphosphate (Supplementary Figure S3C) but not monophosphate (Figure 4D) inhibited IL‐1β release and caspase‐1 activation more potently than ATP, with an IC50 value for PPi of 1.03±0.05 mM (n=3). The non‐hydrolysable bisphosphonate, clodronate (Supplementary Figure S3D), exerted the same inhibition and was 3–5‐fold more potent than PPi or ATP, with an IC50 value of 0.43±0.14 mM. (Figure 4D and E). Both PPi and clodronate were effective to inhibit MTX‐mediated IL‐1β release from wild‐type M1 or M1/M2 intermediate polarized macrophages but neither inhibited ATP‐mediated release from these cells (Figure 4F). Finally, in HEK cells heterologously expressing P2X receptors or endogenously expressing P2Y receptors, neither PPi (5 mM) nor clodronate (1 mM) alone had any effect on membrane currents (data not shown) or cytosolic calcium transients, and nor did they significantly alter the ATP‐evoked responses in these cells (Supplementary Figure S3E).
Clodronate and other bisphosphonates, when encapsulated in liposomes, have been a widely used mechanism to selectively destroy macrophages (van Rooijen and Sanders, 1994) because macrophages phagocytose the liposomes; lysosomal phospholipases break down the liposomes and the resulting free intracellular bisphosphonates initiate apoptosis (van Rooijen and Sanders, 1997). Free bisphosphonates have not been found to accumulate intracellularly (Pennanen et al, 1995). Nevertheless, we asked whether the application of clodronate as used in our experiments (up to 30 min applications) may have toxic actions on macrophage that would explain the inhibition of IL‐1β by measuring LDH release from all experiments (Supplementary Figure S3B). No increase in LDH release over basal conditions was observed and in all cases LDH release was <1–3% of total LDH levels (n=10–14). This result and the observation that clodronate alone did not alter intracellular calcium levels (Supplementary Figure S3E) provide conclusive evidence against a toxic action of clodronate on these macrophages.
ATP differentially regulates actin polymerization and inflammasome localization during macrophage polarization gradient
P2X7R and MTX stimulation of M1 polarized macrophages are associated with rapid and dramatic cytoskeletal re‐arrangements (Pfeiffer et al, 2004; Verhoef et al, 2004). We noticed a striking difference in the actions of ATP on cell morphology over the M1–M2 polarization gradient, and therefore asked whether this may be associated with their actions to inhibit inflammasome activation. We examined F‐actin distribution before and 5 min after the addition of ATP (5 mM), MTX (0.2 nM) or ATP with MTX in M1 polarized wild type and P2X7R−/− macrophage, in intermediate state 3 polarized wild‐type macrophage and in M1 polarized wild‐type macrophage treated with A740003 (Figure 5A). ATP induced intense F‐actin distribution to only one side of the cell with extensive ramifications radiating outward over the other side; MTX induced an equally intense but uniform F‐actin localization around the entire cell whereas co‐application produced a randomly radiating network of F‐actin (Figure 5A, panels i–iv; Supplementary Figure S4C). In striking contrast, ATP induced an intense intracellular clustering of F‐actin in macrophage from intermediate state 3 polarization as well as from acute and chronic P2X7R‐depleted macrophages (Figure 5A, panels vi, x, xiv), whereas MTX produced essentially the same F‐actin pattern as observed in M1 polarized wild‐type macrophages (Figure 5A, panels vii, xi, xv). Co‐application of ATP and MTX to intermediate state 3 polarized macrophages produced both intracellular F‐actin clustering and radiating F‐actin ramifications (Figure 5A, panel viii) but co‐application to P2X7R‐depleted M1 macrophages produced the same pattern of intense intracellular clustering as seen with ATP alone in these macrophages (Figure 5A, panel xii; Supplementary Figure S4C). PPi and clodronate induced a similar pattern of intracellular clustering of F‐actin when co‐applied with MTX in P2X7R‐depleted macrophages and prevented polymerization of F‐actin to the cell edge (Supplementary Figure S4A and C).
To investigate whether the dynamics of F‐actin polymerization were important for the inhibition of inflammasome activation by ATP and PPi, we treated M1 polarized P2X7R−/− macrophages with cytochalasin B (CB) to disrupt coordinated actin cytoskeleton assembly. In agreement with earlier studies demonstrating that actin‐based cytoskeletal pathways are not directly involved with IL‐1β processing and release (Hornung et al, 2008), we found that CB itself did not induce IL‐1β release nor did it alter MTX‐induced release (Figure 5B and C). However, CB partially reversed (by 30–50%) the blockade of MTX‐induced IL‐1β by ATP and PPi (Figure 5B and C) and prevented intracellular actin clustering (Supplementary Figure S4B). To gain further insight into potential mechanisms underlying this earlier undetected involvement of the actin cytoskeleton in NLRP3‐inflammasome inhibition, we used a caspase‐1‐specific fluorescent probe to directly identify active caspase‐1 within the macrophage. No active caspase‐1 was detected in unstimulated M1 polarized macrophages from P2X7R+/+ (Figure 6A, upper panels) or P2X7R−/− mice (Figure 6B, left panel). Stimulation of M1 polarized wild‐type macrophages with ATP, or P2X7R−/− macrophages with MTX, for 5 min led to robust caspase‐1 activation in discrete intracellular aggregates that are likely to represent sites of active inflammasome complexes (Figure 6A, lower panels and B, middle panel). Several of these active caspase‐1 aggregates were observed within the actin barrier at the edge of the cell (Figure 6A and B, arrows), suggestive of potential caspase‐1/IL‐1β release sites. ASC immunostaining of unstimulated M1 macrophages showed a uniform cytosolic distribution (Figure 6A, upper middle panel), whereas ASC formed clusters after ATP stimulation, many of which co‐localized with active caspase‐1 (Figure 6A, arrowheads), confirming their status as an activated inflammasome complex. However, we also observed many ASC‐only and caspase‐1‐only aggregates on ATP stimulation, which may indicate heterogeneous subpopulations of inflammasome‐like complexes (Figure 6A, lower middle panel). We did not observe any co‐localization of active caspase‐1 with the lysosomal marker, cathepsin L (Figure 6A, right panels), which is in agreement with recent studies showing that the early release of IL‐β does not occur through lysosomal secretory pathways (Brough and Rothwell, 2007).
Modulation of redox signalling by ATP and PPi during macrophage polarization
We hypothesized that the inhibition of IL‐1β processing and release by ATP and PPi that was not reversed by treatment with CB (i.e. the actin‐independent inhibition) may result from reduced ROS production because P2X7R stimulation of M1 polarized macrophages is known to increase ROS production and is associated with NLRP3‐inflammasome activation (Dostert et al, 2008; Hewinson et al, 2008) (Supplementary Figure S4D) and ATP, PPi and bisphosphonates may act directly or indirectly as oxygen radical scavengers (Serretti et al, 1993; Kachur et al, 1997; Dombrecht et al, 2006). Both ATP and MTX induced similar rapid (within 2–3 min) increases in ROS production from M1 polarized macrophages (Figure 7A and E) but in M2 polarized macrophages the enhanced ROS production was significantly delayed (no significant increase occurred before 8–10 min) and reduced (Figure 7B and E). MTX similarly increased ROS production from M1 polarized P2X7R−/− macrophages (Figure 7C and E), but MTX‐induced ROS production in these cells was prevented by ATP, PPi and clodronate with IC50 values not significantly different from those obtained for inhibition of IL‐1β release (Figure 7C–E), and similar to the ROS scavenger N‐acetyl cystein (Figure 7E). Similarly, PPi and clodronate were able to block MTX‐induced ROS production in M1 wild‐type macrophages, but fail to block P2X7R‐induced ROS production from the same cells (Figure 7E), this could explain the underlying mechanism for the differential activation of the inflammasome by ATP and MTX observed between M1 and intermediate polarized macorphages (Figures 3A and 4F). Finally, MTX stimulation of ROS production and its inhibition by ATP and PPi were not significantly altered by treatment with cytochalasin B (Figure 7F).
The concept of monocytes/macrophages existing in functionally distinct phenotypes from pro‐ to anti‐inflammatory states has been well established by detailed studies on M1 (classically activated) and M2 (alternatively activated) macrophages (Gordon, 2003; Scotton et al, 2005; Martinez et al, 2006, 2008). However, only recently it has begun to be appreciated that macrophages are able to reversibly and dynamically switch from one activation state to the other (Stout and Suttles, 2004; Porcheray et al, 2005; Gratchev et al, 2006). Here, we have used a polarization gradient of cytokines and endotoxin to rapidly generate a 5‐stage in vitro model of macrophage polarization from M1 through M2. In states 1–4, intracellular IL‐1β remained high and all stimuli except ATP continued to activate the inflammasome and release IL‐1β. In intermediate polarization states, as well as in P2X7R‐deficient macrophages, ATP no longer signalled to caspase‐1 activation; rather, ATP now acted through its diphosphate and/or triphosphate chains independently of any other purine receptor to block activation of the NLRP3 inflammasome by other activators. This was accomplished by two independent mechanisms: a direct reduction in ROS production and a rapid trapping of the inflammasome complex because of dramatic intracellular clustering of the actin cytoskeleton. Our results suggest this uncoupling of P2X7R from caspase‐1 activation may prove to be a key trigger in the switch from a pro‐inflammatory macrophage towards its alternative functions in the resolution of inflammation. Moreover, we have identified a new cellular mechanism by which the inflammasome can be inhibited through physiological alterations in the cytoskeleton.
The characteristics of P2X7R‐induced release of IL‐1β from M1 activated macrophages are well established: intracellular pro‐caspase‐1 is rapidly cleaved and mature IL‐1β, as well as caspase‐1 and other inflammasome components (e.g. ASC) are released within minutes (2–10 min) after receptor stimulation (Mariathasan et al, 2004, 2006; Ferrari et al, 2006; Pelegrin and Surprenant, 2006). In this study we compared these established properties of P2X7R activation in M1 macrophages with those of macrophages exposed to a gradient of M1–M2 activation. We have focused on the intermediate polarization state (state 3) in wild‐type mouse macrophages because it provides us with a representation of a macrophage on the cusp between M1 and M2 phenotypes. In state 3, intermediate polarized macrophages intracellular levels of pro‐IL‐1β were similar to M1 macrophages, and P2X7R mRNA levels, protein levels and protein localization patterns showed no clear differences from M1 macrophages. Initial upstream signalling events, which occur in the seconds to minute time scale and represent opening of the cation‐permeable ion channel (calcium, sodium influx/potassium efflux) and the P2X7R‐pannexin‐1‐associated large pore (dye‐uptake) (Pelegrin and Surprenant, 2006, 2007), were also the same as in M1 macrophages. E. coli, MTX or nigericin‐induced IL‐1β release was the same as, or significantly increased, in these macrophages but ATP‐induced release and caspase‐1 activation were decreased by 50–80% because of an uncoupling of P2X7R signalling to ROS production and caspase‐1 activation. Most strikingly, when ATP was co‐applied with any of these other NLRP3‐inflammasome activators, caspase‐1 activation and IL‐1β release were virtually abolished. Taken together, these results suggest that a specific uncoupling of P2X7R from the downstream caspase‐1 cascade was occurring in intermediate state polarized macrophages without changes in receptor activation per se or its initial ion channel/dye‐uptake pore activation.
We initially supposed that such an uncoupling of P2X7R from the inflammasome allowed ATP to now has a negative action on IL‐1β release induced by other stimuli through one, or more, other purine receptor. However, we found that ATP blocked E. coli, MTX and nigericin‐induced caspase‐1 activation and IL‐1β release in these macrophages by an action that did not involve other purine or adenosine receptors. This is a valid conclusion because agonist profiles did not correspond to any of these receptors; generic P2X and P2Y antagonists (suramin and PPADS) did not alter the inhibition by ATP observed in P2X7R‐deficient macrophages and, unequivocally, because pyrophosphates (tri and di, but not mono, phosphates) as well as the non‐hydrolysable bisphosphonate, clodronate, inhibited caspase‐1 activation and IL‐1β with potency order of clodronate>triphosphate≈diphosphate>ATP≫ADP, suggesting that the diphosphate group resident in ADP could partially block inflammasome without involvement of PPi metabolism. It is tempting to speculate that ecto‐nucleotidases may play a critical role in the switch of a macrophage from an M1 to an M2 state during inflammatory resolution but future studies of expression levels and activities of the large family of these molecules (Zimmermann, 2000) are likely to provide insight into whether specific ecto‐nucleotidases may have a critical function in the switch of a macrophage from an M1 to an M2 state during inflammatory resolution. In any event, our results show that when P2X7Rs are fully functional, that is when they couple to inflammasome activation in M1 macrophages, the actions of ATP are solely stimulatory on IL‐1β release. That is, neither PPi nor clodronate inhibited ATP‐mediated IL‐β release from M1 macrophages nor even from state 3 intermediate polarized macrophages in which the P2X7R‐specific IL‐1β release was still present although at a much reduced level, while they effectively blocked MTX‐ and E. coli‐induced inflammasome activation. These results further support our suggestion that a specific uncoupling of otherwise functional P2X7Rs from the caspase‐1 cascade is the key event in the switch of ATP from inducing (through P2X7R) to inhibiting (through PPP/PPi) IL‐1β release (Figure 8). Recent pharmacological studies have shown that ROS production in M1 macrophages is a critical step for activation of the NLRP3 inflammasome in response to ATP, toxins or crystals (Dostert et al, 2008; Hewinson et al, 2008) and P2X7R activation recruits to the plasma membrane and activates NADPH oxidase complex (Hewinson et al, 2008). However, a recent study has not found genetic support for the role of ROS in NLRP3‐inflammasome activation (Hornung et al, 2008). Here, we found that in M2 macrophages P2X7R activation induces only a weak ROS response that may be owing to a specific uncoupling from recruitment of the NADPH oxidase complex without alterations in P2X7R up‐stream signalling, correlating with the low level of NLRP3 activation and IL‐1β release found in intermediate polarized macrophages after ATP/P2X7R stimulation.
By imaging individual macrophages using a caspase‐1‐specific fluorescent probe and co‐labelling for actin or ASC as an integral component in the inflammasome complex, we found that P2X7R activation in M1 macrophages not only induced rapid actin cytoskeletal rearrangements along a spatial gradient but also rapid formation of active caspase‐1/ASC aggregates, which we interpret as likely sites of functionally active inflammasomes. Several of these aggregates tightly abutted the external surface of actin filaments outlining the plasma membrane, suggesting potential sites of release. In striking contrast, ATP applied to P2X7R‐depleted M1 macrophages, or to state 3 intermediate polarized macrophages, produced a rapid and intense intracellular clustering of actin filaments. A similar pattern of intracellular clustering of actin was observed when ATP, pyrophosphate or clodronate were co‐applied with MTX to these macrophages with little or no activated caspase‐1 fluorescence signals. Cytochalasin B blocked this intracellular aggregation of actin and reversed by 30–50% the inhibition of release observed in the presence of MTX plus ATP or pyrophosphate. Thus, we have identified intracellular clustering of actin filaments as a novel mechanism for inhibition of inflammasome activation (Figure 8).
This study has direct and immediate clinical and physiological relevance. Clodronate and other bisphosphonates have been in clinical use for decades, primarily for the treatment of osteoarthritis and other inflammatory or metastatic bone diseases (Maksymowych, 2002; Russell et al, 2008). Their primary mechanism of action is generally considered to be one of effective chelation of calcium crystal deposits on bone mineral surfaces, resulting in inhibition of osteoclast apoptosis and bone resorption (Russell et al, 2008). Conversely, several studies have provided evidence that they may also exert an anti‐inflammatory action in bone and immune cells independent of their calcium chelation properties (Pennanen et al, 1995; Maksymowych, 2002). The pyrophosphate actions to inhibit IL‐1β release identified in this study may underlie these anti‐inflammatory actions of bisphosphonates. A novel, highly selective and potent P2X7R antagonist, AZ9056, is currently in clinical trials for rheumatoid arthritis (McInnes et al, 2007). Studies to date point to the primary mechanism of action of such P2X7R antagonists as being through inhibition of IL‐1β release (Labasi et al, 2002; Pelegrin, 2008). Our present work continues to support this conclusion but now raises the possibility that a secondary effect of P2X7R inhibition is to further inhibit IL‐1β release evoked by other activators by allowing pyrophosphates—either through extracellular ATP metabolism or directly through the phosphate chains of the intact nucleotides—to directly prevent inflammasome activation by other bacterial or endogenous inflammatory mediators through inhibition of ROS production and by intracellular trapping of the inflammasome complex. Clinical investigations into the potential of co‐therapy with P2X7R antagonists and bisphosphonates in chronic inflammatory diseases may prove fruitful.
Materials and methods
Cells and reagents
Key reagents and their sources: E. coli LPS O55:B5, ATP, ADP, AMP, adenosine, PPi, triphosphates, N‐acetyl cystein, α‐tocopherol, nigericin (Sigma); MTX (Alexis); clodronate, caspase‐1 substrate (Na‐5‐tetramethylrhodaminyl‐YVADAC(S‐acrylodan)‐OH), cytochalasin B (Calbiochem); IFN‐γ, IL‐4 and IL‐1β (PeproTech). E. coli DH5α strain, DAPI, phalloidin‐Texas Red and OxyBURST H2HFF‐BSA (Invitrogen). Abs for ELISAs and cathepsin L were from R&D, for P2X7R from Alamone Laboratories, caspase‐1 p10 rabbit polyclonal and β‐actin from Santa Cruz Biotechnology, ASC from Alexis, IL‐1β mAb from the Biological Resources Branch, National Cancer Institute. All HRP‐conjugated secondary Abs were from DAKO Cytomation. A740003 P2X7R antagonist was synthesized by ArtMolecule.
C57BL/6J (P2X7R+/+) and P2X7R‐deficient (P2X7R−/−) mice in C57BL/6J background were used (Chessell et al, 2005). All experiments were performed under the Animals (Scientific Procedures) Act 1986. Peritoneal macrophages were obtained as described earlier (Pelegrin and Surprenant, 2006). Briefly, the peritoneal cavity was gently lavaged with phosphate‐buffered saline (PBS, Invitrogen). The recovered buffer from two to three mice was pooled, cells were collected by centrifugation (250 g, 5 min) and plated on coverslips, or in 12‐well plates at a density of 106 cell/well, or in 24‐well plates at a density of 0.5 × 106 cell/well, or in 96‐well plates black with clear bottom at a density of 1.5 × 106 cell/well in RPMI 1640 media (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen), 100 units/ml penicillin and 100 μg/ml streptomycin (Invitrogen). The macrophages were allowed to adhere overnight (37°C, 5% CO2) and washed with fresh medium to remove unattached cells before use.
Cell culture and stimulation
Macrophages plated on 12‐well plates were primed with fresh medium supplemented with different stimuli to obtain a gradient of polarity phenotypes (Supplementary Figure S1A). For clarity, the different polarity phenotypes have been represented in the figures and in the text with a number, a colour and a shape. LPS (1 μg/ml) and IFN‐γ (20 ng/ml) for 4 h were used to differentiate to M1 phenotype (polarity state 1, square, red). IL‐4 (20 ng/ml) was used for 4 h to differentiate to M2 phenotype (polarity state 5, round, blue). A combination of LPS/IFN‐γ/IL‐4 for 4 h was used to differentiate an M1/M2 intermediate macrophage polarization phenotype (polarity state 3, diamond, green). To study the polarity changes and to achieve intermediate polarity states, cells were stimulated first for 4 h with IL‐4, washed and then stimulated for further 4 h with LPS/IFN‐γ (polarity state 2, star, pink); or stimulated first for 4 h with LPS/IFN‐γ, washed and then stimulated for further 4 h with IL‐4 (polarity state 4, triangle, yellow). Macrophages were stimulated for 15 min with agonists as described earlier (Pelegrin and Surprenant, 2007). Alternatively, to examine the blocking of IL‐1β release, macrophages were incubated with 0.01–5 mM of ATP, PPi, triphosphate or clodronate or with 5 mM of ADP, AMP, AD or Pi or with 20 mM N‐acetyl cystein or 10 μM α‐tocopherol together with MTX (0.2 nM). E. coli stimulation was performed as described earlier (Pétrilli et al, 2007). Macrophages were stimulated either with LPS/IFN‐γ (polarity state 1), IL‐4 (polarity state 5) or with a combination of LPS/IFN‐γ/IL‐4 (polarity state 3) for 2 h, washed in Optimem and challenged with E. coli for 1 h in the presence or absence of ATP (0.1–5 mM); the cells were then washed and placed in Optimem complemented with 100 units/ml penicillin and 100 μg/ml streptomycin for further 4 h.
Western blot, ELISAs, intracellular calcium and dye‐uptake assays
Detailed methods used for western blot analysis, ELISAs, Fura2 calcium imaging and ethidium bromide uptake have been described earlier (Pelegrin and Surprenant, 2006). Blots were analysed by densitometry measurements using NIH ImageJ software (http://rsb.info.nih.gov/ij/).
Immunocytochemistry and microscopy
Macrophages stimulated on coverslips were washed twice with PBS, fixed with 4% formaldehyde for 30 min at room temperature, and then washed three times with PBS. Cells were blocked with 3% bovine serum albumin and permeabilized with 0.1% triton X‐100 in PBS for 30 min at room temperature before incubating with rat anti‐mouse F4/80 antibody (1:500), rabbit anti‐P2X7R (1:500), rabbit anti‐ASC (1:250) or with goat anti‐cathepsin L (1:250) for 18–20 h at 4°C or for 2 h at room temperature. Cells were washed and incubated with appropriate FITC‐, TRITC‐ or Cy3‐conjugated secondary antibody (1:200) for 2 h at room temperature, then rinsed in PBS and incubated for 15 min with 300 nM of DAPI and in some experiments with phalloidin‐Texas Red (1:100). To stain for active caspase‐1, macrophages were activated with 5 mM ATP or 0.2 nM MTX for 5 min at 37°C, washed and incubated with the fluorochrome inhibitor of caspase‐1, green fluorescent peptide 5‐carboxyfluorescein‐Tyr‐Val‐Ala‐Asp‐fluoromethyl ketone (FAM‐YVAD‐fmk), according to the manufacturer's recommendations (Immunochemistry Technologies). Cells were fixed with 4% formaldehyde, blocked, permeabilized and co‐stained for F‐actin (phalloidin‐Texas Red), ASC or cathepsin L. All coverslips were mounted on slides with ProLong Gold Antifade Reagent (Invitrogen). Images were acquired at room temperature on a Delta Vision RT (Applied Precision) restoration microscope using a 60x/1.42 Plan Apo or 100x/1.40 Uplan Apo objectives and the 360/475 nm, 490/528 nm and 555/617 nm filter sets (Chroma 86000v2). The images were collected using a Coolsnap HQ (Photometrics) camera with a Z optical spacing of 0.2 μm. Raw images were then deconvolved using Softworx software and maximum intensity projections of these deconvolved images are shown in the results. ImageJ software (http://rsb.info.nih.gov/ij/) was used to quantify images.
Quantitative reverse transcriptase–PCR analysis
Detailed methods used for qRT–PCR have been described earlier (Pelegrin and Surprenant, 2006). Specific primers were purchased from Qiagen (QuantiTech Primer Assays), for each primer set the efficiency was >95% and a single product was seen on melt curve analysis. Relative expression levels were calculated using the 2−ΔΔCt method normalizing to GAPDH expression levels for each treatment and the fold increase in expression was relative to the smallest expression level.
Caspase‐1 activity and reactive oxygen species measurement
Macrophages were plated in a 96‐well plate black with clear bottom and primed to reach different polarity states. For casapse‐1 activity, cells were pre‐incubated with 37.5 μM TMR‐YVADAC(AD) at 37°C for 1 h and washed twice with E‐total. Fluorescence was recorded using FlexStation 3 over 30 min at 1 min intervals at 380 nm for excitation and 525 nm for the emission. ATP (5 mM) was added into the wells automatically by the machine at designated time points. For ROS generation, polarized macrophages were washed with PBS and incubated with ATP, clodronate, PPi (0.1–5 mM) or N‐acetyl cystein (20 mM) and/or MTX (0.2 nM) in the presence of 10 μg/ml H2HFF‐BSA and fluorescence emmision by the oxidation of H2HFF‐BSA was recorded by FlexStation 3 over 60 min at 1 min intervals at 492 nm for excitation and 520 nm for the emission.
Average results are expressed as the mean±s.e.m. from the number of assays indicated (from at least three separate cultures). Data were analysed by an unpaired two‐tailed Student's t‐test to determine difference between groups using Prism and Instat (GraphPad) software.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
Review Process File
We thank Elizabeth Martin, Austen Sitko and Mari‐Carmen Banos‐Gregori for technical support. This work was supported by the Wellcome Trust.
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial‐NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non‐commercial and no modifications or adaptations are made.
- Copyright © 2009 European Molecular Biology Organization