Endogenous prostaglandin E2 amplifies IL-33 production by macrophages through an E prostanoid (EP)2/EP4-cAMP-EPAC-dependent pathway

When activated through toll-like receptors (TLRs), macrophages generate IL-33, an IL-1 family member that induces innate immune responses through ST2 signaling. LPS, a TLR4 ligand, induces macrophages to generate prostaglandin E2 (PGE2) through inducible COX-2 and microsomal PGE2 synthase 1 (mPGES-1) (1). We demonstrate that IL-33 production by bone marrow-derived murine macrophages (bmMFs) requires the generation of endogenous PGE2 and the intrinsic expression of EP2 receptors to amplify NF-κB-dependent, LPS-induced IL-33 expression via exchange protein activated by cAMP (EPAC). Compared with WT cells, bmMFs lacking either mPGES-1 or EP2 receptors displayed reduced LPS-induced IL-33 levels. A selective EP2 agonist and, to a lesser extent, EP4 receptor agonist potentiated LPS-induced IL-33 generation from both mPGES-1-null and WT bmMFs, whereas EP1 and EP3 receptor agonists were inactive. The effects of PGE2 depended on cAMP, were mimicked by an EPAC-selective agonist, and were attenuated by EPAC-selective antagonism and knockdown. LPS-induced p38 MAPK and NF-κB activations were necessary for both IL-33 production and PGE2 generation, and exogenous PGE2 partly reversed the suppression of IL-33 production caused by p38 MAPK and NF-κB inhibition. Mice lacking mPGES-1 showed lower IL-33 levels and attenuated lung inflammation in response to repetitive Alternaria inhalation challenges. Cumulatively, our data demonstrate that endogenous PGE2, EP2 receptors, and EPAC are prerequisites for maximal LPS-induced IL-33 expression and that exogenous PGE2 can amplify IL-33 production via EP2 and EP4 receptors. The ubiquitous induction of mPGES-1-dependent PGE2 may be crucial for innate immune system activation during various IL-33 driven pathologic disorders.

IL-33 is a recently discovered member of the IL-1 cytokine family that binds to TLR/IL1R 2 superfamily receptor ST2 (2)(3)(4). IL-33 is constitutively expressed in the nuclei of epithelial cells at barrier sites such as the skin, lung, and intestine and is also expressed by fibroblastic reticular cells and endothelial cells within lymphoid tissues (5). It interacts with histones, promotes chromatin compaction, and retards NF-B transcriptional activity (5). It can also be released in response to cellular damage or activation and, after proteolytic processing to an active cytokine, can bind to a heterotrimeric receptor composed of T1/ST2 and IL-1RAcP, influencing both innate and adaptive type 2 immune responses (2). In addition to barrier cells, macrophages and dendritic cells can inducibly express IL-33 and contribute to the amplification of inflammatory responses in sepsis, cardiovascular disease, and allergy (6 -9). However, little is known about the mechanisms that control such inducible IL-33 expression.
Among the resident hematopoietic cells involved in innate immunity, macrophages are noteworthy for inducible expression of IL-33 in response to LPS in vitro (10 -12) and also upon infection with H3N1 or by TLR7/8 agonist stimulation (13). These same stimuli also induce macrophages to generate prostaglandin E 2 (PGE 2 ), a lipid mediator that plays critical roles in pyrexia, pain sensation, and inflammation. LPS-stimulated macrophages generate PGE 2 from endogenous arachidonic acid that is converted sequentially to prostaglandin H 2 (PGH 2 ) and then to PGE 2 , primarily by the inducible enzymes COX-2 and microsomal PGE 2 synthase 1 (mPGES-1), respectively (1). When released, PGE 2 acts through four types of E prostanoid receptors (EP 1 -EP 4 ) and can function as both autocrine as well as paracrine mediators (14). In addition to generating PGE 2 , macrophages also express EP receptors and respond to endogenous and exogenous PGE 2 ex vivo. In particular, by eliciting signaling through cAMP and PKA-dependent transcription factors, PGE 2 can amplify the production of IL-6 by macrophages and of IL-23 by dendritic cells but can also suppress the generation of IL-23 and IL-12 by human monocytes. Although exogenous PGE 2 can potentiate IL-33 expression by dendritic cells, no studies have addressed whether it regulates IL-33 expression by macrophages. Given the potent capacity for macrophages to generate PGE 2 , we hypothesized that mPGES-1-dependent endogenous PGE 2 may regulate IL-33 production in response to LPS/TLR4-dependent activation and have feedback effects in the presence of stimuli.
In this study, we show that mPGES-1-derived endogenous PGE 2 amplifies the generation of IL-33 by bmMFs upon stimulation by LPS. Early PGE 2 signaling through EP 2 receptors and EPAC, but not PKA, drives this amplification, which is synergistic with the requisite p38 MAPK and NF-B pathways needed to induce IL-33 expression. In addition, exogenous PGE 2 can markedly enhance the LPS-stimulated IL-33 production through both EP 2 and EP 4 receptors. We also demonstrate that endogenous PGE 2 is required for maximal induction of lung IL-33 expression and the attendant immunopathology in response to intranasal Alternaria alternata (Alternaria) inhalation. These amplification mechanisms may also play an important role in sepsis, viral infections, and other circumstances in which PGE 2 and IL-33 play roles in orchestrating innate immunologic responses.

Endogenous PGE 2 is required for maximal IL-33 production in response to LPS
To determine the role of endogenous PGE 2 in LPS-mediated IL-33 production, we stimulated bmMFs with LPS at concentrations of 0.5 and 1.0 g/ml. Supernatants were collected for the measurement of PGE 2 , and cell lysates were used to measure IL-33 by ELISA. Preliminary experiments demonstrated that IL-33 protein was not detected in resting, unstimulated bmMFs and showed maximal LPS-stimulated IL-33 protein expression by WT macrophages at 8 h (data not shown). Stimulation with LPS dose-dependently induced the release of PGE 2 by WT cells, which was significantly reduced in cells lacking mPGES-1 (Fig. 1A). Peak induction of IL-33 protein was significantly impaired in bmMFs from mPGES-1 knock-out mice ( Fig. 1B and supplemental Fig. 1A). Exogenous PGE 2 did not induce IL-33 production on its own but restored IL-33 generation by mPGES-1 knock-out cells to levels equivalent to WT cells at concentrations as low as 0.001 M (supplemental To determine whether the requirement for endogenous PGE 2 to amplify IL-33 generation was due to EP receptor signaling, we used a combination of molecular and pharmacologic approaches. LPS stimulation of bmMFs lacking EP 2 receptors demonstrated significantly lower IL-33 protein expression compared with WT cells, similar to the response of the mPGES-1 KO cells (supplemental Fig. 1B and Fig. 1, A and D). However, the potentiation of IL-33 expression by EP 2 KO cells in response to exogenous PGE 2 was ϳ80% reduced compared with WT (Fig. 1D) and mPGES-1 KO cells (supplemental Fig.  1B). Stimulation of WT and mPGES-1 KO bmMFs with the EP 2 receptor-selective agonist AE1-259-01 (1.0 M) in combination with LPS significantly enhanced IL-33 production but failed to do so in EP 2 KO cells (Fig. 1E). The selective EP 4 receptor agonist AE-329 (1.0 M) modestly potentiated LPS-induced IL-33 expression in all three genotypes (Fig. 1E). Similar inductions in LPS-induced IL-33 protein by EP 2 and EP 4 agonists in WT and mPGES-1 KO cells were also observed by Western blotting (WB) (Fig. 1F). EP 2 KO cells responded to the EP 4 receptor agonist but not to the EP 2 agonist (Fig. 1F). Neither the EP 1 receptor-selective agonist D004 (1.0 M) nor the EP 3 receptor agonist AE248 (1.0 M) had an effect on LPS-stimulated production of IL-33 (Fig. 1G).

LPS induces the coordinate expression of PGE 2 pathway constituents needed to amplify IL-33 expression
To assess the potential relationship between IL-33 induction and the inducible constituents of the PGE 2 synthetic pathway, we monitored the expression of the relevant transcripts by qPCR following stimulation of bmMFs with LPS. LPS-induced expression of IL-33 mRNA in WT cells was evident at 2 h and peaked at 4 h. The induction was significantly blunted in both mPGES-1 KO and EP 2 KO cells compared with WT cells ( Fig.  2A). Stimulation of WT cells with LPS elicited a strong, transient up-regulation of EP 2 receptor mRNA peaking at 2 h ( Fig.  2B) and also induced a more sustained increase in EP 4 receptor mRNA that peaked at 4 h (Fig. 2C). LPS induced increases in both COX-2 ( Fig. 2D) and mPGES-1 (Fig. 2E) mRNA expression that peaked at 4 and 8 h, respectively. The absence of EP 2 receptors did not significantly alter the induced expression of COX-2, mPGES-1, or EP 4 , and the absence of mPGES-1 did not alter the expression of COX-2 or either receptor (Fig. 2, B-E).

cAMP mediates potentiation of LPS-induced IL-33 production
Both EP 2 and EP 4 receptors activate adenylyl cyclase to induce cAMP accumulation (15). To determine whether cAMP was involved in potentiation of IL-33 production, we stimulated bmMFs with LPS in the absence or presence of the pharmacologic adenyl cyclase activator forskolin (50.0 M) or a cell-permeable cAMP analog, 8-bromo-cAMP (100.0 M). Both agents significantly potentiated LPS-induced IL-33 expression in mPGES-1 KO and EP 2 KO cells and also potentiated the response of WT cells (Fig. 3A). WBs confirmed that PGE 2 , 8-bromo-cAMP, EP 2 , and EP 4 agonists each increased the induction of LPS-induced IL-33 protein levels in WT cells (Fig.  3B). EP 2 KO cells responded to 8-bromo-cAMP but showed a reduced response to PGE 2 (Fig. 3C). Neither forskolin nor 8-bromo-cAMP altered the LPS-induced increases in COX-2 protein (Fig. 3, B and C).

PGE 2 -enhanced IL-33 production involves EPAC but not PKA or PI3K
cAMP activates PKA and downstream transcription factors such as CREB. To determine whether PKA and associated transcription factors were involved in IL-33 induction, we used cAMP analogs selective for PKA and EPAC, respectively. Both the PKA-selective cAMP analogue (6-bnz-cAMP, 25.0 M) and the EPAC-selective cAMP analogue 8-pCPT-2-O-Me-cAMP-AM (25.0 M) potentiated the production of IL-33 by LPS-stimulated bmMFs (Fig. 4A), although the potentiation was signifi-cantly greater for 8-pCPT-2-O-Me-cAMP-AM. To investigate further, we stimulated bmMFs in the presence or absence of PKA inhibitors. H89 (10.0 M) failed to inhibit the production of IL-33 by LPS-stimulated bmMFs of any genotype and also did not alter its potentiation by exogenous PGE 2 (Fig. 4B). Similarly, a second PKA inhibitor, KT5720, failed to alter LPS-in-duced IL-33 production (supplemental Fig. 1C). Moreover, the modest potentiation of IL-33 production by 6-bnz-cAMP was insensitive to H89 and KT5720, suggesting an off-target effect of this agonist (supplemental Fig. 1D). Stimulation of WT bmMFs with LPS in the presence of an EPAC-selective inhibitor (ESI09, 10.0 M) significantly decreased the production of Figure 1. Endogenous PGE 2 is required for IL-33 production in response to LPS and it signals through EP 2 and EP 4 . A, PGE 2 levels in cell supernatants from LPS-stimulated WT and mPGES-1 KO bmMFs after 8 h. B, IL-33 protein percentage production levels in corresponding cell lysates by ELISA. C, potentiation of IL-33 production by exogenous PGE 2 at concentrations in mPGES-1 KO and WT cells. D, exogenous PGE 2 induction to LPS-stimulated to EP 2 KO bmMFs. E, enhanced IL-33 production by the EP 2 receptor-selective agonist AE1-259-01 and the selective EP 4 receptor agonist AE-329. F, WB showing LPS-induced IL-33 protein and COX-2 levels in WT, mPGES-1 KO, and EP 2 KO cells in response to EP 2 and EP 4 agonists. G, IL-33 production in the presence of the EP 1 receptor-selective agonist D004 and the EP 3 receptor agonist AE248. Data are presented as mean Ϯ S.E. of at least three independent experiments. Statistical significance was determined using unpaired t test and one-sample t test (comparing fixed 100%). p Ͻ 0.05 was considered statistically significant. *, p Ͻ 0.05; **, p Ͻ 0.005. IL-33 to a level comparable with those of mPGES-1 KO and EP 2 KO bmMFs and did not alter IL-33 generation in either knockout strain (Fig. 4C). The addition of H89 to ESI09 produced no further inhibition of LPS-induced IL-33 production compared with ESI09 alone (Fig. 4D). In contrast, LPS-induced production of IL-6 was significantly inhibited by both H89 and ESI09, with an additive effect of the two inhibitors (Fig. 4E). ESI09 also modestly but significantly reduced PGE 2 -and AE1-259-01-enhanced IL-33 production by WT bmMFs (Fig. 4, F and G) but had no effect on potentiation by AE-329 (Fig. 4G).
Given the ability of EP 4 receptors to elicit PI3K activation (16 -18), we examined the effect of wortmannin (1.0 M) on LPS-induced IL-33 production and its potentiation by PGE 2 . Wortmannin had no effect on LPS-stimulated IL-33 production by bmMFs, irrespective of the absence or presence of PGE 2 ( Fig. 4H). Additionally, we used inhibitors of PKB/AKT signaling that is downstream of PI3K, namely FPA 124 (100.0 nM) and 10-DEBC (5.0 M), and again found no effect on IL-33 production both in absence or presence of PGE 2 (supplemental Fig. 1, E and F).

Gene silencing confirms that PGE 2 enhances IL-33 production in part through EPAC
To verify the involvement of EPAC in the potentiation of LPS-induced IL-33 expression, we performed EPAC gene-silencing studies. We used Rapgef3 and Rapgef4 siRNA pools to inhibit the expression of EPAC1 and EPAC2, respectively. We studied the effect of EPAC knockdown on enhancement of LPS-induced IL-33 expression by PGE 2   EPAC1 siRNA produced a modest statistically insignificant effect (Fig. 5B). qPCR confirmed that each siRNA pool successfully suppressed the expression of their respective targets (Fig.  5, C and D). Co-transfection of bmMFs with both siRNAs resulted in profound suppression (ϳ70%) of IL-33 expression. However, the efficacy of EPAC2 knockdown under these conditions was severely limited (data not shown).

PGE 2 -mediated potentiation of LPS-stimulated IL-33 production is independent of p38 MAPK and NF-B activation
Stimulation of macrophages with LPS via TLR4 activates p38 MAPK and NF-B, which synergize to induce the expression of proinflammatory cytokines and enzymes (19,20). We therefore sought to determine whether endogenous or exogenous PGE 2 modulated MAPK or NF-B activation to potentiate LPS-induced IL-33 expression. Two different inhibitors of p38 MAPK, SKF 86002 (1.0 M) and SB 706504 (3.3 M), inhibited LPSinduced IL-33 production (Fig. 6A). Stimulation with LPS induced p38 phosphorylation in all three bmMF genotypes (Fig.  6B). The NF-B inhibitor RO 106-9920 (5.0 M) significantly suppressed the LPS-induced expression of IL-33 in all three genotypes (Fig. 6C). LPS stimulation strongly induced phosphorylation of IB in WT cells (Fig. 6D) and also in both KOs (Fig. 6E). Both the p38 inhibitor SB 706504 and the NF-B inhibitor RO 106-9920 significantly suppressed LPS-induced production of PGE 2 (supplemental Fig. 1G). Exogenous PGE 2 partially reversed the suppression of IL-33 induction by both p38 inhibition and NF-B inhibition (Fig. 6, A and C). Neither PGE 2 nor selective EP 2 or EP 4 agonists altered the phosphorylation of p38 (Fig. 6B) or IB, whereas 8-Br-cAMP tended to decrease p38 phosphorylation (Fig. 6D). Neither the ERK1/2 inhibitor FR 180204 (1.0 M) nor the JNK inhibitor SP 600125  (100.0 M) altered LPS-induced IL-33 expression in any genotype either in the absence or presence of PGE 2 (Fig. 6, F and G).

Endogenous PGE 2 is necessary for inducible IL-33 expression and type 2 immunopathology in response to Alternaria
Alternaria alternata is a mold that induces severe asthma exacerbations in humans and marked IL-33-dependent immunopathology in mice. To determine whether PGE 2 is involved in this immunopathology, we administered an extract of Alternaria to WT and mPGES-1 KO mice intranasally. Compared with WT mice, mPGES-1-null mice showed significantly reduced lung inflammation, as indicated by total lung cells and eosinophils (Fig. 7, A and B). The numbers of group 2 innate lymphoid cells (ILC2s), a major target of IL-33, were reduced in the lungs of mPGES-1 KO mice compared with WT mice (Fig. 7C). Single-lung cell suspensions from Alternariatreated mPGES-1 KO mice expressed significantly lower levels of IL-33 mRNA than WT controls (Fig. 7D), whereas IL-33 transcripts from PBS-challenged WT and mPGES-1 KO mouse cells were below detection levels. Similar trends in inflammation, ILC2 expansion, and IL-33 expression were identified in EP 2 receptor knock-out mice (n ϭ 1 experiment, data not shown).

Discussion
Originally identified as a DNA-binding protein, IL-33 is now well recognized as an inducer and amplifier of innate immune responses by activating ST2-expressing hematopoietic and structural cells. Although IL-33 is strongly implicated in type 2 immune responses to allergens and parasites, driving ILC2 expansion and cytokine production, it is also elevated in the serum of patients with myocardial infarction, congestive heart failure, and trauma/septic shock (21). Experimental models also demonstrate that IL-33 is critical for the induction of local inflammation at the onset of colitis and LPS-induced endotoxic shock (8), both of which also involve prominent production of PGE 2 (22,23). Macrophages are important innate hematopoietic effector cells that are major sources of PGE 2 during inflammatory responses because of their capacity for inducible expression and function of COX-2 and mPGES-1. Macrophages can also generate IL-33 inducibly, and this induction can be facilitated by pharmacologic enhancement of intracellular cAMP levels. Because PGE 2 can modulate the generation of certain cytokines through both autocrine and paracrine pathways (and frequently through cAMP), we sought to determine whether the endogenous PGE 2 synthetic pathway and EP receptors mediated an amplification step for IL-33 expression in response to LPS, the ligand for TLR4, and a key inducer of macrophage activation in sepsis and antimicrobial responses.
We induced IL-33 protein and mRNA expression by bmMFs, which do not constitutively express IL-33, using LPS as a physiologically relevant TLR4 ligand. To identify potential contributions from endogenous PGE 2 , we compared the responses of WT cells to those derived from mice lacking mPGES-1. mPGES-1 KO cells showed significantly reduced LPS-mediated induction of IL-33 at both the RNA and protein levels (Figs. 1B and 2A) as well as the anticipated impairment in PGE 2 generation during the same time frame compared with WT control cells (Fig. 1A). The impairment of IL-33 induction in bmMFs lacking EP 2 receptors, which bind PGE 2 with high affinity, was comparable (Ͼ50% of the control, Fig. 1D) with that observed in mPGES-1 KO cells (Fig. 1B). Notably, EP 2 receptor mRNA expression was rapidly induced by LPS (Fig. 2B), peaking before the maximal induction of COX-2 (Fig. 2D), mPGES-1 (Fig. 2E), and IL-33 ( Fig. 2A) mRNAs. Both exogenous PGE 2 (Fig. 1C) and an EP 2 receptor-selective agonist (Fig. 1, E and F) restored LPSinduced IL-33 expression in mPGES-1 KO cells, but not in EP 2 receptor KO cells, to WT levels. Collectively, these observations are consistent with a critical role for endogenous PGE 2 in augmenting IL-33 expression, with EP 2 receptors being functionally dominant. The rapid up-regulation of EP 2 receptor expression by LPS may ensure maximal cellular sensitivity to PGE 2 prior to the peak levels of its production, permitting a substantial autocrine contribution to IL-33 expression.
Because both EP 2 and EP 4 receptors activate adenylyl cyclases, we suspected that the cAMP-dependent effectors PKA and/or EPAC would account for the effects of endogenous and exogenous PGE 2 . Indeed, pharmacologic activation of adenylyl cyclase with forskolin or stimulation with the cell-permeable cAMP mimic 8-Br-cAMP enhanced the LPS-induced expression of IL-33 in all three genotypes (Fig. 3A-3C), effectively bypassing the deficiency in endogenous EP receptor signaling. TLR4 signaling activates the PKA-dependent transcription factor CREB in peritoneal macrophages and a macrophage cell line, and this activation (and the associated IL-33 generation) is potentiated by cAMP-elevating agents. Several lines of evidence suggest that this is not the most relevant pathway involved in PGE 2 -mediated potentiation of IL-33 production by bmMFs. First, the absence of mPGES-1 and EP 2 receptors did not alter LPS-induced CREB phosphorylation, and neither exogenous PGE 2 nor 8-Br-cAMP potentiated this phosphorylation (supplemental Fig. 2A). Second, two different PKA inhibitors (H89 and KT5720, Fig. 4B and supplemental Fig. 1C, respectively) failed to suppress LPS-induced IL-33 expression in any genotype either in the absence or presence of exogenous PGE 2 . In contrast, H89 substantially suppressed the induced expression of IL-6, another cAMP-facilitated cytokine, under the same conditions (Fig. 4, D and E). Third, potentiation of LPS-induced IL-33 generation by a PKA-selective cAMP analogue was modest compared with that induced by the EPACselective agonist (Fig. 4A) and was not inhibited by either PKA antagonist, suggesting an off-target effect (supplemental Fig. 1D).
In contrast to the lack of effect of PKA inhibition, EPAC inhibition suppressed the LPS-induced expression of IL-33 in WT cells (Fig. 4C) to levels equivalent to those induced in cells lacking either mPGES-1 or EP 2 receptors. Notably, EPAC inhibition also decreased the potentiation of LPS-induced IL-33 by exogenous PGE 2 (Fig. 4F) and by an EP 2 agonist (Fig. 4G) but not by an EP 4 agonist (Fig. 4G), suggesting the involvement of an additional EP 4 -driven pathway activated by exogenous PGE 2 . The results of the siRNA knockdown studies were consistent with a significant contribution from EPAC to the PGE 2and cAMP-dependent potentiation of LPS-induced IL-33 expression (Fig. 5). We performed EPAC1 and EPAC2 double knockdown as well by electroporating both EPAC1 and EPAC2 siRNAs to WT bmMFs. As expected, EPAC1 and EPAC2 double knockdown significantly inhibited the production of both IL-33 mRNA and IL-6 secretion (ϳ70% for each), but there was a limitation of the assay, as EPAC1 and EPAC2 double electroporation reduced the efficiency of EPAC2 knockdown (data not shown). Thus, although endogenous PGE 2 promotes LPS-induced IL-33 expression primarily through EP 2 receptors and EPAC, exogenous PGE 2 can activate an additional cAMP-dependent pathway that depends, at least in part, on EP 4 receptors. Furthermore, this additional pathway does not involve PI3K (Fig. 4H) or its downstream effectors (Supplemental Figs. 1, D and E, and 2B).
TLR signaling elicits expression of cytokines and other proinflammatory proteins by activating NF-B-dependent transcription and p38 MAPK-dependent mRNA stabilization. Our studies suggest that these pathways were also essential for LPSmediated induction of IL-33 expression (Fig. 6, A-E) but were not modulated by PGE 2 or EP receptor signaling. The lack of difference in LPS-induced IB and p38 phosphorylation between WT, EP 2 KO, and mPGES-1 KO cells demonstrates that endogenous PGE 2 is not required for full activation of these pathways. Indeed, both exogenous PGE 2 and pharmacologic activation of cAMP-dependent signaling pathways slightly attenuated p38 and NF-B activation. The modulation of NF-B by cAMP pathways is complex and dependent on both stimulus and cell type, and our findings are consistent with previous reports regarding LPS-stimulated macrophages (24). Importantly, as shown in supplemental Fig. 1B, exogenous PGE 2 elicits significant potentiation of LPS-induced IL-33 production at low nanomolar doses. The exquisite sensitivity of these cells to potentiation of IL-33 production by PGE 2 likely permits the EPAC-dependent signal to override the modest PGE 2 -dependent suppression of the requisite NF-B and p38 signals. Notably, although selective inhibitors of NF-B and p38 markedly suppressed LPS-induced IL-33 expression, exogenous PGE 2 still induced substantial potentiation of IL-33 expression under these conditions. Given that synergism between signaling pathways is most evident at submaximal levels of input, the modest suppression of NF-B by PGE 2 may enhance the extent to which its subsequent activation of EPAC amplifies the response. Because NF-B and p38 activation are also required for endogenous PGE 2 production (supplemental Fig. 1G), the effect of the inhibitors is likely to reflect a loss of both direct (IL-33 transcription and mRNA stabilization) and indirect (potentiation by endogenous PGE 2 ) inputs required for maximal IL-33 expression.
Finally, we sought an in vivo context to determine whether PGE 2 is necessary to amplify IL-33-driven immunopathology. Although a single dose of Alternaria extract to naive mice induces the release of preformed IL-33 from epithelial cells, repeated intranasal dosing up-regulates lung IL-33 expression (and attendant ILC2 expansion and eosinophil recruitment) in part by inducing IL-33-expressing macrophages (25)(26)(27)(28)(29). Indeed, lung inflammation, ILC2 expansion, and IL-33 expression were all lower in mPGES-1 KO mice than in WT controls (Fig. 7). These findings support the physiological function of PGE 2 in amplifying IL-33 expression, which, in turn, drives the downstream IL-33-dependent effectors.
Our study demonstrates that endogenous PGE 2 , produced through the inducibly expressed COX-2/mPGES-1 system, uses EP 2 receptors and EPAC to orchestrate a pathway that amplifies IL-33 expression downstream of p38 and NF-B in mouse bmMFs (Fig. 8). To our knowledge, this is the first study showing the involvement of exogenous PGE 2 in up-regulating LPS-induced IL-33 production in macrophages. This autocrine system may facilitate IL-33-dependent features during early innate immune responses. We postulate that PGE 2 derived from paracrine sources may amplify macrophage production of IL-33 during Th2 inflammation, inflammatory processes in sepsis, asthma, inflammatory bowel disease, and endotoxin shock, where PGE 2 production is an integral part of the pathology (22,23,30,31). COX-2 inhibitors, which prevent the synthesis of PGE 2 , show efficacy in inflammatory diseases such as arthritis but can also increase the risk of adverse cardiovascular events. Given that IL-33 is proinflammatory in many contexts but also cardioprotective, it is tempting to speculate that some of the observed clinical effects of COX-2 inhibitors may reflect the loss of PGE 2 -dependent amplification mechanisms for IL-33 expression.

Bone marrow macrophages
All mouse strains used were described previously (14) and were on the C57BL/6 background and housed at Charles River. Bone marrow was harvested humanely from 6-to 8-week-old female mice with procedures approved by the Animal Care and Use Committee of the Dana-Farber Cancer Institute. Bone marrow cells were cultured with 50 ng/ml of recombinant murine MCSF (rmMCSF) for 7 days at 37°C in a humidified CO 2 (5%) incubator with fresh medium changed on day 2 and day 5. On day 7, bmMFs were harvested by using StemPro Accutase (Thermo Fisher Scientific, Grand Island, NY) and plated on tissue culture plates at 1 ϫ 10 6 cells/ml in RPMI 1640 complete medium. When adhered, cells were stimulated with LPS with or without pharmacological agents for 8 h unless stated otherwise. Supernatants were collected for measurements of PGE 2 and IL-6, and cells were lysed with Pathscan cell lysis buffer from Cell Signaling Technology by two cycles of freeze-thaw, and lysates were analyzed either by ELISA or WB.

ELISA and Western blotting
ELISAs for IL-6, PGE 2 IL-33, and phospho-CREB, and phospho-AKT were done according to the protocol of the manufacturer. WT bmMFs produced varying quantities (40 -150 pg/ml) of IL-33 in response to LPS, well within the detection limits of the ELISA kit (IL-33 ELISA, R&D Systems, detection range of 31.2-2000 pg/ml). Because the relative potentiation of IL-33 production by PGE 2 (its reductions in mPGES-1 and EP 2 KO cells) showed minimal interexperimental variability, we normalized the data to LPS-stimulated WT cells for each experiment and displayed the mean data as a percent change. Representative amounts of IL-33 produced by bmMFs upon LPS stimulation are shown in supplemental Fig. 1A. For WB, lysates were prepared and quantified using BCA protein estimation from Thermo Fisher Scientific. 30 -50 g of total protein for each sample with 4ϫ NuPage sample preparation buffer was run on Novex 4 -20% Tris/glycine gels from Thermo Fisher Scientific and post-run transferred onto a methanol-charged PVDF membrane. Membranes were blocked for 2 h with 5% skim milk in TBST (TBS-Tween 20) and processed further according to the recommendations of the manufacturer for primary and secondary antibodies.

Real-time PCR and gene silencing
For real-time PCR studies, kits and reagents from Qiagen were used. RNAs were extracted using the RNeasy kit and postquantification subjected to first-strand cDNA synthesis using the RT 2 first strand kit. qPCR assays were done with the respective primers using RT 2 SYBR Green qPCR Master Mix on Stratagene qPCR systems. Relative mRNA levels were determined in comparison with GAPDH. For Rapgef3-and Rapgef4-silencing experiments, ON-TARGETplus pools of the respective siRNAs, along with a non-targeting control, were electroporated to bmMFs at a concentration of 500 nM/1 ϫ 10 6 cells using the Nucleofector mouse macrophage electroporation kit from Lonza (Atlanta, GA). 24 h after electroporation, cells were supplied with RPMI 1640 complete medium with rmMCSF and allowed to recover for another 24 h. After a total of 48 h, cells were subjected to LPS stimulation along with other agonists.

Intranasal inhalation challenge and flow cytometry
C57B/L6 WT and mPGES-1 KO or EP2 KO mice (9 -11 weeks old) were given 12 g of A. alternata cellular extract (Alternaria) in 20 l of PBS or PBS alone by intranasal inhalation on days 0, 3, 6, and 9 as described previously (32) and euthanized 18 h after the last dose. Lungs were manually divided into pieces ϳ1.0 mm 3 in size, followed by digestion in RPMI containing 428 units/ml Collagenase IV and 20 mg/ml DNase I (30 min, 37°C). After digestion, suspensions were strained and passed through 70 M sterile filters and washed with PBS, followed by RBC lysis. After RBC lysis, the suspensions from individual mice were washed, strained again, and counted (for total cell counts). At this point, about 2 million cells from each mouse were aliquoted for RNA extraction later. From the remaining cells, 1 million cells per mouse were washed again and blocked (1 h, 4°C) with 1% rat anti-mouse CD16/CD32 (BD Biosciences) and 10% FBS and then stained (1 h, 4°C) with appropriate antibodies. Single-cell suspensions were stained with CD45 PercPCy5.5, lineage antibodies-FITC (CD19, CD3, CD11b, CD11c, Ly6G/C, Nk1.1, and FceR1), Siglec-F PE, and Thy 1.2 APC. Post-staining, cells were fixed with 4% paraformaldehyde, and data were acquired using a FACSCanto II flow cytometer with FACSDiva software. Analysis was done with FlowJo (Tree Star, Ashland, OR).

Statistical analysis
Data are presented as mean Ϯ S.E. of at least three independent experiments unless otherwise stated. Statistical significance was determined using unpaired t test and one-sample t test (comparing fixed 100%). p Ͻ0.05 was considered statistically significant (*, p Ͻ 0.05; **, p Ͻ 0.005).