Prostaglandin E2 controls metabolism and reduces lung immune cell responses

Rag2^−/−,28 EP2(Ptger2)^−/−,29 EP4(Ptger4)^−/−30 mice were maintained under specific pathogen‐free conditions in accredited animal facilities in the University of Edinburgh. Vav^CreEP4^fl/fl and Thy1.2^iCreEP4^fl/fl mice were generated by crossing Vav‐Cre mice31 and Thy1.2^iCre mice,32 respectively, to EP4‐floxed mice33 and maintained under specific pathogen‐free conditions in accredited animal facilities in the University of Edinburgh. All the genetically modified mouse strains are at the C57BL/6 background. Wild‐type female C57BL/6 mice were purchased from Harlan UK or bred in animal facilities within the University of Edinburgh. Sex‐matched mice aged >6 weeks were used in experiments. Mice were randomly allocated into different treatment groups and analysed individually. No animal was excluded for analysis except one mouse in Figure 2A which was overlooked at the step of ex vivo restimulation of lung single cells with PMA, ionomycin and Brefeld A (for performing intracellular cytokine staining). All animal experiments were conducted in accordance with the U.K. Animals (Scientific Procedures) Act of 1986 with local institutional ethical approval by the University of Edinburgh Animal Welfare and Ethical Review Body.

Antibodies to mouse CD45 (clone 30‐F11), CD19 (clone 6D5), NK‐1.1 (clone PK136), CD4 (clone RM4‐5), CD90.2 (clone 30‐H12), CD3e (clone 145‐2C11), CD11b (clone M1/70), CD11c (clone N418), ICOS (clone7E.17G9), IL‐5 (clone TRFK5), IL‐13 (clone eBio13A), Ki‐67 (clone 16A8) and Ly6G (clone 1A8) were purchased from eBioscience or Biolegend. Antibodies to mouse ST2(IL33‐R) (clone DJ8) and SiglecF (clone E50‐2440) were purchased from mdbioproducts and BD respectively. LIVE/DEAD™️ Fixable Blue Dead Cell Stain Kit for UV excitation, UltraComp eBeads™️ and mouse IL‐5/IL‐13 ELISA kits were purchased from Thermo Fisher Scientific. Recombinant mouse IL‐33 were purchased from Biolegend. PGE₂, Butaprost (free acid) and L‐902688 (EP4 agonist) were obtained from Cayman, while PF‐04418948 was purchased from Abcam. Indomethacin, phorbol myristate acetate (PMA), ionomycin, brefeldin A, dibutyryl cyclic adenosine monophosphate (db‐cAMP) and 3‐isobutyl‐1‐methyxanthine (IBMX) were purchased from Sigma. Complete RMPI consisted of RPMI 1640 (Gibco) supplemented with 10% FBS, 1% Penicillin/Streptomycin, 1% L. glutamine and 50 μM β‐mercaptoethanol.

A variety of different mouse lines including C57BL/6 wild‐type, EP4^−/−, Vav^CreEP4^fl/fl, and their appropriate control mice were anaesthetised by inhalation of isoflurane and the pro‐allergic cytokine IL‐33 (200 ng per treatment) was administered via intratracheal administration once daily for consecutive 3 days. When indicated, EP2 or EP4 agonist (each 10 μg) or vehicle control (PBS) were co‐administered with IL‐33. Indomethacin (5 mg/kg/day), PF‐04418948 (10 mg/kg/day) or vehicle (0.5% EtOH) was administered in drinking water started one day before IL‐33 administration. 2‐Deoxy‐d‐glucose (2‐DG, 1 g/kg/day) (Abcam) was administered via intraperitoneal injection (one injection per day throughout the whole experiment). All animals were culled via anaesthetic overdose by i.p injection of 200 μl Pentoject Pentobarbital Sodium solution (200 mg/ml) and subsequent exsanguination. Bronchoalveolar lavage (BAL) was collected (x3 lavages with 800 μl ice cold PBS) followed by aseptic lung dissection for subsequent tissue digestion. Both male and female mice were used in Figures 1E–G, 2 and 5N,O, while only female WT mice were used in Figures 1B,C, 3, 5.

The murine lungs were digested in Liberase TL (Roche) and DNase I (Sigma) at 37°C for 35 min. Digested tissue was passed through a 100 μm cell strainer and red blood cells lysed using ACK lysing buffer. Viable and dead lung cell counts were determined using 0.4% Trypan Blue solution and a TC10™ Automated Cell Counter (BioRad). BAL samples were centrifuged and supernatants from the first BAL retrieval harvested, snap frozen on dry ice and stored at −80°C for subsequent cytokine detection using mouse IL‐5/IL‐13 ELISA kits according to manufacturer's instructions. Red blood cells were lysed from BAL cell pellets using ACK lysing buffer and cell counts performed as before.

For in vitro experiments, all the work was carried out aseptically and single cell suspensions were obtained from lungs and bone marrow of untreated Rag2^−/− mice. Whole lung and bone marrow cells were cultured in complete RPMI with a cytokine cocktail (IL‐2, IL‐7, and IL‐33) with various reagents indicated in the figure legends at 37°C, 5% CO₂ for 3–4 days. IL‐5 and IL‐13 in the supernatants were analysed by ELISA. In some experiments, CD90.2⁺ cells were pre‐sorted from Rag2^−/− lung and bone marrow single cell suspensions using an EasySep™ Mouse CD90.2 Positive Selection Kit II (Stemcell Technologies) according to manufacturer's instructions. Live Lineage(CD11c/CD11b/NK1.1)⁻CD45⁺CD90.2⁺ST2⁺ ILC2s were sorted using a BD FACS Aria II and then cultured in complete RPMI containing 50 μM β‐mercaptoethanol with IL‐2, IL‐7, and IL‐33 with or without PGE₂ in a round bottom 96‐well plate at 37°C in 5% CO₂ for 3 days.

For surface staining only, lung and BAL single cell suspensions were washed in PBS and stained with LIVE/DEAD™️ Fixable Blue Dead Cell Stain Kit. Cells were then blocked with IgG from rat serum and surfaced stained. For lung and BAL neutrophil/eosinophil detection, cell suspensions were labelled with CD45, CD11b, CD11c, Ly6G, and SiglecF. For BAL ILC2 detection, cell suspensions were labelled with a cocktail of lineage negative (Lin‐) markers consisting of CD3e, CD19, CD11b, CD11c and NK1.1, as well as CD45, CD4, CD90.2, ICOS, and ST2. In some experiments, anti‐mouse CD200R3 Ab was added to the lineage marker list. For intracellular staining, single cell lung suspensions were stimulated with a cocktail of PMA, Ionomycin and Brefeldin A (or Golgiplug) and cultured in complete RPMI for 4 h at 37°C with 5% CO₂. After stimulation, lung cells were Live/Dead stained, blocked and surface stained for ILC2s as before. Next, lung cells were fixed with intracellular fixation buffer (eBioscience), and then stained for intracellular IL‐5, IL‐13 and Ki‐67 for 20 min at 4°C. To measure c‐Myc expression and STAT5 and S6 phosphorylation, lung and bone marrow CD90.2+ cells from naïve Rag2−/− mice were cultured with IL‐2, IL‐7 and IL‐33 forr 7–10 days, and then live CD45 + Lin‐ST2+ ILC2s were sorted using flow cytometry and cultured with or without PGE₂ or db‐cAMP in the presence or absence of the cytokine cocktail (IL‐2/IL‐7/IL‐33) for further 3 days. Cells were fixed and stained with antibodies against c‐Myc (clone E5Q6W), pSTAT5 (clone C71E5), or pS6 (clone C71E5, all from Cell Signaling Technology) using the Foxp3 staining kit (eBioscience). Cell samples were acquired on a BD LSR Fortessa analyser, and results were analysed by FlowJo™.

The Seahorse XFe96‐well metabolic analyser (Agilent) was used to investigate the effect PGE₂ may have upon ILC2 glycolysis in vitro. RAG2^−/− mice were subjected to i.t administrations of IL‐33 (one i.t. per mouse per day) on days 0, 2 and 4. Lung and BAL single cell suspensions were prepared on day 6 and pooled together for sorting ILC2s as described above. 4 × 10⁴ ILC2s were rested overnight in a 96‐well round bottom plate in complete RPMI (supplemented with L‐glutamine) and IL‐2 and IL‐7 (both 10 ng/ml). On the next day, IL‐33 (10 ng/ml), PGE₂ (1 μM) or DMSO (0.01% v/v as vehicle) was added to media and cultured for a further 24 h. The next day, ILC2 counts were performed on each treatment to confirm >95% viability. Approximately 4 × 10⁴ ILC2s re‐suspended in 50 μl Seahorse XF RPMI pH 7.4 (Agilent) (+1 M glucose, 100 mM pyruvate and 200 mM glutamine) were added in triplicate to wells of a Seahorse XF96 V3 PS cell culture microplate (Agilent) coated 24 h prior with Cell‐Tak™️ cell adhesive (Sigma). Precise measurements of glycolysis in ILC2s were carried out using the Seahorse XF Glycolytic Rate Assay Kit (Agilent; 103344‐100) according to manufacturer's instructions. The real time oxygen consumption rate (OCR), extracellular acidification rate (ECAR), and proton efflux rate (PER) of ILC2s during glycolysis was directly measured, including glycolytic rates for basal conditions and compensatory glycolysis after injections of 0.5 μM Rotenone/Antimycin A (inhibits mitochondrial respiration) and 50 mM 2‐DG (inhibits glycolysis) respectively into all treatment wells.

To explore the effect of cAMP on ILC2 gene expression, we first downloaded raw sequence data in FASTQ format from the GEO database (accession number: GSE131996↗). FastQC was used to assess the quality of FASTQ files (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/↗). Sequence reads were mapped to mouse genome using STAR 2.7.34 QualiMap35 was used to assess the quality of mapped data and featureCounts was employed to count uniquely mapped fragments against genomic features defined by the GENCODE annotation file (Mus_musculus.GRCm39.103.gtf). The counts were further processed for differential expression gene analysis in R (4.0.3). Differentially expressed genes were analysed by DESeq236 and significance was identified using adjusted p‐value < .05 and the absolute value log2 fold change ≥1. Ensembl version 103 was used for gene annotation. Gene Set Enrichment Analysis37 was performed for functional analysis using the Molecular Signatures Database v7.4. Hallmark, KEGG and REACTOME curated gene sets were used as references for screening enriched pathways. p values were adjusted and enrichment scores were normalised. An adjusted p‐value < .05 (Benjamini–Hochberg) was used as an indicator of significance in pathway analysis.

All data were expressed as mean ± SD (in vivo) or SEM (in vitro). For certain experiments where both sexes of animals were used, absolute cell numbers were normalised by sex and data were presented as fold changes. Statistical significance between two groups was examined using an unpaired, 2‐tailed Student's t‐test. One‐way and two‐way ANOVA with post‐hoc Holm–Sidak's multiple comparisons tests were used to evaluate statistical significance between multiple groups. Statistical analyses were performed using Prism 8 software (Graphpad) and significance was accepted at p < .05.

At the steady state, PGE₂ is expressed in most tissues including the lung and its levels are elevated by inflammatory stimuli. We thus investigated whether endogenous PGE₂ suppressed lung ILC2 immune responses. We administered intratracheally IL‐33 into the lungs of wild‐type (WT) C57BL/6 mice. Mice also received indomethacin, an NSAID that blocks the biosynthesis of all PGs including PGE₂, or vehicle control in drinking water (Figure 1A). In agreement with a previous report,38 indomethacin significantly increased IL‐33‐mediated accumulation of CD45⁺Lineage(Lin)⁻ST2⁺ ILC2s to the lung and enhanced type 2 cytokine (i.e., IL‐5 and IL‐13) production from ILC2s (Figure 1B,C). However, indomethacin treatment had no effect on the accumulation of eosinophils in the lung (data not shown), possibly because indomethacin inhibits all PG synthesis and some PGs have direct actions on eosinophils. These results suggest that endogenous PGs suppress lung ILC2 responses.

As indomethacin inhibits biosynthesis of all PGs, we examined whether PGE₂ is involved in the regulation of lung ILC2 responses, and if yes, through which receptor(s). We examined published datasets39 and found that ILC2s isolated from various tissues including lung, bone marrow, skin, and intestine have considerably higher gene expression of EP4, followed by EP2 and EP1, and that EP3 has much lower expression levels in ILC2s (Figure S1). Given that down‐regulation of EP2 is associated with NERD,14 we first asked whether blockade of endogenous PGE₂‐EP2 signaling influenced lung ILC2 responses. We injected IL‐33 into the lungs of EP2‐deficient29 or control mice and measured lung ILC2 activation (Figure 1D). As expected, EP2‐deficiency increased lung ILC2 accumulation and production of type 2 cytokines (IL‐5 and IL‐13) in response to IL‐33 (Figure 1E,F), although naïve EP2 deficient mice had similar lung ILC2s compared to naive WT mice (Figure S2A,B). EP2 (Ptger2) gene expression was reduced by half in EP2‐heterozygous mice (Figure S2C), which led to increased IL‐5‐producing ILC2s but did not affect ST2⁺ total ILC2s or IL‐13‐producing ILC2s (Figure 1F), indicating that EP2 expression levels may differently control distinct ILC2 subpopulations. EP2 deficiency also increased lung Lin⁺CD90.2⁺ST2⁺ Th2 cells, but type 2 cytokine production from T cells was not significantly affected by the loss of EP2 (Figure 1E,F). In agreement with the increase in ILC2s, EP2‐deficient mice had more eosinophils after IL‐33 treatment (Figure 1G and Figure S3). Our data suggest that endogenous PGE₂‐EP2 signaling restricts IL‐33‐dependent lung ILC2 responses.

To test whether blocking PGE₂‐EP4 signaling modulates lung ILC2 responses, we administered IL‐33 to global EP4‐deficient30 and control mice (Figure 2A). Unexpectedly, EP4 deficiency did not enhance ILC2 accumulation in the lung but reduced type 2 cytokine production compared to WT and heterozygous mice (Figure 2B,C). Global EP4 deficiency also did not increase eosinophils in the lung (Figure 2D). We then investigated whether specific deletion of EP4 in immune cells including ILC2s enhances lung ILC2 responses. As there were no mouse lines that selectively target only ILC2s, we generated Vav^CreEP4^fl/fl mice by crossing EP4‐floxed mice33 to Vav‐cre mice31 which drives ablation of EP4 in all hematopoietic‐derived immune cells (including ILC2s) (Figure S4A). Naïve Vav^CreEP4^fl/fl mice had similar lung ILC2s to naïve control EP4^fl/fl mice (Figure S4B,C). A previously published report showed that Vav‐dependent EP4 deficiency enhanced lung ILC2 function and eosinophilic inflammation.27 However, we found that mice with conditional EP4 deficiency in all immune cells had comparable lung ILC2 responses after IL‐33 administration, albeit with a moderate reduction of lung ST2⁺ ILC2 accumulation, compared to control EP4^fl/fl and Vav^CreEP4^fl/+ mice (Figure 2E,F). Homozygous Vav^CreEP4^fl/fl mice had a trend to reduce eosinophils (Figure 2G), which was likely due to a direct effect of EP4 on Vav‐expressing eosinophils. Both EP4^−/− and Vav^CreEP4^fl/fl mice ablated EP4 from the germline, which may have a potential effect on ILC (including ILC2) development. To exclude this possibility, we employed an inducible EP4 deficient mouse line by crossing Thy1.2^iCre mice32 to EP4‐floxed mice to generate Thy1.2^iCreEP4^fl/fl mice, which do not have EP4 expression in Thy1.2‐expressing cells (i.e., all ILC and T cells) after administration of tamoxifen. Again, Thy1.2‐iCre‐driven inducible EP4 deficiency in lymphocytes did not alter IL‐33‐induced lung ILC2 responses (Figure 2H,I). Thy1.2‐iCre‐driven inducible EP4 deficiency significantly reduced lung eosinophils (Figure 2H–J), which may be due to the effects of EP4 signaling on other Thy1.2‐expressing cells such as the neuron, stromal or endothelial cells. Thus, our results suggest that endogenous PGE₂‐EP4 signaling does not suppress lung ILC2 responses.

To examine the effects of exogenous activation of EP2 and EP4 on ILC2 responses in vivo, we intratracheally administered IL‐33 together with highly selective agonist for EP2 [butaprost (free acid)] or EP4 (L‐902688) into C57BL/6 mice (Figure 3A). Administration of IL‐33‐induced accumulation of eosinophils and neutrophils in the bronchoalveolar lavage (BAL), while such accumulation was almost completely attenuated by co‐administration of the EP4 agonist (Figure 3B). The EP4 agonist had no effects on IL‐33‐induced lung ILC2 responses in EP4KO mice (Figure S5), indicating that its inhibitory actions were not due to the off‐target effects. Co‐administration of an EP2 agonist had no significant effects on IL‐33‐dependent accumulation of eosinophils and neutrophils (Figure 3C). Flow cytometric analysis suggested that IL‐33‐induced recruitment of lineage‐negative ST2⁺ ILC2s in the airway was impeded by both EP4 and EP2 agonists (Figure 3D,E). Analysis of lung single cells showed that IL‐33 increased numbers of total ST2⁺ ILC2s and their production of type 2 cytokines (e.g., IL‐5 and IL‐13), with the accumulation of ILC2s was again inhibited by both EP4 and EP2 agonists (Figure 3F–I).

Engagement of EP2 and EP4 activates the intracellular cAMP pathway. To test if the elevation of cAMP contributes PGE₂ suppression of lung ILC2 responses, we administered 3‐isobutyl‐1‐methylxanthine (IBMX), a non‐selective inhibitor of cyclic nucleotide phosphodiesterases (PDEs) that increases the intracellular cAMP level through blocking cAMP degradation. Like EP2 or EP4 agonists, IBMX significantly reduced lung ILC2 numbers and type 2 cytokine production (Figure 3E,G,I). It is worth to note that although the EP4 agonist markedly suppressed IL‐5 production from both IL‐5⁺IL‐13⁻ and IL‐5⁺IL‐13⁺ ILC2s in response to IL‐33 administration, neither the EP2 agonist nor IBMX reduced IL‐5 production from lung ILC2s (Figure 3H,I). This may be the cause for unchanged eosinophils by EP2 agonist and IBMX as eosinophil activation is driven by IL‐5. In addition, the recent studies suggested that platelet activation and platelet‐adherent leukocytes were associated with NERD,40, 41 which might link to disruption of eicosanoid balance such as overproduction of cysteinyl leukotriene and reduction of PGE₂.14, 41 As platelet adherent promotes lung ILC2 effector function,42, 43 we examined whether PGE₂‐EP2‐cAMP signaling had an effect on platelet‐adherent (i.e., CD41‐positive) ILC2s using flow cytometry. IL‐33 treatment indeed increased the numbers of CD41⁺ ILC2s in the lung, and this was abrogated by EP2 agonist and IBMX, although neither IL‐33 nor EP2 agonist or IBMX affected CD41 adherent at the single cell level (Figure S6). Thus, exogenously activated the PGE₂‐EP2/EP4‐cAMP pathway suppresses lung ILC2 responses.

We next examined the underlying mechanisms for PGE₂ suppression of ILC2 responses. To this end, we firstly stimulated whole lung cells isolated from Rag2^−/− mice (which have no T or B cells but have ILCs including ILC2s) in vitro with a cytokine cocktail including IL‐33, IL‐2 and IL‐7 for 3–5 days. This cytokine cocktail markedly increased viable CD45⁺ immune cells, expanded the Lin⁻ST2⁺ ILC2 population, and induced IL‐5 and IL‐13 production from ILC2s in the whole lung cell cultures (Figure S7A). As seen in human ILC2 cell cultures,26 addition of PGE₂, EP2 and EP4 agonists significantly reduced the numbers of live Lin⁻ST2⁺ ILC2s and overall IL‐5 and IL‐13 production (Figure S7B,C). Among live ILC2s, PGE₂ and EP2 agonist significantly inhibited IL‐5 and IL‐13 double positive cells but increased IL‐5 single expressing cells, while EP4 agonist did not influence cytokine production from live ILC2s (Figure S7B). These results suggest that PGE₂ inhibits lung ILC2 cell activation and effector function.

To investigate whether PGE₂ directly controls ILC2 activation, we sorted ILC2s from lung (Figure 4A–D) and bone marrows using flow cytometry (Figure S8) and cultured them with the cytokine cocktail including IL‐2, IL‐7 and IL‐33 for 3 days. Addition of PGE₂ significantly decreased ILC2 viability, cellular size [evidenced by reducing forward scatter (FSC)], and IL‐13 production (Figure 4A–E and Figure S8). Activation of the PGE₂‐EP2/EP4‐cAMP pathway also down‐regulated ST2 expression (Figure 4E,F). In agreement with above results from whole Rag2^−/− lung cell culture and a previous report showing that cAMP inhibits ILC2 production of IL‐13 rather than IL‐5,44 neither PGE₂ nor EP agonists inhibited IL‐5 production (Figure 4C–E). These results indicate that PGE₂ may reduce IL‐13 expression from IL‐5‐ and IL‐13‐double positive cells and turn them into IL‐5‐single positive ILC2s (Figure 4E). In agreement with results from whole Rag2^−/− immune cell cultures, EP4 agonist had no effects on IL‐5‐single or IL‐5/IL‐13‐double expressing ILC2s but reduced IL‐13‐single expressing ILC2s (Figure 4E). Other cells such as eosinophils and basophils can also produce IL‐5, which may be the reason for PGE₂ suppression of IL‐5 production in whole Rag2^−/− lung cell culture but not in purified ILC2 cultures. ILC2s have been reported produce the regulatory cytokine IL‐10,45, 46, 47 and PGE₂ promotes IL‐10 production from multiple cell types such as Th2 or macrophages.10 We thus examined if PGE₂ impacted on IL‐10 production from ILC2s. Unlike in other cell types, PGE₂ still suppressed IL‐10 production from in in vitro cultured ILC2s (Figure S9). Moreover, PGE₂ inhibition of ILC2 activation and effector function (e.g., ST2 expression and cytokine production) was mimicked by dibutyryl‐cyclic AMP (db‐cAMP, a cell‐permeable cAMP analog) and IBMX (Figure 4E,F). Pre‐treatment of Rp‐8‐CPT‐cAMP, a cAMP antagonist, abrogated PGE₂ suppression of ST2 expression although Rp‐8‐CPT‐cAMP had no impacts on PGE₂ inhibition of ILC2 viability and cell size (Figure S10). These results suggest that PGE₂ directly acts on ILC2s and controls ILC2 survival, growth and function through EP2/EP4 and the downstream cAMP signaling.

To understand the underlying mechanisms for PGE₂ suppression of ILC2 responses, we examined a published dataset44 and analysed the effects of cAMP on ILC2 gene expression. Gene set enrichment analysis showed that, in addition to restricting cell cycle progression as reported previously,44 cAMP markedly enhanced apoptosis‐associated pathways (Figure 4G). Expression of pro‐apoptosis associated genes (e.g., Bad, Bak1, caspases, Pdcd1) were upregulated by cAMP independently of IL‐33 (Figure 4H). In contrast, IL‐33‐induced expression of anti‐apoptotic genes (e.g., Bcl2l1, Bcl3, Trp53, Ptk2) were down‐regulated by cAMP (Figure 4H). This differential cAMP‐altered expression of pro‐ and anti‐apoptotic genes contributes to overall reduced cell survival. Especially, in consistent to down‐regulation of the hallmark c‐Myc pathway gene expression by cAMP (Figure 4G), our data show that PGE₂‐cAMP significantly reduced c‐Myc protein expression in ILC2s (Figure 4I).

Moreover, IL‐33 elevated expression of the STAT5 pathway genes including cytokines/receptors (Il2ra, Il9, Csf2, and Csf2r), JAK family members (Jak1 and Tyk2) and Stat5a and Stat5b themselves, which were all down‐regulated by cAMP (Figure 4H). Indeed, PGE₂ and cAMP restrained STAT5 activation in ILC2s cultured with IL‐2/IL‐7 and IL‐33 (Figure 4J). STAT5 is important for ILC2 homeostasis, accumulation and effector function at lymphoid and nonlymphoid tissues.48, 49, 50 Similar to findings observed in T cells,51 cAMP up‐regulated gene expression involved in the IL‐12 pathway (e.g., Il12rb2) and the interferon pathway, the latter included cytokine receptors (Il18r, Il18rap, Ifng, Ifngr1, Ifnar1, and Ifnar2) and downstream interferon‐stimulated or gamma‐activated genes (Ifit1, Irf8, Isg15, and Hif1a) (Figure 4H). Interferon signaling down‐regulates ILC2 activity and restricts type 2 immune responses.52, 53 IL‐33 alone had no impacts on IL‐12 and interferon pathway gene expression (Figure 4H). The interferon‐ and IL‐12‐activated STATs (e.g., STAT1/2/4) repress ILC2 activation by antagonizing STAT5.54 Similarly, cAMP enhanced STAT3 expression (Figure 4H) which usually competes with STAT5 for binding to gene promoters.55 Furthermore, STAT5 is recruited to the c‐Myc enhancer,56 which may contribute to cAMP down‐regulation of the MYC pathway (Figure 4G,I). MYC plays an important role in cellular metabolism, and it has been reported to be critical in ILC2‐mediated airway inflammation.57 In addition, cAMP upregulated cellular senescence signature genes (Eomes, Foxo1, Havcr3, and CD244a) and lymphocyte exhaustion markers (Tnfrsf18, Tigit, Ctla4 and Pdcd1) but down‐regulated cyclin‐dependent kinase inhibitors (Cdkn2a and Cdkn2b) (Figure 4K), suggesting that cAMP may foster ILC2 senescence and exhaustion.

In this study, we have established that PGE₂ reduces ILC2 survival, cell size, proliferation, and c‐Myc expression (Figure 4), all are hallmark events of cellular metabolism. Similarly, in vivo activation of EP4 and, possibly EP2, by their respective agonists reduced Ki‐67‐expressing ST2⁺ ILC2s in lungs of mice treated with IL‐33 (Figure 5A,B). In contrast, EP2 deficiency increased Ki‐67⁺ST2⁺ ILC2s in the lung (Figure 5C,D). Furthermore, IL‐33 stimulation induced phosphorylation of the ribosomal protein S6 (pS6), a key signature of cellular metabolic activation. This is again reverted by PGE₂ and cAMP (Figure 5E,F). These results may imply changes in energy metabolism of ILC2s by PGE₂‐cAMP signaling. To examine if PGE₂ directly regulates ILC2 metabolism, we measured metabolic changes in cultured lung ILC2s using the Seahorse assays. Indeed, PGE₂‐reduced oxygen consumption rate (OCR, an indicator of mitochondrial phosphorylation), extracellular acidification rate (ECAR, an indicator of aerobic glycolysis) and glycolytic proton efflux rate (PER, a real‐time indicator of glycolysis rate) of ILC2s (Figure 5G,H). Furthermore, PGE₂ suppressed not only basal glycolysis (i.e., physiological mitochondrial respiration) but also compensatory glycolysis when mitochondrial respiration is inhibited (Figure 5I). Thus, our results suggest that PGE₂ directly inhibits IL‐33‐dependent ILC2 energy metabolism in vitro.

To determine whether PG signaling controls ILC2 responses through inhibiting ILC2 metabolism in vivo, we treated WT C57BL/6 mice with IL‐33 and indomethacin (Figure 5J). Some mice received 2‐deoxy‐D‐glucose (2‐DG), a glucose analogue that inhibits glycolysis through blocking glucose hexokinase. 2‐DG has no impact on IL‐33‐ or helminth infection‐induced ILC2 activation,58, 59 but it reduces ILC2 responses when ILC2s are over‐activated such as in PD‐1‐deficient mice.59 In agreement with these previous reports,58, 59 2‐DG affected neither IL‐33‐induced ILC2 accumulation and type 2 cytokine production nor eosinophil recruitment in mice with intact PG signaling (i.e., without indomethacin‐treatment) (Figure 5K–M). However, in indomethacin‐treated mice where biosynthesis of endogenous PGs including PGE₂ was blocked, 2‐DG strikingly reduced ILC2 and Th2 numbers and type 2 cytokine production in the airway (Figure 5K,L). Accordingly, overall lung inflammation was also decreased by 2‐DG in indomethacin‐treated mice as evidenced by reduction of both eosinophils and neutrophils in the BAL fluid (Figure 5M). 2‐DG did not affect lung ILC2 numbers nor cytokine production in both indomethacin‐ or vehicle‐treated mice except the reduction of lung eosinophils in mice treated with indomethacin but not vehicle control (Figure S11). As indomethacin inhibits production of all PGs, we wondered if PGE₂, especially the EP2 receptor, inhibits lung ILC2 responses through inhibiting ILC2 metabolism. To address this, we treated mice with PF‐04418948, a highly selective EP2 antagonist,60 with or without 2‐DG (Figure 5J). In agreement with our observations in EP2KO mice, the EP2 antagonist also significantly increased numbers of ILC2s, Th2 cells, eosinophils and neutrophils in the BAL fluid (Figure 5N,O). Similar to findings in mice treated with indomethacin, augmented ILC2 responses in EP2 antagonist‐treated mice were also markedly reduced by 2‐DG although it remained no effects on ILC2 responses in mice treated with vehicle control (Figure 5N,O). Taken together, these results suggest that PGE₂ impedes ILC2 energy metabolism and constrains IL‐33‐induced lung type 2 immune responses.