Cyclooxygenase-2 negatively regulates innate lymphoid cell type 2 differentiation and function during allergic lung inflammation

Asthma is a chronic airway disease characterized by bronchial hyperresponsiveness, reversible airflow limitation, and airway inflammation. Classically, acquired immunity and antigen-specific Th2 cells, which secrete cytokines such as IL-4, IL-5, and IL-13, are believed to play pivotal roles in the pathogenesis of asthma (1). IL-4 induces differentiation of naive Th cells to Th2 cells; IL-5 induces development, recruitment, and activation of eosinophils (2); and IL-13 promotes the hyperproduction of mucus and airway remodeling (3).

Type 2 innate lymphoid cells (ILC2) have emerged as a major component of type 2 inflammation in mice and humans. These cells were first identified in 2010 as IL-13–producing non-B/non-T innate effector cells crucial in type 2 immune responses during helminth infections (4). ILC2 cells develop from common lymphoid progenitor cells (CLP) in the bone marrow and require the inhibitor of DNA binding 2 (Id2), Notch, and IL-7 signals (5–8). In addition, ILC2 cells can be generated from CLP cells in vitro in the presence of Delta-like ligand-expressing OP9 stromal cells and exogenous IL-7 and IL-33 (9). ILC2 cells are enriched in the blood, lung, and other tissues of allergic individuals, where they initiate and exacerbate type 2 immune responses (10, 11) and secrete large amounts of IL-5 and IL-13 that are primarily responsible for induction of asthmatic pathology (12–16). In addition to their cytokine-secreting function, recent reports have identified ILC2 cells as antigen-presenting cells (17, 18). For example, ILC2-driven antigen presentation is necessary for development of anthelmintic Th2 responses and significantly enhances allergic airway inflammation in response to the allergen ovalbumin (OVA) (19). Exposure to the fungal allergen Alternaria alternata induces release of the epithelial alarmin IL-33, which activates lung-resident ILC2s to produce type 2 cytokines that initiate eosinophilic airway inflammation (20). The precise molecular signaling pathways and the role of prostaglandins in regulating these ILC2 cell functions remain unclear.

In mice, ILC2 cells express receptors for epithelial cytokines such as IL-25, IL-33, and thymic stromal lymphopoietin (TSLP) that regulate their function. ILC2 cells respond to these cytokines by producing IL-5 and IL-13, which potentiate disease phenotypes in several inflammatory models (21). The relative contribution and importance of each of these cytokines in directing the ILC2 cell phenotype have not been established. It has been suggested that IL-33 may be more potent than IL-25 and TSLP in directing allergic inflammation in the lungs via ILC2 cell activation, while redundancy among these 3 cytokines was observed in models examining chronic inflammation and fibrosis (22).

Cyclooxygenase-2 (COX-2) is expressed at low levels under basal conditions but becomes induced by various stimuli such as bacterial LPS, proinflammatory cytokines, and growth factors. COX-2 metabolizes arachidonic acid (AA) to prostaglandin (PG) H₂, which is further metabolized by specific PG synthases to the biologically active lipids known as prostanoids, including PGD₂, PGE₂, PGF_2α, prostacyclin (PGI₂), and thromboxane (TXA2) (23). These prostanoids elicit various effects through their respective cognate G-protein coupled receptors called DP, EP, FP, IP, and TP, respectively. These receptors have been found on various cells throughout the immune system, including T cells, B cells, DCs, and macrophages (24). PGs play an important role in modulating the inflammatory response, especially in allergic airway inflammation, chronic infections, and cancer (23–25). COX-2, PG synthases, and prostanoid receptors represent important targets for treating a wide range of human diseases from asthma to autoimmunity to cancer (25).

COX-2 plays a critical but distinct role in T cell maturation and lymphocyte development (26). Our previous work showed that COX-2 regulates Th9 and Th17 cell differentiation and function during allergic lung inflammation (27–29). Separate studies have demonstrated that PGD₂, PGE₂, or PGI₂ can either enhance or impair ILC2 cell differentiation and function (30–32). Given these apparently contradictory effects of PGs, it remains unknown whether COX-2 inhibition or genetic disruption would positively or negatively regulate ILC2 cell differentiation or function during allergic lung inflammation. In this study, we used COX-2^–/– mice, EP2^–/– mice, selective COX-2 inhibitors, and selective EP2 receptor agonists and antagonists to examine lung ILC2 cell differentiation and function during OVA- and Alternaria-induced allergic lung inflammation. Our results indicate that, during allergic lung inflammation, COX-2–derived PGE₂ directly signals through the EP2 receptor to suppress epithelial IL-33 and TSLP expression and reduce ILC2 cell differentiation in vivo and in vitro. Independent of its effect on epithelial cytokines, COX-2–derived PGE₂ also directly impairs ILC2 cell differentiation in vitro. Taken together, these results suggest that altered PGE₂ signaling is the most determinative factor for ILC2 differentiation and function in allergic mice with COX-2 inhibition or COX-2 genetic disruption. Defining the role of COX-2–derived prostanoids in ILC2 cell differentiation and function may lead to a better understanding of the role of nonsteroidal antiinflammatory drugs (NSAIDs), and selective COX-2 inhibitors in particular, in asthma progression; it may also lead to the development of new therapies for allergic lung diseases in humans.

COX-2^–/– mice have increased lung ILC2 cells during OVA-induced allergic inflammation in vivo. To investigate the role of COX-2 on lung ILC2 cells in vivo, we first compared lung ILC2 cell numbers after OVA-induced allergic lung inflammation between COX-2^+/+ and COX-2^–/– mice. Mice were sensitized with OVA (20 μg/mouse/day) in the presence of adjuvant (Alum) on days 0 and 1. After 14 days, mice were exposed to vehicle or 1% OVA/PBS via inhalation for 5 consecutive days (30 min per day). Twenty-four hours after the last exposure, mouse lungs, BALF, blood, and lymph nodes were collected and analyzed (Figure 1A). ILC2 cells (defined as CD45⁺Lin^– CD90.2⁺CD161^–ROR_γt^–CD11b^–) in the lung, BALF, blood, and lymph nodes were quantified as a percentage of CD45⁺Lin^– cells as previously described (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.189205DS1) (33). Compared with vehicle, OVA sensitization/exposure significantly increased lung ILC2 cells (Figure 1B), as well as the percentage of ILC2 cells expressing the cytokines IL-5 and IL-13 and the transcription factor GATA3 (Figure 1D and Supplemental Figure 2A). Importantly, ILC2 cells were significantly increased in COX-2^–/– lungs (9.3% ± 0.7%) compared with COX-2^+/+ lungs (4.7% ± 0.4%) after OVA challenge (Figures 1, B and C). ILC2 cells from lungs of COX-2^–/– mice also had a significantly higher percentage of cells expressing IL-5, IL-13, and GATA3 than ILC2 cells from lungs of COX-2^+/+ mice (Figure 1D and Supplemental Figure 2B). ILC2 cells in BALF, blood, and lymph nodes were not significantly different between COX-2^+/+ and COX-2^–/– mice after OVA exposure (Supplemental Figure 1B). Together, these results indicate that disruption of COX-2 results in increased lung ILC2 cells with increased IL-5, IL-13, and GATA3 expression during OVA-induced allergic inflammation in vivo.

Figure 1

COX-2^–/– lungs have increased ILC2 cells during allergic inflammation induced by OVA and Alternaria in vivo. (A) Schematic of the in vivo allergic lung inflammation models used. COX-2^+/+ and COX-2^–/– mice were sensitized to OVA in the presence of adjuvant (Alum) via i.p. injection (Days 0 and 1) and then exposed to 1% OVA via aerosol for 5 consecutive days (Day 14 to 18). Alternatively, mice were exposed to Alternaria extract by i.n. injection (20 μg/mouse/day) for 4 days. Mice were sacrificed 24 hours after the last exposure and lungs were collected for analysis. (B) ILC2 cells as a percentage of CD45⁺Lin^– cells were quantified by FACS in lungs from vehicle- and OVA-treated COX-2^+/+ and COX-2^–/– mice (n = 5 per group, *P < 0.05). (C) Representative FACS plots of OVA-treated COX-2^+/+ and COX-2^–/– lungs showing ILC2 cells (GATA3⁺ CD11b^–). (D) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺KLRG1⁺ cells in lungs from OVA-treated COX-2^+/+ and COX-2^–/– mice were quantified by FACS analysis. (E) After PBS or Alternaria exposure, ILC2 cells as a percentage of CD45⁺Lin^– cells, or IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺KLRG1⁺ cells in lungs from COX-2^+/+ and COX-2^–/– mice were quantified by FACS analysis. n = 3–5/group, *P < 0.05 by 2-way ANOVA.

COX-2^–/– mice have increased lung ILC2 cells during Alternaria-induced allergic inflammation in vivo. COX-2 disruption also resulted in increased ILC2 cells in response to Alternaria. COX-2^+/+ and COX-2^–/– mice were exposed to Alternaria extract (20 μg/mouse/day i.n.) for 4 days (Figure 1A) similar to previous experiments (34). Twenty-four hours after the last exposure, ILC2 cells from mouse lungs, BALF, blood, and lymph nodes were analyzed. Compared with vehicle, Alternaria exposure increased lung ILC2 cells, and the percentage of ILC2 cells that express IL-5, IL-13, and GATA3 (Figure 1E). Importantly, COX-2^–/– lungs had increased ILC2 cells and increased ILC2 cells expressing IL-5, IL-13, and GATA3 compared with COX-2^+/+ lungs after Alternaria exposure (Figure 1E). ILC2 cells in BALF, blood, and lymph nodes were also higher in COX-2^–/– mice compared with COX-2^+/+ mice after Alternaria (Supplemental Figure 1C). Together, these results indicate that COX-2 disruption results in increased ILC2 cells with increased expression of IL-5, IL-13, and GATA3 during Alternaria-induced allergic inflammation in vivo.

COX-2 disruption increases ILC2 cell differentiation from CLP cells in vitro. To examine the role of COX-2 in ILC2 cell differentiation in vitro, we isolated CD127⁺CD135⁺Sca-1 CLP cells from bone marrows (Supplemental Figure 3A) of COX-2^+/+ and COX-2^–/– mice. These CLP cells were then differentiated to ILC2 cells via coculture with OP9-DL1 feeder cells (35, 36) over 10 days and stimulation with SCF, IL-7, and IL-33 (37, 38) as depicted in Supplemental Figure 3B. Treatment with SCF, IL-7, and IL-33 increased the level of COX-2 (Ptgs2) mRNA by ~50% relative to vehicle (Figure 2A). Importantly, a significantly higher percentage of CLP cells from COX-2^–/– mice differentiated to ILC2 cells than CLP cells from COX-2^+/+ mice (1.350 ± 0.13% versus 0.6 ± 0.12%), respectively; P < 0.05) (Figure 2B). Moreover, a higher percentage of COX-2^–/– CLP cells differentiated with SCF, IL-7, and IL-33 expressed IL-5, IL-13, and GATA3 compared with COX-2^+/+ CLP cells (Figure 2C). In agreement with these observations, CLP cells treated with the selective COX-2 inhibitor NS-398 expressed higher levels of IL-5, IL-13, and GATA3 after SCF, IL-7, and IL-33 treatment (Figure 2D). Together, these data suggest that COX-2 is induced after SCF, IL-7, and IL-33 treatment and that COX-2 disruption increases ILC2 cell differentiation from CLP cells in vitro.

Figure 2

COX-2 disruption increases ILC2 cell differentiation from bone marrow–derived CLP cells in vitro. CLP cells were isolated from bone marrows of COX-2^+/+ and COX-2^–/– mice and treated with either vehicle or SCF + IL-7 + IL-33 in vitro to induce ILC2 cell differentiation. (A) COX-2 (Ptgs2) mRNA levels measured by qPCR in CLP cells from COX-2^+/+ mice treated with either vehicle or SCF + IL-7 + IL-33 in vitro. (B) FACS analysis of GATA3⁺ cells from COX-2^+/+ and COX-2^-/- CLP cells treated with either vehicle or SCF + IL-7 + IL-33 in vitro. (C) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺KLRG1⁺ cells derived from COX-2^+/+ and COX-2^–/– CLP cells. (D) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺KLRG1⁺ cells derived from COX-2^+/+ CLP cells differentiated in the presence of vehicle or NS398. n = 3–5 per group, *P < 0.05 by t test (A and C) or 1-way ANOVA (B and D).

COX-2 products inhibit ILC2 cell differentiation from bone marrow–derived CLP cells in vitro. We hypothesized that COX-2 negatively regulates ILC2 cell differentiation by generating prostanoids, which act through their cognate receptors (39). To define which prostanoids are able to regulate ILC2 cell differentiation of CLP cells, we first examined the expression of the prostanoid receptors during ILC2 differentiation in vitro. SCF, IL-7, and IL-33 treatment of CLP cells significantly induced expression of EP2 (Ptger2), EP3 (Ptger3), IP (Ptgir), and TP (Tbxa2r) receptors (Supplemental Figure 4A). The increases in expression of EP2, EP3, IP, and TP receptors were confirmed by surface staining and FACS analysis (Supplemental Figure 4B). EP1 (Ptger1), EP4 (Ptger4), and FP (Ptgfr) receptors were expressed but were either decreased or unchanged during ILC2 cell differentiation. DP (Ptgdr) and DP2/CRTH2 (Ptgdr2) receptors were expressed at or below the limit of detection (Supplemental Figure 4A). This suggests that PGE₂-EP_1-4, PGF_2α-FP, PGI₂-IP, and/or TXA₂-TP signaling might regulate CLP cell differentiation to ILC2 cells.

To assess which prostanoids regulate ILC2 cell differentiation, COX-2^+/+ CLP cells were differentiated to ILC2 cells in vitro with SCF, IL-7, and IL-33 in the presence or absence of vehicle (ethanol) or 5–1,000 nM PGE₂, PGD₂, PGI₂, PGF_2α, or cTXA₂. While high concentrations of several prostanoids reduced ILC2 cell differentiation, only PGE₂ suppressed IL-5, IL-13, and GATA3 protein expression and Il5, Il13, and Gata3 mRNA levels at biologically relevant concentrations as low as 50 nM (Figure 3A and Supplemental Figures 5 and 6). These data suggest that PGE₂ is likely the most relevant COX-2–derived prostanoid, which inhibits ILC2 cell differentiation. Therefore, we focused on the role of PGE₂-EP receptor signaling in regulating ILC2 cell differentiation.

Figure 3

Effect of prostanoids and receptor modulators on ILC2 cell differentiation in vitro. CLP cells were isolated from bone marrows of COX-2^+/+ mice and treated with either vehicle or SCF + IL-7 + IL-33 in vitro to induce ILC2 cell differentiation. (A) The percentage of IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells was quantified by FACS in COX-2^+/+ CLP cells differentiated in vitro in the presence of vehicle or 100 nM of PGD₂, PGI₂, PGF_2α, PGE₂, or TXA₂. (B) The percentage of IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells was quantified by FACS in COX-2^+/+ or COX-2^–/– CLP cells differentiated in vitro in the presence of vehicle or 100 nM of PGE₂. (C) The percentage of IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells was quantified by FACS in COX-2^+/+ CLP cells differentiated in vitro in the presence of vehicle, 100 nM PF-04418948 (EP2 antagonist), or 100 nM MF498 (EP4 antagonist). n = 3–5 per group, *P < 0.05 by 1-way ANOVA (A and C) or multiple t tests with Holm Šídák correction (B).

PGE₂-EP2 signaling is critical in regulating lung ILC2 cell differentiation in vitro and in vivo. To further examine the role of PGE₂ in suppression of ILC2 cell differentiation, we differentiated CLP cells isolated from COX-2^+/+ and COX-2^–/– mice to ILC2 cells in the presence or absence of PGE₂ in vitro. PGE₂ suppressed differentiation of COX-2^+/+ CLP cells to ILC2 cells as determined by the percentage of ILC2 cells expressing IL-5, IL-13, or GATA3 protein (Figure 3B). PGE₂ also suppressed the enhanced differentiation of COX-2^–/– CLP cells to ILC2 cells (Figure 3B). To examine the role of PGE₂-EP2 receptor signaling in ILC2 cell differentiation, we treated COX-2^+/+ CLP cells with the selective EP2 agonist butaprost. The EP2 agonist significantly inhibited ILC2 cell differentiation (Supplemental Figure 7). Conversely, treatment with an EP2 antagonist (PF-04418948) increased ILC2 cell differentiation (Figure 3C). In contrast, the EP4 antagonist MF498 suppressed ILC2 cell differentiation (Figure 3C). These data suggest that PGE₂ inhibits ILC2 cell differentiation primarily through EP2 receptor activation in vitro.

To examine the role of PGE₂ in regulating lung ILC2 cell differentiation in vivo, PGE₂ was delivered by osmotic minipumps during OVA-induced allergic lung inflammation (Figure 4A). While vehicle-treated COX-2^–/– mouse lungs had increased ILC2 cells compared with vehicle-treated COX-2^+/+ mouse lungs, PGE₂ treatment decreased COX-2^–/– mouse lung ILC2 cells to levels that were similar to those observed in COX-2^+/+ mice (Figure 4B). Consistent with these data, the increased expression of Il5 and Il13 mRNAs in COX-2^–/– lungs relative to COX-2^+/+ lungs were also attenuated by administration of PGE₂ (Figure 4C). These data suggest that the lung ILC2 cell phenotype during OVA-induced allergic lung inflammation in COX-2^–/– mice can be rescued by PGE₂ treatment in vivo.

Figure 4

PGE₂, EP2 agonist, EP2 antagonist, and COX-2 inhibitor regulate ILC2 cells during allergic lung inflammation in vivo. (A) COX-2^+/+ and COX-2^–/– mice were treated with either vehicle, PGE₂, butaprost, PF-04418948, or CAY10404 via Alzet minipump during OVA-induced allergic lung inflammation. (B) ILC2 cells were quantified by FACS in lungs from allergic COX-2^+/+ and COX-2^–/– mice treated with vehicle and COX-2^–/– mice treated with PGE₂. n = 5–7 per group, *P < 0.05 by ANOVA. (C) Il5, Il13, and Il33 mRNA levels were quantified by qPCR in lungs of allergic COX-2^+/+ and COX-2^–/– mice treated with vehicle and allergic COX-2^–/– mice treated with PGE₂. (D and E) ILC2 cells (D) and IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells (E) in lungs of COX-2^+/+ mice treated with butaprost (EP2 agonist), PF-04418948 (EP2 antagonist), or CAY10404 (COX-2 inhibitor) during OVA-induced allergic lung inflammation. n = 6–9 per group, *P < 0.05 by 1-way ANOVA.

We next tested the effects of the selective EP2 receptor agonist (butaprost), the selective EP2 receptor antagonist (PF-04418948), and the selective COX-2 inhibitor (CAY10404) on regulation of lung ILC2 cells during OVA-induced allergic lung inflammation in vivo. The EP2 receptor agonist significantly decreased lung ILC2 cells after OVA exposure (Figure 4D) and tended to reduce IL-5⁺ and IL-13⁺ ILC2 cells. Conversely, administration of either the EP2 receptor antagonist or the COX-2 inhibitor significantly increased lung ILC2 cells and increased IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells (Figure 4, D and E).

To confirm the function of the EP2 receptor in the regulation of ILC2 cells in vivo, we induced allergic lung inflammation in EP2^+/+ and EP2^–/– mice. After OVA treatment, lungs from allergic EP2^–/– mice had significantly increased ILC2 cells compared with lungs from allergic EP2^+/+ mice (12.8 ± 4.2 versus 5.4 ± 2.3, n = 6; P < 0.05) (Figure 5A). In addition, lungs from OVA-treated EP2^–/– mice had increased IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells than lungs from OVA-treated EP2^+/+ mice (Figure 5B). Similarly, lungs from Alternaria-treated EP2^–/– mice had significantly increased ILC2 cells compared with lungs from Alternaria-treated EP2^+/+ mice (12.4 ± 2.2 versus 4.5 ± 1.8, n = 3–5; P < 0.05) (Figure 5C). Lungs from Alternaria-treated EP2^–/– mice also had increased IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells than lungs from Alternaria-treated EP2^+/+ mice (Figure 5C). Furthermore, in vitro differentiation of EP2^–/– bone marrow–derived CLP cells to ILC2 cells was significantly greater than differentiation of EP2^+/+ bone marrow–derived CLP cells as measured by increased expression of Il5, Il13, and Gata3 mRNAs (Figure 5D). Together these data indicate that COX-2–derived PGE₂ inhibits lung ILC2 cells via signaling through the EP2 receptor in vitro and in vivo.

Figure 5

EP2^–/– mice have increased ILC2 cells during allergic lung inflammation in vivo and in vitro. (A) Lungs from OVA-treated EP2^–/– mice have increased ILC2 cells compared with lungs from OVA-treated EP2^+/+ mice as determined by FACS. n = 6 per group, *P < 0.05. (B) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells were quantified by FACS in lungs from OVA-treated EP2^+/+ and EP2^–/– mice. n = 6 per group, *P < 0.05. (C) Lungs from Alternaria-treated EP2^–/– mice have increased ILC2 cells compared with lungs from Alternaria-treated EP2^+/+ mice as determined by FACS. n = 3–5 per group, *P < 0.05. (D) After in vitro differentiation with SCF + IL-7 + IL-33, bone marrow-derived CLP cells from EP2^–/– mice have increased Il5, Il13, and Gata3 mRNA levels compared with bone marrow–derived CLP cells from EP2^+/+ mice. n = 3 per group, *P < 0.05 by t test (A and B) or 2-way ANOVA (C and D).

Allergic COX-2^–/– mice have altered lung ILC2 cell gene expression. To examine whether COX-2 disruption alters ILC2 cell gene expression during allergic lung inflammation, we performed single-cell RNA-seq on isolated lung cells from COX-2^+/+ and COX-2^–/– mice exposed to OVA as described in Figure 1A. Uniform manifold approximation and projection (UMAP) dimensional reduction of 6,955 lymphoid T cells identified 18 distinct cell clusters (Figure 6A). These clusters, including the ILC2 cell cluster (red, #8), were identified by characteristic lineage specific markers (Supplemental Figure 8A). After OVA exposure, the number of ILC2 cells in COX-2^–/– mice was increased relative to COX-2^+/+ mice (Figure 6B and Supplemental Figure 8B). OVA exposure caused changes in expression of 323 genes in ILC2 cells from COX-2^+/+ mice (155 genes downregulated, 168 genes upregulated), including induction of IL-2, Th2, and Th17 signaling pathways (Supplemental Figure 8C, Supplemental Figure 9A, and Supplemental Table 2). By comparison, OVA exposure caused changes in expression of 1,386 genes in ILC2 cells from COX-2^–/– mice (612 genes downregulated, 774 genes upregulated), including induction of T cell migration and cytokine stimulus pathways (Supplemental Figure 8C and 9B). Relative to ILC2 cells from allergic COX-2^+/+ mice, ILC2 cells from allergic COX-2^–/– mice had differential expression of 260 genes (118 genes downregulated, 142 genes upregulated) (Figure 6C). Pathway analysis suggested that ILC2 cells from allergic COX-2^–/– mice had upregulation of multiple pathways involved in leukocyte migration and proliferation (Figure 6D and Supplemental Table 2). Significantly increased expression of Ccl5 (RANTES) in COX-2^–/– ILC2 cells is a key component of many of these pathways. ILC2 cells from allergic COX-2^–/– mice had downregulation of several pathways involved in apoptosis, telomere maintenance, and both lipid and protein phosphorylation (Figure 6D and Supplemental Table 2). Together, these data suggest that COX-2 disruption, in addition to increasing the number of ILC2 cells in allergic lungs, also alters the ability of ILC2 cells to proliferate, migrate, and produce or respond to cytokine signals.

Figure 6

COX-2 disruption alters ILC2 gene expression profiles during allergic lung inflammation. Lung cells were isolated from vehicle and OVA-exposed COX-2^+/+ and COX-2^–/– mice as shown in Figure 1A. Cells from 3 mice per treatment group were pooled and single-cell RNA-seq was performed. (A) Uniform manifold approximation and projection (UMAP) of 6,955 lymphoid cells identified 18 cell clusters, including ILC2 cells (red, #8). (B) The percentage of lymphoid cells identified as ILC2 cells in allergic COX-2^+/+ and COX-2^–/– mice is shown. (C) Volcano plot (–Log₁₀ [P value] versus Log₂ FC) indicates genes differentially expressed between OVA-exposed COX-2^+/+ and COX-2^–/– ILC2 cells. (D) Normalized enrichment scores (NES) from GSEA indicate the top 5 pathways significantly increased (orange) or decreased (blue) in OVA-exposed COX-2^–/– ILC2 cells versus OVA-exposed COX-2^+/+ ILC2 cells.

COX-2–derived PGE₂ inhibits lung epithelial cytokine expression in vitro and in vivo. Activated epithelial cells produce IL-33, TSLP, and IL-25 to induce lung ILC2 cell differentiation during OVA-induced allergic lung inflammation (40). While vehicle-treated COX-2^–/– mouse lungs had increased Il33 mRNA compared with vehicle-treated COX-2^+/+ mouse lungs, PGE₂ treatment decreased COX-2^–/– mouse lung Il33 to levels that were similar to those observed in COX-2^+/+ mice (Figure 4C). To further investigate whether COX-2–derived PGE₂ regulates these epithelial cytokines in the allergic lung, we examined IL-33, TSLP, and IL-25 expression in frozen lung tissue sections after OVA sensitization/exposure by confocal microscopy using immunofluorescence conjugated IL-33 and TSLP antibodies (Figure 7A and Supplemental Figure 11). Intensity quantification of the resulting confocal images indicates that COX-2^–/– lung epithelial cells have significantly higher IL-33 and TSLP expression compared with COX-2^+/+ lung epithelial cells (Figure 7A). We consistently observed increased Il33 and/or Tslp mRNA levels in OVA-treated COX-2^–/– lungs compared with OVA-treated COX-2^+/+ lungs (Figure 4C and Figure 7B). Similarly, Alternaria exposure increased Il33 and Tslp mRNA levels in COX-2^+/+ lungs. This induction was further enhanced in Alternaria-treated COX-2^–/– lungs (Figure 7C). In most of our experiments, Il25 mRNA levels were either unchanged or at the limit of detection (not shown). Together, these data suggest that COX-2 negatively regulates lung epithelial cytokines during OVA- and Alternaria-induced allergic lung inflammation.

Figure 7

COX-2 suppresses airway epithelial IL-33 and TSLP expression in vivo and in vitro. (A) After OVA-induced allergic inflammation, lung epithelial IL-33 and TSLP were visualized by confocal microscopy using anti-mouse antibodies to EpCAM (green), IL-33 or TSLP (red), and DAPI (blue), and the average pixel intensity in Epcam⁺ cells was quantified. Epithelial expression of both IL-33 and TSLP are increased in COX-2^–/– mice relative to COX-2^+/+ mice. (B and C) Levels of Il33 and Tslp mRNAs in lungs from OVA-treated (B), and PBS or Alternaria-treated (C), COX-2^+/+ and COX-2^–/– mice were measured by qPCR. (D) Purified COX-2^–/– airway epithelial cells have increased levels of Il33 and Tslp after LPS treatment. In contrast, levels of Il25 remain unchanged. (E) PGE₂ significantly inhibits Il33 and Tslp expression in purified airway epithelial cells from COX-2^+/+ mice treated with LPS. n = 3–12 per group, *P < 0.05 by t test (A and B), multiple t tests (C and D) or 1-way ANOVA (E).

To quantify the effect of COX-2 on epithelial cytokine production, we purified airway epithelial cells from COX-2^+/+ and COX-2^–/– lungs and examined epithelial cytokine production after a 4-hour treatment with bacterial LPS in vitro. LPS treatment increased Il33 and Tslp mRNA levels in both COX-2^+/+ and COX-2^–/– epithelial cells; however, COX-2^–/– epithelial cells had a significantly greater increase in Il33 and Tslp expression than COX-2^+/+ epithelial cells (Figure 7D). Importantly, treatment of purified airway epithelial cells with PGE₂ attenuated LPS-induced Il33 and Tslp expression (Figure 7E). We also examined prostanoid receptor expression in lung epithelial cells after LPS treatment. While all receptors were detectable, LPS significantly induced lung epithelial expression of the IP (Ptgir) and EP2 (Ptger2) receptors (Supplemental Figure 10). Taken together, these data suggest that COX-2–derived PGE₂ can also inhibit expression of the lung epithelial cytokines (IL-33 and TSLP), which promote ILC2 cell differentiation.

IL-33 increases lung ILC2 cells in both COX-2^+/+ and COX-2^–/– mice in vivo. COX-2–derived PGE₂ downregulates both the epithelial production of ILC2-inducing IL-33 and TSLP, and the ability of CLP cells to differentiate to ILC2 cells in response to those epithelial cytokines. We sought to determine whether exogenous IL-33 would restore the ILC2 cell phenotype in COX-2^–/– lungs in vivo. We gave IL-33 intranasally to COX-2^+/+ and COX-2^–/– mice for 5 consecutive days and analyzed ILC2 cell numbers and function 24 hours after the fifth dose (Figure 8A). Both COX-2^+/+ and COX-2^–/– lungs exhibit very low ILC2 cell numbers with vehicle treatment (1.03 ± 0.50 versus 0.87 ± 0.25, n = 3). In contrast, IL-33 treatment significantly increased lung ILC2 cells in both COX-2^+/+ and COX-2^–/– lungs to a comparable degree (3.2 ± 1.1 versus 3.3 ± 0.6, n = 6) (Figure 8B). IL-33 also induced comparable amounts of Il5 and Il13 mRNAs in COX-2^+/+ and COX-2^–/– mice (Figure 8C). One interpretation is that exogenous IL-33 normalized the ILC2 cell phenotype in COX-2^–/– mice, which would suggest a major effect of COX-2 disruption was to increase epithelial cytokines during lung inflammation. An alternative explanation may be that the exogenous IL-33 instillation overwhelmed the ILC2 cell differentiation or recruitment such that the resulting ILC2 cell responses were no longer modulable by changes in PG signaling.

Figure 8

COX-2^+/+ and COX-2^–/– lungs have similar ILC2 cell numbers after IL-33–induced inflammation in vivo. (A) Schematic of the IL-33–induced lung inflammation model. Mice were given 1.5 μg of IL-33 in 50 μL of PBS i.n. for 5 consecutive days. Lungs were collected and analyzed 24 hours after the last IL-33 dose. (B) Lung ILC2 cell numbers were analyzed by FACS. (C) Levels of Il5, Il13, and Gata3 mRNAs during IL-33–induced lung inflammation were quantified by qPCR (n = 3–6 per group, *P < 0.05 by 1-way ANOVA).

Effect of PGE₂ and EP receptor modulators on human ILC2 cell differentiation in vitro. To determine whether PGE₂ similarly regulates human ILC2 cell differentiation, lymphocytes isolated from peripheral blood of healthy human donors were cultured with IL-33 and IL-2 to induce differentiation toward an ILC2 cell phenotype. Stimulation with IL-33 and IL-2 significantly increased the percentage of IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells compared with vehicle-treated controls (Figure 9A). Importantly, treatment with either PGE₂ or the selective EP2 receptor agonist butaprost significantly attenuated this increase. Conversely, treatment with the selective EP2 receptor antagonist PF-04418948 significantly increased the percentage of IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells, whereas the selective EP4 receptor antagonist MF498 had no significant effect (Figure 9B). Consistent with our observations in mouse ILC2 cells, these findings indicate that PGE₂ negatively regulates human ILC2 cell differentiation, predominantly through PGE₂ activation of the EP2 receptor.

Figure 9

Effect of prostanoids and receptor modulators on human ILC2 cell differentiation in vitro. CLP cells were isolated from peripheral blood of normal healthy human donors and treated with either vehicle or IL-33 + IL-2 in vitro to induce ILC2 cell differentiation. (A) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺ cells were quantified by FACS in ILC2 cells differentiated from CLP cells in vitro in the presence of vehicle, 100 nM butaprost, or 100 nM PGE₂. (B) IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells as a percentage of CD90⁺ cells were quantified by FACS in ILC2 cells differentiated from CLP cells in vitro in the presence of vehicle, 100 nM PF-04418948 (EP2 antagonist), or 100 nM MF498 (EP4 antagonist). n = 5 per group, *P < 0.05 by paired 1-way ANOVA. Lines connect samples from individual patients.

COX-2 and its eicosanoid products play important roles in regulating inflammation, innate immunity, and adaptive immunity (41). We previously reported that COX-2–derived eicosanoids regulate Th9 and Th17 cell differentiation and function during allergic lung inflammation (27, 29). ILC2 cells are non-B/non-T cells that secrete Th2-like cytokines including IL-5 and IL-13 that play critical roles in allergic lung disease pathogenesis. Therefore, we hypothesized that COX-2 and its products may also regulate lung ILC2 cells during allergic lung inflammation. We used COX-2^–/– mice, EP2^–/– mice, selective COX-2 inhibitors, and selective EP2 receptor agonists and antagonists to examine lung ILC2 cell differentiation and function during OVA- and Alternaria-induced allergic lung inflammation. We found the following: (a) allergic COX-2^–/– mice have increased lung ILC2 cells, more cells expressing the ILC2 cytokines IL-5 and IL-13, and the ILC2 transcription factor GATA3, and differential ILC2 gene expression profiles compared with COX-2^+/+ mice in vivo; (b) these effects are observed in both OVA- and Alternaria-induced models of allergic airway inflammation; (c) differentiation of CLP cells to ILC2 cells in vitro is accompanied by increased COX-2 expression; (d) CLP cells from COX-2^–/– mice or from COX-2^+/+ mice treated with a selective COX-2 inhibitor exhibit increased ILC2 cell differentiation and function in vitro; (e) COX-2^–/– epithelial cells produce increased amounts of the ILC2-inducing cytokines IL-33 and TSLP in vitro and in vivo; (f) treatment with PGE₂ or a selective EP2 receptor agonist suppresses ILC2 cell differentiation and function in vitro and in vivo; (g) allergic EP2^–/– mice have increased lung ILC2 cells and increased ILC2 cell expression of IL-5, IL-13, and GATA3 compared with EP2^+/+ mice in vivo; (h) PGE₂ suppresses lung epithelial expression of Il33 and Tslp; (i) IL-33 can restore the lung ILC2 cell phenotype in COX-2^–/– mice; and (j) PGE₂ and EP2 receptor activation similarly suppress IL-33/IL-2-induced differentiation of human peripheral blood-derived ILC2 cells. Together, these findings indicate that COX-2–derived PGE₂ regulates lung ILC2 cell differentiation and function during allergic inflammation via a mechanism that involves the EP2 receptor, IL-33 and TSLP.

NSAIDs, including COX-2 selective inhibitors, are the most commonly prescribed antiinflammatory medications in the world. COX-2 catalyzes the committed step in the biosynthesis of PGE₂, PGD₂, PGF_2α, PGI₂, and TXA₂ (42). COX-2–derived prostanoids regulate the functions of immune and other cell types through specific G-protein coupled receptors including the EP, DP, FP, IP, and TP receptors (43, 44). In addition to the relatively uncommon but severe condition of aspirin-exacerbated respiratory disease (AERD), NSAID use is more generally associated with increased risk of asthma exacerbation (45). Although the precise molecular mechanisms that underlie this association remain enigmatic, several possibilities have been suggested, including increased production of proinflammatory leukotrienes via “shunting” of AA down the lipoxygenase pathway and decreased production of antiinflammatory PGs within the airway (46). Nevertheless, it is widely accepted that the acute and transient antiinflammatory, antipyretic, and antinociceptive benefits of NSAIDs in asthmatic patients are more than outweighed by their prolonged and pronounced effects on the immune system, namely effects on T cell differentiation and function leading to increased production of proinflammatory cytokines such as IL-4, IL-5, and IL-13 that initiate and perpetuate the allergic asthmatic response.

COX-derived prostanoids regulate many immune cell types during allergic lung inflammation. For example, PGE₂, PGF_2α and PGI₂ increase Th17 cell differentiation (47), while PGE₂ and PGD₂ suppress Th9 cell differentiation (27). TXA₂ inhibits both Th2 and Th9 cell differentiation (47), possibly by inhibiting the immune synapse between antigen-presenting cells and naive CD4⁺ T cells (48). PGD₂ recruits eosinophils and mast cells to the lung during allergic inflammation and potentiates their degranulation. Conversely, PGE₂ attenuates eosinophilic and mast cell–induced inflammation (49). PGI₂ inhibits DC activation and cytokine production, which impairs antigen presentation to naive CD4⁺ T cells (50).

In our experiments, multiple PGs were capable of suppressing ILC2 cell differentiation and function in vitro; however, only PGE₂ had effects on ILC2 cells at physiologically relevant concentrations as low as 50 nM. Interestingly, we found that while ILC2 cells expressed multiple prostanoid receptors, a selective EP2 receptor agonist inhibited and a selective EP2 receptor antagonist enhanced ILC2 cell differentiation and function in vitro and in vivo. Importantly, systemic administration of PGE₂ via osmotic minipump rescued the lung ILC2 cell phenotype in allergic COX-2^–/– mice in vivo. Moreover, EP2^–/– mice had significantly increased lung ILC2 cells during allergic lung inflammation in vivo and EP2^–/– bone marrow–derived CLP cells exhibited increased ILC2 cell differentiation compared with EP2^+/+ CLP cells in vitro. Notably, these findings were not restricted to the OVA model. We observed similar increases in lung ILC2 cells and IL-5⁺, IL-13⁺, and GATA3⁺ ILC2 cells in both COX-2^–/– and EP2^–/– mice following Alternaria extract exposure. Furthermore, Alternaria-induced expression of the epithelial cytokines Il33 and Tslp was enhanced in COX-2^–/– lungs, paralleling our observations in OVA-treated mice. The consistency of these findings across 2 distinct models of allergic lung inflammation suggests that COX-2–derived PGE₂ functions as a broadly acting negative regulator of ILC2 cell responses rather than a pathway-specific regulator. Our evidence supports the concept that COX-2 negatively regulates ILC2 cells via PGE₂-EP2 signaling by multiple pathways. Importantly, PGE₂ and EP2 agonism or antagonism similarly regulate human ILC2 cell differentiation.

Our results support and expand upon prior studies showing that COX-derived prostanoids are important regulators of ILC2 cell differentiation, activation, and effector function. Maric and coworkers reported that PGE₂ suppressed blood and tonsillar ILC2 cell activation via EP2 and EP4 receptors in vitro (30). Consistent with our findings, Zhou and coworkers demonstrated that pharmacologic COX inhibition enhanced Alternaria-induced pulmonary ILC2 cell responses and IL-33 release in vivo, while both PGE2 and PGI2 suppressed epithelial IL-33 release in vitro (34). Zhou et al. also showed that PGI₂ inhibited ILC2 cell function via the IP receptor in an in vivo mouse model of fungal allergen exposure (32). In addition, PGD₂ was shown to be essential for cytokine-induced ILC2 cell activation via the CRTH2 receptor in vitro (31, 51). Thus, COX-derived prostanoids can have both ILC2-promoting and ILC2-inhibiting effects. Since COX-2^–/– mice have higher levels of ILC2 cells in vivo, our data suggest that, on balance, COX-2–derived prostanoids inhibit ILC2 differentiation, activation, and/or recruitment during allergic lung inflammation.

It is interesting to note that allergic COX-2^–/– mice exhibit a 1.5-fold increase in lung ILC2 cells compared with COX-2^+/+ mice but a much more pronounced 6- to 20-fold increase in lung expression of Il5 and Il13 in vivo. Moreover, we observed a 2-fold increase in differentiation of COX-2^–/– CLP cells to ILC2 cells compared with COX-2^+/+ and a similar 2-fold increase in expression of Il5 and Il13 in vitro. This suggests that other non-ILC2 cells may be contributing to the increased lung expression of Il5 and Il13 after allergen challenge in vivo. In this regard, we have previously shown that allergic COX-2^–/– mice exhibit increased Th2 cell differentiation and function (27). Thus, the effects of COX-2 products on proinflammatory cytokine levels in vivo are complex and likely involve influences on multiple different immune cell types.

Pathway analysis of single-cell RNA-seq data suggests that COX-2^–/– ILC2 cells have enhanced signaling in pathways that promote T cell chemotaxis and T cell/lymphocyte migration after OVA-induced allergic lung inflammation. In particular, Ccl5 was the most significantly induced gene in COX-2^–/– ILC2 cells compared with COX-2^+/+ ILC2 cells. CCL5 (RANTES) is increased in asthmatics and acts as a pulmonary chemoattractant for multiple cell types, including DCs, T cells, and eosinophils during allergic inflammation (52, 53). Thus, COX-2^–/– lungs have both an increased number of ILC2 cells and ILC2 cells that more strongly recruit other cells types that exacerbate allergic inflammation.

Allergen-activated airway epithelial cells secrete IL-33, IL-25, and TSLP, which are known to stimulate ILC2 cell differentiation (54). We found that COX-2^–/– airway epithelial cells have increased IL-33 and TSLP expression in vivo after OVA challenge or Alternaria-induced inflammation compared with COX-2^+/+. Moreover, LPS treatment of purified airway epithelial cells from COX-2^–/– lungs in vitro resulted in increased Il33 and Tslp expression compared with COX-2^+/+. Furthermore, PGE₂ significantly attenuated airway epithelial cell induction of Il33 and Tslp following LPS treatment in vitro. These results indicate that COX-2-derived PGE₂ inhibits IL-33 and TSLP production by airway epithelial cells. Therefore, increased levels of IL-33 and TSLP produced by airway epithelial cells in allergic COX-2^–/– mice likely contribute to the ILC2 cell phenotype. The fact that IL-33 administration was able to induce similar numbers of ILC2 cells in lungs of both COX-2^+/+ and COX-2^–/– mice further suggests a central role of this epithelial-derived cytokine in the observed ILC2 cell phenotype.

Our results are consistent with a model that involves effects of COX-2–derived PGE₂ on both epithelial cells and CLP cells (Figure 10). OVA exposure leads to upregulation of COX-2 in both epithelial cells and CLP cells. In COX-2^+/+ epithelial cells, increased COX-2 results in increased PGE₂ production and increased signaling through the EP2 receptor which results in decreased production of IL-33 and TSLP, leading to reduced differentiation of CLP cells to ILC2 cells. The increased PGE₂ also has a direct effect on CLP cells by suppressing their ability to differentiate to ILC2 cells in response to epithelial-derived IL-33 and TSLP. In contrast, COX-2^–/– epithelial and CLP cells are unable to upregulate COX-2 and, therefore, produce less PGE₂ in response to OVA. This results in reduced signaling through the EP2 receptor, leading to increased production of IL-33 and TSLP and to increased differentiation of CLP cells to ILC2 cells.

Figure 10.

Proposed model for the regulation of lung ILC2 cells by COX-2 derived PGE₂ during allergic lung inflammation. In COX-2^+/+ mice (left), OVA or Alternaria induces upregulation of COX-2, resulting in increased PGE₂ signaling to airway epithelial and CLP cells. Enhanced PGE₂-EP2 signaling reduces production of IL-33 and TSLP from airway epithelial cells and inhibits ILC2 cell differentiation from CLP cells. In COX-2^–/– mice (right), OVA or Alternaria cannot upregulate COX-2 resulting in lower PGE₂ levels. Decreased PGE₂-EP2 signaling results in higher IL-33 and TSLP production in airway epithelial cells and reduces the suppressive effect of PGE2-EP2 signaling during ILC2 cell differentiation.

An important limitation of the current study is that the cellular source and precise mechanism of PGE₂-mediated regulation of ILC2 cell responses in vivo remain uncertain. In previous studies, COX-2^–/– BALF had ~10%–40% lower PGE₂ levels, though these results were not statistically significant (27, 55, 56). Important changes in autocrine or paracrine PG signaling may not be captured in total BALF analyses. While our in vitro findings support the capacity for autocrine PGE₂ signaling in ILC2 cells, they do not establish whether epithelial cells or ILC2 cells are a major source of PGE₂ in vivo. Similarly, our data do not conclusively establish whether autocrine PGE₂/EP2 signaling regulates epithelial IL-33 and TSLP production in vivo. Future studies with cell-selective disruption of COX-2 and/or EP2 in ILC2 and epithelial cells will be required to address these possibilities.

In conclusion, we show that COX-2–derived prostanoids negatively regulate lung ILC2 cell differentiation and function during allergic lung inflammation through 2 separate pathways. PGE₂ signaling via the EP2 receptor inhibits airway epithelial cell formation of IL-33 and TSLP to limit the major epithelial cytokine drivers of ILC2 cell differentiation. PGE₂ also directly suppresses CLP cell differentiation into ILC2 cells. While multiple prostanoids have been shown to regulate ILC2 cell differentiation and function, our studies suggest that COX-2–derived PGE₂ is the predominant PG involved during OVA- and Alternaria-induced allergic lung inflammation. Our findings also suggest the possibility that COX-2–derived PGE₂ may represent a potentially new therapeutic target for allergic lung disease in humans.

Sex as a biological variable. Only male mice were used in this study to reduce variability. Female rodents are known to have increased allergic responses compared with males (57). The effect of COX-2 disruption or COX-2 inhibition is robust in both sexes. We expect, but cannot definitively conclude, that these findings are relevant to females.

Mice. C57BL/6J mice were purchased from Jackson Laboratories. Male, 6- to 10-week-old COX-2^–/– mice were maintained on a 129B6 F1 genetic background (B6;129P2-Ptgs2^tm1Unc) (58). EP2^–/– mice on a C57BL/6J background were purchased from The Jackson Laboratory. All experiments used littermate and COX-2^+/+ or EP2^+/+ (WT) as controls.

Reagents. The selective COX-2 inhibitors NS-398 and CAY10404, synthetic prostanoids PGE₂, PGD₂, PGI_2, and PGF_2α, stable TXA₂ analog carbocyclic TXA2 (cTXA₂), EP2 receptor agonist (butaprost), EP2 receptor antagonist (PF-04418948) and EP4 receptor antagonist (MF498) were purchased from Cayman Chemical. IL-7, IL-33, and SCF (stem cell factor) were purchased from PepPro Tech.

Airway inflammation models. As previously described (27), mice were immunized with OVA and Alum (adjuvant) by i.p. injection on Days 0 and 1; 14–18 days later, mice were exposed to 1% OVA in sterile saline (or to saline vehicle only) via inhalation for 30 minutes per day for 5 consecutive days. In some experiments, PGD₂, PGE₂, PGI₂, or PGF_2α were delivered 2 days before OVA inhalation exposure via s.c. implanted osmotic mini pumps (model 1007D; Alzet). Alternatively, mice were dosed with 20 μg of Alternaria extract (Stallergenes Greer; XPM1D3A2.5) in 50 μL PBS by i.n. instillation for 4 consecutive days, similar to previous experiments (34). For both models, mice were euthanized 24 hours after the last OVA or Alternaria extract exposure to collect bronchoalveolar lavage fluid (BALF), blood, lungs and lymph nodes. IHC staining of lung tissues, analysis of cytokines, and determination of eicosanoid levels were performed as previously described (29).

Flow cytometry. Single-cell suspensions were incubated with anti-CD16/32 Fc receptor–blocking antibody (eBioscience) for 15 minutes at 4°C, followed by surface staining with fluorochrome-conjugated antibodies for 15–20 minutes at 4°C. For identification of ILCs, a lineage (Lin) cocktail was used comprising antibodies against CD3 (145-2C11), CD5 (53-7.3), CD19 (1D3), TCRβ (H57-597), TCRγδ (eBioGL3), FcεRI (MAR-1), CD8α (53-6.7), CD11b (M1/70), NK1.1 (PK136), CD4 (RM45), and CD49b (DX5). Cells were additionally stained with antibodies against CD45 (30-F11), CD45.1 (A20), CD45.2 (clone 104), and Thy1.2 (CD90.2; 30-H12). For intracellular staining of transcription factors and cytokines, including RORγt (Q31-378), GATA3 (TWAJ), and EOMES (Dan11mag), cells were fixed and permeabilized prior to staining. ILC2 cells were identified using the following gating strategy: lymphocytes were first gated based on forward and side scatter, followed by exclusion of doublets and dead cells. Lineage⁺ (Lin⁺; CD3ε⁺, CD5⁺, CD8α⁺, CD19⁺, TCRγδ⁺, TCRβ⁺, or FcεRI⁺) cells were excluded. Viable CD45⁺Lin^– cells were subsequently gated to identify ILC2 cells as NK1.1^–RORγt^–CD11b^–GATA3⁺ cells. For detailed phenotypic and functional analysis, cells were stained with a FITC-conjugated lineage cocktail (BioLegend, 133302), BUV395 anti-CD45 (BD Biosciences, 564279), BV711 anti-CD90.2 (BioLegend, 105349), BV421 anti-CD127 (IL-7Rα; BioLegend, 135024), and PerCP/Cy5.5 anti-KLRG1 (BioLegend, 138418). Surface staining was performed at 4°C for 20 minutes, and cell viability was assessed using Live/Dead APC-Cy7 (BD Biosciences). Intracellular staining was performed using the Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer’s instructions. Antibodies included APC anti–IL-5 (BD Biosciences, 554396), anti–IL-13 (eBio13A; eFluor 660), and anti-GATA3 (BioLegend, 653806) (32). PG receptor expression was assessed using antibodies against EP2, EP3, IP, and TP receptors (all from Cayman Chemical). Appropriate isotype controls (rat IgG) were included. Data were acquired on FACSAria, LSR II, or LSRFortessa flow cytometers (BD Biosciences) and analyzed using FlowJo software (Tree Star).

ILC2 cell differentiation in vitro. ILC2 cell differentiation was performed as previously described (59). A pool of cells enriched for CLP cells was obtained from bone marrow using an EasySep Mouse Hematopoietic Progenitor Cell Isolation Kit (STEMCELL, 19856). A highly purified population of CLP cells (CD127⁺CD135⁺Sca-1⁺) was purified from bone marrow by FACS sorting. CLP cell differentiation to ILC2 cells was performed by treating with SCF, IL-7, and IL-33. Briefly, highly purified CLP cells were cultivated in DMEM supplemented with 10% FBS, 80 μM 2-mercaptoethanol, 10 mM HEPES, 1 mM sodium pyruvate plus nonessential amino acids (Gibco, Thermo Fisher Scientifics), 8 mg/mL glutamine, 100 U/mL penicillin, 0.4 mg/mL gentamicin and 100 μg/mL streptomycin. CLP cells were plated into 96-well microtiter plates previously coated with mitomycin C-treated (Sigma-Aldrich) mitotically inactive OP9 or OP9-DL1 feeder cells. CLP cells were differentiated to ILC2 cells in vitro for 6 days in media containing IL-7 (30 ng/mL), SCF (30 ng/mL), and IL-33 (20 ng/mL). To determine cytokine production, the cells were analyzed after 5–7 days of in vitro culture and were stimulated for 4 hours with a combination of phorbol 12-myristate 13-acetate (PMA, 3.7 μg/mL, Sigma-Aldrich) and ionomycin (50 ng/mL, Sigma-Aldrich) in the presence of Golgi Stop (BD Biosciences). In some experiments, PGE₂, PGD₂, PGI₂, PGF_2α, or cTXA₂, (0–1,000 nM), COX-2 inhibitor NS-398 (10 μM), EP2 agonist (butaprost, 300 or 1,000 nM), EP2 antagonist (PF-04418948, 1 μM) or EP4 antagonist (MF498, 1 μM) were added to the culture media.

Immunofluorescence staining of lung tissue. Mice were euthanized by i.p. injection of pentobarbital (115 mg/kg). BALF was collected using 1 mL of HBSS. The lungs were then intratracheally instilled with 50% Sakura Tissue-Tek O.C.T. compound (International Medical Equipment) at approximately 25 cm H₂O and subsequently frozen on dry ice for cutting in the dorsal plane at the bifurcation of the first bronchi in 5 μm sections. Lung sections were fixed in methanol with 0.3% H₂O₂ for 10 minutes at 4°C to reduce nonspecific staining and permeabilized with Triton X-100 (0.8%) (Sigma-Aldrich) for 10 minutes at room temperature. Slides were rinsed with PBS before and after the fixation and permeabilization. After lung sections were blocked with a PBS solution containing 5% BSA, they were simultaneously immunostained with FITC labeled anti-EpCAM, PE-labeled anti–IL-33, DAPI or FITC labeled anti-TSLP, PE-labeled anti-EpCAM, and/or DAPI for 60 minutes at room temperature. Lung sections were imaged using an Axioplan 2 fluorescence microscope (Carl Zeiss) with a digital camera (AxioCam MRC or MRM, Zeiss) or a confocal microscope (LSM 510, Carl Zeiss). IL-33⁺Epcam⁺DAPI⁺ cells or TSLP⁺Epcam⁺DAPI⁺ cells were counted in 2 randomly selected regions (objective, ×40) of each section from multiple lung slides (n = 5–6) and quantified using confocal microscopy.

qPCR. Homogenized whole lung tissues or single cell suspensions were resuspended in Trizol (Invitrogen, Thermo Fisher Scientifics) and stored at –80°C. RNA extraction was performed according to the protocol provided by the manufacturer. RNA concentrations were determined using a Nanodrop 8000 device (Thermo Fisher Scientifics). Reverse transcription of up to 2 μg total RNA was performed using the High-Capacity cDNA Reverse Transcription kit according to the manufacturer’s protocol (Applied Biosystems, Thermo Fisher Scientifics). TaqMan gene expression assays (Applied Biosystems) were performed on an Applied Biosystems ViiA7 Real-Time PCR Instrument to determine the expression of those murine genes. Primers used throughout the study are shown in Supplemental Table 1. According to the cycling threshold, results were normalized as an n-fold difference to the housekeeping gene, Gapdh. Calculation of mRNA levels was performed with SDS2.1 software (Applied Biosystems).

scRNA-seq library preparation and next-generation sequencing. Lung specimens were minced mechanically into small pieces and then digested for 1 hour at 37°C with 0.2% pronase and 0.08% collagenase-II/dispase in DMEM with 0.01% DNase I. Dissociated cells in suspension were passed through a 70 μm strainer and centrifuged at 500g for 5 minutes at 4°C. Cell pellets were resuspended in Red Blood Cell Lysis Buffer for 1–3 minutes on ice and washed with PBS. The resulting cell preparation was suspended in cold DMEM with 2% fetal bovine serum. scRNA-seq libraries were generated using the 10x Chromium platform 3′ v2 library preparation kit (10x Genomics) following the manufacturer’s recommendations. Next-generation sequencing was performed on an Illumina NovaSeq 6000.

Single-cell data analysis. Reads with read quality less than 30 were filtered out, and CellRanger Count v6.1.1 (10x Genomics) was used to align reads onto the GRCm38 reference genome. The Seurat package in R was used to perform dimensionality reduction with UMAP clustering. Genes were considered differentially expressed if the adjusted P < 0.05. To identify genes that were differentially expressed between OVA COX-2^–/– and OVA COX-2^+/+ lungs, we used the negative binomial model to test for differences in expression with a log fold change (FC) cutoff of 0.25 and P < 0.05. Enrichment for Gene Ontology biological process and pathways were performed using gene set enrichment analysis (GSEA). For GSEA significantly enriched gene sets were those with an adjusted P value of < 0.01 and the normalized enrichment score (NES).

Airway epithelial cell purification. Airway epithelial cells were purified as previously described (60). Briefly, lungs were treated with collagenase II and DNase I digestion at 37°C for 1 hour. Single cell suspensions were incubated with antibodies against CD45, CD31, and CD16/32 to remove nonepithelial cells. Airway epithelial cells were then sorted using anti-EpCAM antibodies.

Human ILC2 cell analysis. Human peripheral venous blood samples were collected after informed consent at the NIEHS/NIH Clinical Research Unit. Healthy subjects were excluded if they had received cytotoxic drugs in the previous 3 months or had taken steroid or nonsteroidal anti-inflammatory medications within 72 hours of sample collection. Human peripheral blood mononuclear cells (PBMCs) were isolated from 100 mL blood. In total, 1-2 × 10⁶ cells/mL of PBMCs were cultured in 24-well culture plate with IL-33 (10 ng/mL) and IL-2 (20 U/mL), with or without 100 nM of PGE₂, 100 nM of PF-04418948 (EP2 antagonist), 100 nM of MF498 (EP4 antagonist), 100 nM of butaprost (selective EP2 receptor agonist), or 100 nM of CAY10404 (selective COX-2 inhibitor). After 6 days in culture, ILC2 cells were collected and restimulated with 50 ng/mL PMA (Sigma-Aldrich), 1 μg/mL ionomycin (Sigma-Aldrich), and 0.07% GolgiStop (BD Biosciences) for 4 hours at 37°C. ILC2 cells were then stained with antibodies to lineage markers (CD3, CD14, CD16, CD19, CD56, and FceRIa) and antibodies to CD127 and CRTH2, fixed and permeabilized using the methods described in the above section. ILC2 cells were then intracellularly stained with anti-GATA3, anti–IL-5, and anti–IL-13. Human ILC2 cells were gated on the Lin^–CD127⁺CRTH2⁺ population, before being further analyzed as GATA3⁺, IL-5⁺, and IL-13⁺ ILC2 cells (61).

Statistics. Data are presented as mean ± SEM. Statistical comparisons among treatment groups were performed by randomized-design 1-way ANOVA followed by the Newman-Keuls post hoc test for more than 2 groups, multiple 2-tailed t tests using Holm-Šídák correction for multiple comparisons, or by unpaired 2-tailed Student’s t test for 2 groups using Prism software (GraphPad), as appropriate. Statistical significance was defined as P < 0.05.

Study approval. All mice were maintained under specific pathogen–free conditions and handled according to protocols approved by the NIEHS Animal Care and Use Committee and the Responsible Federal Health Authorities of the State of North Carolina. All animal experiments were performed according to Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011). Human studies were overseen by the NIH IRB and the approved protocol 000152 ENEX.

Data availability. RNA-seq data have been deposited at the National Library of Medicine, GEO dataset accession number GSE318742. Values for all data points in graphs are reported in the Supporting Data Values file.