Prostaglandin E2 constrains systemic inflammation through an innate lymphoid cell–IL-22 axis

See allHide authors and affiliations

Science  18 Mar 2016:
Vol. 351, Issue 6279, pp. 1333-1338
DOI: 10.1126/science.aad9903

A prostaglandin barrier to inflammation

Blood-borne bacterial infections and severe trauma can send the immune system into overdrive, causing it to pump out inflammatory mediators, sometimes at lethal doses. Duffin et al. now report on a role for prostaglandins in keeping systemic inflammation in check. Systemic inflammation correlates with decreased production of the prostaglandin E2 (PGE2). Blocking PGE2 signaling in mice led to severe inflammation associated with the translocation of gut bacteria. PGE2 acts on innate lymphoid cells, which produce interleukin-22, a secreted protein that helps promote intestinal integrity.

Science, this issue p. 1333


Systemic inflammation, which results from the massive release of proinflammatory molecules into the circulatory system, is a major risk factor for severe illness, but the precise mechanisms underlying its control are not fully understood. We observed that prostaglandin E2 (PGE2), through its receptor EP4, is down-regulated in human systemic inflammatory disease. Mice with reduced PGE2 synthesis develop systemic inflammation, associated with translocation of gut bacteria, which can be prevented by treatment with EP4 agonists. Mechanistically, we demonstrate that PGE2-EP4 signaling acts directly on type 3 innate lymphoid cells (ILCs), promoting their homeostasis and driving them to produce interleukin-22 (IL-22). Disruption of the ILC–IL-22 axis impairs PGE2-mediated inhibition of systemic inflammation. Hence, the ILC–IL-22 axis is essential in protecting against gut barrier dysfunction, enabling PGE2-EP4 signaling to impede systemic inflammation.

Systemic inflammation commonly develops from locally invasive infection, is characterized by dysregulation of the innate immune system and overproduction of proinflammatory cytokines, and can result in severe critical illness (e.g., bacteremia, sepsis, and septic shock) (1, 2). Despite much research on systemic inflammation, our understanding of the precise mechanisms for its control remains incomplete and represents an unmet clinical need (13). Prostaglandins (PGs) are bioactive lipid mediators generated from arachidonic acid via the enzymatic activity of cyclooxygenases (COXs) (4). PGs participate in the pathogenesis of inflammatory disease (4, 5), and many inflammatory conditions are treated using nonsteroidal anti-inflammatory drugs (NSAIDs) that inhibit PG synthesis by blocking COXs (6). NSAID therapy is also thought to confer similar beneficial effects in treating severe inflammation, but large randomized and controlled clinical trials have shown that NSAIDs failed to reduce mortality in severe systemic inflammation (7, 8). More importantly, the use of NSAIDs during evolving bacterial infection is associated with more severe critical illness (913). Therefore, it is imperative to define the paradoxical regulatory role of PGs in systemic inflammation (14).

Prostaglandin E2 (PGE2) is one of the most abundantly produced PGs and modulates immune and inflammatory responses through its receptors (EP1 to EP4) (4). We performed a genome-wide gene expression analysis of whole-blood samples from human neonates with sepsis (15) and found that expression levels of PTGES2 (encoding membrane-associated PGE synthase-2) and PTGER4 (encoding EP4) were significantly diminished in the sepsis group compared with noninfected controls (Fig. 1A). The reduced expression of PTGES2 and PTGER4 was associated with increased neutrophil blood count as a marker of inflammation (Fig. 1B). Down-regulation of PTGER4 and PTGS2 (encoding COX2) was similarly observed in patients suffering from systemic inflammatory response syndrome, sepsis, septic shock, or severe blunt trauma. In contrast, expression of HPGD (encoding 15-PGDH, which mediates PGE2 degradation) in these patients was up-regulated compared with noninfected controls (fig. S1). Consistent with this finding, blood monocytes from patients with sepsis and septic shock produced less PGE2 (16). Thus, the PGE2-EP4 pathway is down-regulated in human severe systemic inflammatory disease.

Fig. 1 PGE2-EP4 signaling controls LPS-induced systemic inflammation.

(A) Gene expression of PTGES2 and PTGER4 in whole-blood samples of neonates suffering from sepsis with confirmed bacterial infection (red, n = 27) and matched noninfected controls (blue, n = 35). Line graphs display gene expression (log2 scale) as probability density plots for both group samples. Nonparametric Wilcoxon-Rank-Sum tests (Pdiff) were used to test for differential expression of a gene between infected and control neonates. Fligner-Killeen tests (Pvar) were used to evaluate whether sepsis and control groups have substantively different intersubject variation in gene expression levels. (B) Supervised heat map of clinical neutrophil counts and expression levels for PTGES2 and PTGER4 genes for noninfected healthy controls (n = 12) and bacterially infected neonates (n = 27). A colored scale bar is shown for neutrophil count or gene expression z-score transformed values, respectively. ***P < 0.001 by nonparametric Spearman correlation test performed to analyze the negative association between PTGES2 (correlation coefficient rs = –0.6111) or PTGER4 (rs = –0.6323) gene expression and blood neutrophil counts. (C to F) Serum TNF-α and IL-6 levels (C), spleen size and weight (D), neutrophil counts in peritoneal cavity lavage (E), and liver histology (F) of WT C57BL/6 mice treated with indomethacin (Indo) or vehicle control (Veh) for 5 days, followed by LPS challenge injection for another 2 hours [(C), n = 6 mice per group] or 24 hours [(D) to (F), n = 8 per group]. (G to I) Spleen weight (G), neutrophils (H), and serum TNF-α and IL-6 levels (I) of WT C57BL/6 mice treated with indomethacin and agonists for EP2 or EP4, followed by LPS challenge for 24 hours [(G) and (H)] or 2 hours (I). Data shown as means ± SEM (error bars) are pooled from two independent experiments. Scale bar in (F), 50 μm. *P < 0.05, **P < 0.01 by one-way analysis of variance (ANOVA). NS, not significant; ND, not detected; PBS, phosphate-buffered saline.

To understand the mechanism(s) whereby PGE2 regulates systemic inflammation, we challenged wild-type (WT) C57BL/6 mice with lipopolysaccharide (LPS) to induce systemic inflammation after pretreatment with indomethacin, which effectively suppresses PGE2 production (17, 18). Mice pretreated with indomethacin developed an enhanced cytokine storm [e.g., tumor necrosis factor–α (TNF-α) and interleukin-6 (IL-6)] (Fig. 1C), as well as other inflammatory signs such as splenomegaly (Fig. 1D), peritonitis characterized by accumulation of CD11b+Ly-6G+ neutrophils (Fig. 1E), and low-grade hepatic inflammation, as indicated by sinusoidal lymphocytosis and necrosis or inflammatory foci in the parenchyma (Fig. 1F). Furthermore, coadministration of an EP4 agonist almost completely diminished indomethacin-augmented systemic inflammation (Fig. 1, G to I). Thus, PGE2-EP4 signaling constrains LPS-induced systemic inflammation.

In addition to the inflammatory markers described above, we also detected dissemination of gut bacteria into normally sterile tissues (e.g., liver) of indomethacin-treated mice but not controls (Fig. 2A). The coadministration of an EP4 agonist blocked the dissemination of bacteria to sterile tissues (Fig. 2B). Although pathogenic bacteria spread via the bloodstream usually cause severe systemic inflammation, gut leakage of commensal bacteria, especially in patients with damaged gut epithelium or endothelium, may also trigger systemic inflammation. Therefore, to test whether these disseminated bacteria contribute to indomethacin-augmented systemic inflammation, we treated mice with indomethacin and antibiotics, which are known to effectively deplete gut bacteria (19). Antibiotic therapy reduced indomethacin-facilitated systemic inflammation (Fig. 2, C to F). Indomethacin-dependent systemic inflammation was also observed in Rag1−/− mice (fig. S2) and was again diminished by EP4 agonism (Fig. 2, G to I). Hence, PGE2-EP4 signaling prevents systemic inflammation independently of adaptive immune cells.

Fig. 2 PGE2 control of systemic inflammation involves gut bacterial dissemination and acts independently of adaptive immune cells.

(A) Colony forming units (CFU) present in liver homogenates from WT C57BL/6 mice (n = 6) treated with indomethacin for 5 days, followed by LPS challenge for another 24 hours. (B) CFU present in liver homogenates from WT C57BL/6 mice treated with indomethacin plus an EP4 agonist (n = 6) or vehicle (n = 8) for 5 days, followed by LPS challenge for another 24 hours. (C to F) Spleen size and weight (C), neutrophils (D), liver histology (E), and serum TNF-α and IL-6 levels (F) of WT C57BL/6 mice (n = 6 per group) treated with indomethacin and antibiotics (ABX) for 5 days, followed by LPS challenge for another 24 hours [(C) to (E)] or 2 hours (F). (G to I) TNF-α and IL-6 levels in serum (G) and peritoneal cavity lavage (H) and CFU present in liver homogenates (I) from Rag1−/− mice treated with indomethacin plus an EP4 agonist or vehicle control (n = 8 per group) for 4 days. Data shown as means ± SEM (error bars) are pooled from two independent experiments. Scale bar in (E), 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001 by two-way ANOVA [(C), (D), and (F)] or Mann-Whitney test [(G) to (I)]. ND, not detected.

Given that IL-22 has recently been shown to inhibit inflammatory responses, particularly in the intestine (1922), we hypothesized that IL-22 may mediate PGE2 control of systemic inflammation. To test this hypothesis, we first examined the effect of PGE2 on IL-22 production. Mice treated with indomethacin exhibited a decrease in LPS-induced IL-22 production (Fig. 3A), which was mimicked by an EP4 antagonist (Fig. 3B). Thus, PGE2-EP4 signaling promotes IL-22 production in vivo. Whereas LPS-induced systemic inflammation in WT mice was exacerbated or prevented by indomethacin or an EP4 agonist, respectively (Fig. 1, G to I), both had negligible effects on LPS-induced inflammation in IL-22–deficient mice (23) (fig. S3). IL-22 is therefore required for coupling PGE2-EP4 signaling–dependent control of systemic inflammation. Furthermore, coadministration of recombinant IL-22 (rIL-22) prompted a decrease in bacterial dissemination induced by indomethacin, which positively correlated with reduced systemic inflammation (Fig. 3, C to E), implying that IL-22 mediates PGE2 suppression of gut bacterial dissemination and subsequent systemic inflammation.

Fig. 3 IL-22 mediates PGE2 protection against systemic inflammation and intestinal barrier injury.

(A) Serum IL-22 levels of WT C57BL6 mice treated with indomethacin ( n = 6) or vehicle control (n = 6) for 5 days, followed by challenge with LPS or PBS for 2 hours. (B) Serum IL-22 levels in WT C57BL/6 mice treated with an EP4 antagonist or vehicle (n = 4 per group) in drinking water for 5 days, followed by LPS challenge for another 2 hours. (C to G) Rag1−/− mice were treated with vehicle (n = 9) or indomethacin plus rIL-22 (n = 10) or vehicle-control PBS (n = 10) for 4 days. (C) CFUs present in liver homogenates, (D) neutrophil counts, (E) correlation of liver CFU with systemic inflammation profile and expression profile of genes related to intestinal barrier function, and [(F) and (G)] gene expression of IL-22 and its receptors (F) and antimicrobial peptides (G) in the terminal ileum. Expression was measured by real-time PCR, normalized to the Gapdh gene and presented as relative expression (in log2 fold change) to the vehicle-control group. (H to J) Change in body weight (H), disease activity index (I), and colon length (J) of WT C57BL/6 mice (n = 10 per group) treated with 2% DSS and indomethacin or vehicle control in drinking water plus rIL-22 or vehicle-control PBS for 6 days. Data shown as means ± SEM (error bars) are pooled from two [(A) and (H) to (J)] or three [(C) to (G)] independent experiments or represent one experiment (B). *P < 0.05, **P < 0.01, ***P < 0.001 by Mann-Whitney test [(A), (B), and (D)], one-way ANOVA [(C) and (F) to (J)], or two-tailed Pearson correlation test (E).

As both PGE2 and IL-22 can potentially protect the gut epithelial barrier (1722), we reasoned that inhibition of bacterial dissemination by PGE2 may proceed through augmentation of IL-22 action on gut epithelial cells. We therefore examined gene expression related to barrier function in intestinal tissues. Indomethacin suppressed expression of IL-22 as well as its receptors [i.e., Il22ra1 (encoding IL-22Rα1) and Il10rb (encoding IL-10Rβ)], and their suppression was prevented by rIL-22 (Fig. 3F), which suggests that PGE2 may strengthen IL-22–IL-22R signaling in intestinal epithelial cells. Indeed, IL-22R–target genes [such as Reg3b, Reg3g, and Fut2 (24)], mucins, and tight junctions were similarly down-regulated by indomethacin but rescued by rIL-22 (Fig. 3G and fig. S4). Expression of genes related to IL-22R signaling and barrier function inversely correlated with gut bacterial dissemination (Fig. 3E). Given the protective role of PGE2 in acute colonic mucosal injury (17), we proposed that this protection is mediated by IL-22. Indomethacin exacerbated colitis induced by dextran sulfate sodium (DSS), but rIL-22 had the reverse effect, producing significantly lower values of the colitis disease activity index, as measured by weight loss, diarrhea, and rectal bleeding (Fig. 3, H to J). Our results indicate an important role of PGE2 in regulating the intestinal epithelial barrier function through innate IL-22 signaling.

Flow cytometric analysis showed that IL-22–producing cells were CD45LowLineage(Lin)CD90.2+RORγt+CCR6+ type 3 innate lymphoid cells (ILC3s) (22) in the gut and CD45+LinCD90.2+CD4+ ILC3s in the spleen (fig. S5). We wondered whether these cells are involved in EP4-dependent control of systemic inflammation. Although Rag1−/− mice cotreated with indomethacin and an EP4 agonist (i.e., lacking PG signaling except through EP4) did not develop systemic inflammation (Fig. 2, G to I), depletion of CD90.2+ ILCs restored indomethacin-induced systemic inflammation in these mice (Fig. 4A). To address whether PGE2-EP4 signaling specifically in ILCs controls systemic inflammation, we crossed CD90.2-Cre mice (25) with EP4-floxed mice (26) to generate tamoxifen-induced selective down-regulation of EP4 in CD90.2+ cells (including T cells and ILCs) in heterozygous CD90.2CreEP4fl/+ mice (Fig. 4B). CD90.2CreEP4fl/+ mice produced lower levels of IL-22 in response to LPS, which was associated with augmentation of systemic inflammatory response [e.g., TNF-α production (Fig. 4B)], whereas down-regulation of EP4 in T cells (27) did not affect IL-22 or TNF-α production (fig. S6). These data demonstrate that PGE2-EP4 signaling controls systemic inflammation, at least in part, through the regulation of ILC3s.

Fig. 4 PGE2-EP4 signaling potentiates ILC3 homeostasis and function that contributes to control of the systemic inflammatory response.

(A) Splenic CD90.2+ ILCs, serum TNF-α levels, and neutrophil counts in Rag1−/− mice cotreated with indomethacin and an EP4 agonist plus anti-CD90.2 (αCD90.2) or control immunoglobulin G (IgG) antibodies (n = 8 per group) for 4 days. (B) Expression of Ptger4 mRNA in splenic CD90.2+ cells and serum levels of IL-22 and TNF-α in tamoxifen-treated CD90.2CreEP4fl/+ mice (n = 7) and control C57BL/6 mice (n = 5) at 1.5 hours after LPS challenge. (C to E) Rag1−/− [(C) and (D)] or C57BL/6 (E) mice were treated with indomethacin with or without an EP4 agonist for 4 to 5 days. (C) Percentages and numbers of ILC3s in small intestines. Cells were stimulated ex vivo with IL-23 for 3 hours. Each point represents an individual mouse. (D) Ki-67 expression in gut ILC3s. Each point represents an individual mouse. (E) Gene expression profile in gut ILC3s presented as relative expression to the vehicle-control group. Data are the summary of two independent experiments with eight mice per group. (F) Expression of RORγt and IL-22 by Rag1−/− gut CD45+ lamina propria leukocytes (LPLs) stimulated with IL-23 and PGE2 for 4 hours. Numbers in brackets represent IL-22 geometric mean fluorescence intensities of the RORγt+IL-22+ populations. (G to J) IL-22 production in supernatants by LPLs or ILC3s isolated from small intestines of Rag1−/− mice and then cultured with indicated conditions overnight [(G), (H), and (J)] or for 3 days (I). (K) IL-22 production by human ILC3s cultured with IL-2, IL-23, IL-1β, and PGE2 for 4 days (n = 5 donors). (L) Plasma IL-22 levels in individuals with acute pancreatitis (AP) at the time of hospital admission (n = 48) or healthy individual donors (n = 28). The vertical line represents the mean for each group. Data shown as means ± SEM (error bars) are pooled from two or more independent experiments [(A), (C) to (E), (G), (J), and (K)] or represent one (B) or two [(F), (H), and (I)] independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001 by unpaired Student’s t test [(A), (B), (H), and (I)], one-way ANOVA [(C), (D), (F), (G), and (J)], ratio paired t test (K), or Mann Whitney test (L).

Next, we investigated the effect of PGE2 on ILC3 maintenance. Indomethacin significantly reduced ILC3 numbers in the gut and spleen in the steady state (Fig. 4C and fig. S7, A to C). This reduction was prevented by an EP4 agonist and was mimicked by an EP4 antagonist (Fig. 4C and fig. S7). Consistently, ILC3s express PGE2 receptors with particularly high expression levels of EP4 (fig. S8). Furthermore, indomethacin down-regulated Ki-67 expression, and an EP4 agonist prevented this reduction in ILC3s (Fig. 4D), suggesting that PGE2-EP4 signaling potentiates ILC3 proliferation. PGE2-EP4 signaling also prevented ILC3 apoptosis in vitro but had little effect on ILC3 apoptosis in vivo (fig. S9, A and B). Moreover, down-regulation of EP4 in CD90.2+ ILCs not only decreased RORγt+ ILC3s but also weakened IL-7 responsiveness [e.g., Bcl-2 expression] in ILC3s (fig. S9C). To further understand the effect of PGE2 on ILC3s, we measured gene expression by quantitative polymerase chain reaction (PCR) in LinCD45lowCD90.2hiKLRG1CCR6+ ILC3s sorted from the small intestines of mice treated with indomethacin and an EP4 agonist. PGE2-EP4 signaling up-regulates many ILC3 signature genes, including Rorc, Il23r, Il1r1, Il22, Il17a, Lta, Ltb, Tnfsf11, Tnfsf4, and Bcl2l1 (Fig. 4E).

We next asked whether and how PGE2 regulates ILC3 function. PGE2 promoted IL-23–driven IL-22 production by gut lamina propria leukocytes isolated from Rag1−/− mice in vitro (Fig. 4, F and G), and this was inhibited by indomethacin (Fig. 4H). PGE2 also promoted IL-22 production from splenic CD4+ ILC3s, which was mediated by EP2 and EP4 (fig. S10, A to C). More importantly, PGE2 increased IL-23–driven IL-22 production from highly purified CD45+LinCD90.2+KLRG1CCR6+ ILC3s from intestines and CD3CD11bCD11cCD4+ ILC3s from spleen or bone marrow (Fig. 4I and fig. S10D), suggesting that PGE2 acts directly on ILC3s. Additionally, PGE2 increased IL-22 production from ILC3s in response to IL-1β and IL-7 (fig. S11). PGE2 activated cyclic adenosine monophosphate signaling in ILC3s and, in turn, enhanced IL-22 production (Fig. 4J and fig. S12, A and B). Moreover, PGE2-augmented IL-22 production was dependent on the transcription factor STAT3 (signal transducer and activator of transcription 3), which is critical for IL-22 production by ILC3s (28), but not RORγt or Ahr (fig. S12, C to E). Together, these data demonstrate that PGE2-EP4 signaling promotes ILC3 homeostasis and IL-22 production by stimulating ILC3 proliferation and enhancing cytokine (e.g., IL-7, IL-23, and IL-1) responsiveness.

Finally, we sought to determine whether PGE2 promotes IL-22 production from human ILC3s. To answer this question, we sorted human LinCD161+CD127+CD117+CRTH2 ILC3s from peripheral blood of healthy donors and cultured them with IL-2, IL-23, and IL-1β to induce IL-22 production. Addition of PGE2 similarly up-regulated IL-22 production from sorted human ILC3s (Fig. 4K). Moreover, expression of the human IL22 gene in healthy individuals infused with a bacterial endotoxin (29) positively correlated with expression of PTGS2 and PTGES (encoding mPGES-1) (fig. S13), two inducible enzymes that mediate PGE2 synthesis (4). Acute pancreatitis (AP) is a sterile initiator of systemic inflammation that results in multiple organ dysfunction where gut barrier injury is central to the pathogenesis (30). Given that IL-22 was protective in an animal model of AP (31), we measured IL-22 levels in patients with AP. IL-22 concentrations in plasma were lower in patients with AP compared with healthy controls at the time of hospital admission (Fig. 4L), further confirming that the reduction of IL-22 signaling is associated with development of systemic inflammation.

Here we identify a physiological role of endogenous PGE2-EP4 signaling in activation of the ILC3–IL-22 axis, which functionally contributes to cross-talk between the innate immune system, gut epithelium, and microflora and subsequently constrains systemic inflammation (fig. S14). These findings provide valuable insight toward understanding how inactivation of COXs may be harmful in severe bacterial infection and inflammation (3234). In addition, our results advance a crucial cellular and molecular mechanism for a scenario in which maintaining or augmenting PGE2 signaling—e.g., by inhibiting the PGE2-degrading enzyme 15-PGDH—protects against intestinal barrier injury and potentiates its repair and control of systemic inflammation (35, 36).

Supplementary Materials

Materials and Methods

Figs. S1 to S14

References (3755)

References and Notes

Acknowledgments: We thank J. Allen for critically reading and editing the manuscript; P. J. Brophy and D. Mahad for CD90.2(Thy1.2)-Cre mice; J.-C. Renauld for IL-22–deficient mice; J. Pollard, P. J. Spence, and J. Thompson for Rag1−/− mice; T. Walsh and J. Rennie for human blood samples; Y. X. Fu, X. G. Guo, D. Ruckerl, T. Kendall, J. Hu, Q. Huang, and G. T. Ho for advice, technical assistance, and/or discussion; and S. Johnston and W. Ramsay for cell sorting and analysis. We also thank the Wellcome Trust Clinical Research Facility staff and the Department of Surgery, NHS Lothian. Mice deficient in IL-22 are available from the Jean-Ludwig Institute For Cancer Research under a material transfer agreement with J.-C. Renauld. This work was supported in part by the University of Edinburgh start-up funding (Ch.Y.), the Wellcome Trust Institutional Strategic Support Fund (Ch.Y., J.P.I., S.M.A), MRC UK (J.P.I., S.M.A., R.D., A.G.R.), NIH (grant DK37097 to R.M.B.), a VA Merit Award (1BX000616 to R.M.B.), Health Foundation/Academy of Medical Sciences (D.J.M.), University of Edinburgh and GlaxoSmithKline Discovery Partnerships with Academia (DPAC) collaboration (D.J.M.), a Wellcome Trust Investigator Award (106122 to R.M.M.), the Rainin Trust (Award 13H6 to R.M.M.), CREST of JST (S.N.), European Union FP7 Industry Academia Partnerships and Pathways project ClouDx-I, Chief Scientists Office (grant ETM202 to P.G.), and Biotechnology and Biological Sciences Research Council (grant BB/K091121/1 to P.G.). The data presented in this manuscript are tabulated in the main paper and the supplementary materials. We declare no financial conflicts of interest.

Stay Connected to Science


Navigate This Article