Articles https://doi.org/10.1038/s41590-018-0243-7

Autocrine–paracrine prostaglandin E2 signaling restricts TLR4 internalization and TRIF signaling

Darren J. Perkins 1*, Katharina Richard1, Anne-Marie Hansen1, Wendy Lai1, Shreeram Nallar1, Beverly Koller2 and Stefanie N. Vogel 1*

The unique cell biology of Toll-like receptor 4 (TLR4) allows it to initiate two signal-transduction cascades: a signal depen- dent on the adaptors TIRAP (Mal) and MyD88 that begins at the cell surface and regulates proinflammatory cytokines, and a signal dependent on the adaptors TRAM and TRIF that begins in the endosomes and drives the production of type I interfer- ons. Negative feedback circuits to limit TLR4 signals from both locations are necessary to balance the inflammatory response.

We describe a negative feedback loop driven by autocrine–paracrine prostaglandin E2 (PGE2) and the PGE2 receptor EP4 that restricted TRIF-dependent signals and the induction of interferon-β ​through the regulation of TLR4 trafficking. Inhibition of PGE2 production or antagonism of EP4 increased the rate at which TLR4 translocated to endosomes and amplified TRIF- dependent activation of the transcription factor IRF3 and caspase-8. This PGE2-driven mechanism restricted TLR4–TRIF sig- naling in vitro after infection of macrophages by the Gram-negative pathogens Escherichia coli or Citrobacter rodentium and protected mice against mortality induced by Salmonella enteritidis serovar Typhimurium. Thus, PGE2 restricted TLR4–TRIF sig- naling specifically in response to lipopolysaccharide.

he host innate immune inflammatory response to lipopolysac- Detailed understanding exists of how the kinetics of MyD88- charide (LPS) from Gram-negative bacteria is mediated by the dependent signaling are regulated. MyD88-driven NF-κ​B rapidly recognition of LPS by Toll-like receptor 4 (TLR4) in the con- drives expression of the enzyme A20 that accumulates and causes T 1–3 text of TLR4-bound its co-receptor MD2 . TLR4 is unique among disassembly of the NF-κ​B and MAPK signalosomes through de- the TLRs in that following the recognition of LPS by MD2, two ubquitinase activities as well as ubiquitin-ligase activities22,23. The separate ‘modes’ of signaling can be initiated, each from a distinct negative feedback loop mediated by the induction of A20 is criti- cellular compartment that utilizes distinct signaling intermediates, cal for host homeostasis, as A20-deficient mice succumb to massive which results in discrete transcriptional programs4. Cell-surface inflammation shortly after birth and A20-deficient macrophages TLR4 engages the intracellular adaptors TIRAP (Mal) and MyD88 display prolonged MyD88 signaling with enhanced NF-κ​B and through interactions with its Toll–interleukin 1 (IL-1) receptor MAPK responses24. Notably, however, macrophages deficient in (TIR) domain to activate signaling via the transcription factor NF-κ​ A20 display normal regulation of LPS-driven activation of IRF3, B and kinases of the MAPK family and subsequent transcription which indicates that a distinct feedback loop regulates the TLR4– of genes encoding proinflammatory cytokines (e.g., Tnf and Il1b)5. TRIF pathway24.

A qualitatively distinct signal-transduction cascade is elicited fol- Endogenously produced prostaglandin E2 (PGE2) is a strong can- lowing endosomal translocation of a portion of LPS-bound TLR4– didate for the inhibition of TLR4-dependent production of interfer- MD2 through an incompletely understood action of the cell-surface ons due to its rapid synthesis and secretion in response to LPS25,26 co-receptor CD146–8. The TIR domain of endosomal TLR4 engages and its autocrine mode of action27, as well as its noted immuno- the adaptor TRAM, which recruits the adaptor TRIF and leads to modulatory ability during infection with Gram-negative bacteria28. activation of the kinase TBK1 that leads to phosphorylation of the Additionally, PGE2 has also been shown to display a reciprocally transcription factor IRF3 and, ultimately, the production of type I antagonistic pattern of expression with IFN-α​ and IFN-β​ in some 9–12 29 interferons (e.g., IFN-β​) . In addition, TRIF signaling also activates models of bacterial infection . PGE2 itself is a secreted, bioactive, a second, unique pathway that involves caspase-8 and the kinases signaling lipid that frequently acts in an autocrine–paracrine man- RIP1 and RIP3, which leads to caspase-1-independent processing ner that can influence both inflammatory processes and homeo- of pro-IL-1β13​ –15. The kinetics of activation and termination of the static processes30,31. separate signal-transduction cascades driven by MyD88 and TRIF In search of possible endogenous regulatory circuits that limit are necessarily limited to protect host tissues from excessive inflam- IFN-β ​production dependent on TLR4–TRIF during the response mation and to re-establish homeostasis. While MyD88-dependent to LPS from Gram-negative bacteria, we performed a microar- production of cytokines is crucial for the clearance of many bacte- ray analysis to identify genes expressed differentially in wild-type rial infections16, conversely, the production of type I interferons can primary mouse thioglycollate-elicited peritoneal macrophages promote infection in various models by a mechanism that involves (TEPMs) infected with Salmonella enteritidis serovar Typhimurium a poorly understood immunosuppression17–21. Thus, it is crucial to (ST) relative to the expression of those genes in their IFN-β​-defi- understand the molecular mechanisms that limit the inflammatory cient (Ifnb–/–) counterparts. Subsequent bioinformatics analysis response and establish the balance between TLR4–MyD88 signaling of those data predicted transcriptional regulators whose activities and TLR4–TRIF signaling. were inversely related to IFN-β ​expression during ST infection and

1Department of Microbiology and Immunology, University of Maryland, Baltimore, School of Medicine, Baltimore, MD, USA. 2Department of Genetics, UNC School of Medicine, Chapel Hill, NC, USA. *e-mail: [email protected]; [email protected]

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy

identified the high-affinity PGE2-specific receptor EP4 (encoded by ing of PGE2 by EP4 regulated Ifnb transcription, we measured Tnf Ptger4). We report here that in response to bacterial LPS, PGE2 was and Ifnb mRNA over a 6-hour time course of LPS stimulation with- rapidly released from mouse macrophages and participated in an out or with the EP4 antagonist. Selective blockade of EP4 enhanced autocrine–paracrine regulatory loop through EP4. This PGE2–EP4 transcription of LPS-induced Ifnb mRNA but not that of Tnf mRNA signaling axis specifically limited the TRIF-dependent arm of the (Fig. 1g). To confirm the role of EP4 in regulating LPS-induced TLR4 response to LPS and thereby reduced the production of type I IFN-β,​ we generated bone marrow–derived macrophages (BMDMs) interferons, as well as the TRIF-dependent, caspase-1-independent from wild-type and Ptger4–/– (EP4-null) mice and compared their processing of IL-1β.​ These findings contribute to the understand- production of LPS-induced IFN-β​ protein. Ptger4–/– BMDMs pro- ing of the regulation of endosomal TLR4 signals, as well as how the duced substantially more IFN-β ​than did wild-type BMDMs (Fig. production of type I interferons during bacterial infection might be 1h). These data indicated a role for endogenously produced PGE2 limited by the host. in the selective suppression of IFN-β ​production in response to LPS.

Results Inhibition of PGE2–EP4 enhances the activation of IRF3. To LPS-dependent PGE2 limits IFN-β production. We initially car- explore the mechanism underlying the selective suppression of the ried out a microarray analysis to identify transcripts whose induc- production of IFN-β​ by PGE2–EP4 during the response to LPS, we tion by ST infection in macrophages was increased or inhibited assessed activation of the TRIF-activated transcription factor IRF3 specifically by IFN-β​ expression (wild-type versus Ifnb–/–). With over a 180-minute time course. In vehicle-treated, LPS-stimulated these data, we used the Upstream Regulator Analytic tool of macrophages, activation of IRF3 (assessed as phosphorylated Ingenuity Pathway Assist software to identify transcriptional regu- IRF3) was observed between 60 min and 120 min after stimulation, lators whose activity was predicted to be elevated or inhibited by in agreement with published work35. However, in the presence IFN-β ​during the response to ST. As expected, this analysis identi- of the EP4-specific antagonist, we observed enhanced amplitude fied various known regulators of the production or signaling of type and duration of the activation of IRF3 (Fig. 2a). In the presence of I interferons as being inhibited in the absence of IFN-β​ expression the EP4 antagonist, enhanced LPS-dependent activation of IRF3 (Fig. 1a). Notably, this analysis also indicated that the PGE2-specific was also observed in the human monocytic cell line THP-1 and in receptor EP4 displayed activity inversely related to the production primary human monocytes (Supplementary Fig. 2). To determine of IFN-β ​during the early (6-hour) inflammatory response to ST the extent to which antagonism of EP4 might prolong the activa- infection, a time point at which the innate inflammatory response tion of IRF3, we extended the time course of LPS stimulation and to ST is predominantly TLR4 dependent17. observed substantial phosphorylation of IRF3 as late as 5 h after

To elucidate a possible role for autocrine or paracrine PGE2 stimulation with LPS in the presence of the EP4 antagonist (Fig. 2b). in the regulation of TLR4 signaling in response to LPS, we first To determine the specificity of EP4’s regulation of IRF3, we determined the kinetics of endogenous PGE2 production in LPS- assessed the LPS-dependent activation of MAPKs in the presence stimulated peritoneal macrophages. Activation of the proximal or absence of the EP4 antagonist. We found no modulation of acti- enzyme in the PGE2 biosynthetic cascade, the calcium-dependent vation of the MAPKs Jnk, Erk or p38 in LPS-stimulated macro- phospholipase cPLA2, was assessed by phosphorylation-specific phages by the EP4 antagonist (Fig. 2c). Thus, the inhibitory effect immunoblot analysis. Stimulation of mouse peritoneal macrophages of the PGE2–EP4 signaling axis seemed to be specific for TLR4- with LPS resulted in activation of cPLA2 within 30 min that ended mediated signaling via TRIF. by 90 min (Fig. 1b). Quantitation of extracellular PGE2 by ELISA We additionally assessed the effect of EP4 antagonism on deg- indicated that secreted PGE2 was readily detected within 3 h of LPS radation of the inhibitory cytoplasmic NF-κ​B chaperone Iκ​Bα​ stimulation and that LPS-dependent release of PGE2 was inhibited in response to LPS. The early kinetics of the degradation of Iκ​ by a small-molecule antagonist of the microsomal PGE2 synthase Bα​ were not affected by antagonism of EP4; however, we consis- mPGES, the terminal enzyme in the PGE2 biosynthetic cascade that tently observed a delay in the re-synthesis of Iκ​Bα​ in the presence 32 is essential for LPS-induced production of PGE2 (ref. ) (Fig. 1c). of the inhibitor of EP4 (Fig. 2d). This delayed re-synthesis of Iκ​ Analysis of the concentration of secreted cytokines in supernatants Bα​ was consistent with the results of a published report demon- of LPS-stimulated macrophages treated with vehicle (DMSO) or the strating regulation of the re-synthesis of Iκ​Bα​ by the TRIF path- 36 mPGES antagonist revealed no effect of the inhibition of PGE2 on way . To confirm the results obtained with the EP4 antagonist, we production of the cytokine TNF; however, inhibition of the produc- compared the time course with which IRF3 was activated in wild- –/– tion of PGE2 resulted in an increase in the secretion of IFN-β​ (Fig. type BMDMs versus that in Ptger4 BMDMs. Consistent with the 1d). This result was consistent with our preliminary hypothesis that results of our inhibitor studies, the activation of IRF3 was extended –/– autocrine–paracrine PGE2 might negatively regulate the TLR4- in the LPS-stimulated Ptger4 BMDMs (Fig. 2e). The specific- activated TRIF signaling pathway. Extracellular PGE2 can be sensed ity of EP4 in regulating the activation of IRF3 was confirmed by by any of four specific PGE2 receptors, EP1–EP4, each of which has comparison of the effects of the EP4 antagonist with those of the distinct expression patterns and functions33. We initially focused on EP2- or EP3-specific antagonists. Only inhibition of EP4 enhanced a role for EP4, as it was identified in our bioinformatics analysis and and prolonged the LPS-dependent activation of IRF3 (Fig. 2f). To had previously been linked to the regulation of innate inflamma- control for potential nonspecific effects of our EP4 antagonist, we 34 tory responses . Therefore, macrophages were stimulated with LPS subsequently compared the effects of global inhibition of PGE2 in the presence or absence of a specific, high-affinity antagonist of by the mPGES antagonist on the activation of IRF3 (Fig. 2g) and EP4. Treatment of cells with this EP4 antagonist produced results its effects on the re-synthesis of Iκ​Bα ​ (Fig. 2h). In both instances, identical to those obtained with the mPGES antagonist in terms inhibiting the production of PGE2 globally recapitulated the effects of enhanced production of IFN-β​ with no effect on TNF secretion of the EP4-specific antagonist. Finally, we sought to ascertain if it (Fig. 1e). Analysis of additional, non-secreted proteins dependent were possible to amplify the effects of the endogenous autocrine on MyD88 (i.e., A20) or IFN-β ​(i.e., PKR) confirmed this pattern or paracrine PGE2 regulatory loop by treating macrophage cultures of regulation (Supplementary Fig. 1). In contrast to the increased with exogenous PGE2. The addition of 1 μM​ PGE2 to macrophage production of IFN-β​ seen in LPS-stimulated macrophages treated cultures, concurrent with LPS stimulation, substantially suppressed with the EP4 antagonist, no enhancement of LPS-induced IFN-β​ the activation of IRF3 (Fig. 2i). These experiments indicated that was observed in the presence of specific antagonists of other PGE2 endogenously produced PGE2 regulated TLR4-dependent tran- receptors (EP2 or EP3) (Fig. 1f). To determine if inhibiting the sens- scription of Ifnb by limiting the activation of IRF3.

Nature Immunology | www.nature.com/natureimmunology NATuRE IMMunoLoGy Articles ab c Regulator z-score LPS time (min) 0306090120 180 125 * IRF7 –7.023 100 Vehicle IRF3 –6.744 p-cPLA2 75 LPS

(pg/ml) IFNAR –6.6 2 50 LPS + mPGES Total antagonist GE 25 STAT1 –6.036 cPLA2 P 0 PTGER4 5.162 p-cPLA2/ 0.56 cPLA2 0.005 0.3170.001 0.003 0.005 def * * 600 14 1,200 700 900 Vehicle 500 12 Vehicle 600 Vehicle 750 400 1,000 10 EP4 EP2 antagonist 500 mPGES 600 300 800 8 antagonist EP3 antagonist 400 antagonist 450 600 300 6 200 400 4 300 IFN- β (pg/nl) 100

200 (ng/nl) TNF TNF (pg/ml) TNF

IFN- β (pg/nl) 150

β (pg/ml) 200 IFN- 100 2 0 0 0 0 0 g 20 h 600 200 * WT Vehicle Vehicle 500 –/– 15 EP4 antagonist Ptger4

150 EP4 antagonist 400 (fold) 100 10 300

* 200 Ifnb Tn f (fold) 50 5 IFN- β (pg/nl) 100 0 0 0 024860246 8 Time (h) Time (h)

Fig. 1 | Rapid LPS-dependent PGE2 limits TLR4-mediated production of IFN-β. a, Partial list of transcriptional regulators with predicted differential activity in ST-infected Ifnb−/− primary mouse TEPMs relative to that in ST-infected wild-type primary mouse TEPMs. b, Immunoblot analysis of phosphorylated (p-) or total cPLA2 (left margin) in whole-cell lysates of TEPMs stimulated for 0–180 min (above lanes) with LPS (100 ng/ml throughout); below lanes, densitometry (comparison, left margin). c, ELISA of PGE2 in supernatants of TEPMs pre-treated with vehicle alone (DMSO throughout) or the mPGES antagonist (key) before stimulation for 3 h with LPS. *P =​ 0.0003. d, ELISA of TNF (left) or IFN-β​ (right) in supernatants of TEPMs pre-treated as in c (key) and then stimulated for 6 h with LPS. *P =​ 0.0043. e, ELISA of TNF (left) or IFN-β​ (right) in supernatants of TEPMs pre-treated for 15 min with vehicle or the EP4 antagonist BGC 20-1531 (50 μ​M) (key) before stimulation as in d. *P =​ 0.02. f, ELISA of IFN-β​ in supernatants of TEPMs pre-treated for 30 min with vehicle or with the EP2 antagonist PF-04418948 (50 μ​M) or EP3 antagonist L-798,106 (50-μ​M) in DMSO (key) before stimulation as in d. g, qRT- PCR analysis of Tnf (left) or Ifnb (right) among total RNA from TEPMs pre-treated as in (e) (key) and then stimulated for 0–8 h (horizontal axis) with LPS; results were normalized to those of the control gene Hprt. *P =​ 0.0024. h, ELISA of IFN-β​ in supernatants of wild-type or Ptger4−/− BMDMs (key) stimulated for 6 h with LPS. Data are from three separate experiments (a; n =​ 3), are from one experiment representative of at least four experiments (b; n > 4),​ are from n = 3​ independent experiments (c–g; mean +​ s.e.m. (c–f) or mean ±​ s.e.m. (g)) or are from two experiments (h; mean +​ s.e.m.). P values, two-tailed Student’s unpaired t-test.

PGE2–EP4 regulates the activation of TRIF. The consider- shown that there is an absolute requirement for activation of the able effect of EP4 antagonism on the TLR4-dependent activation tyrosine kinase Syk in a CD14-dependent, but TLR4-independent, of IRF3, an event known to be strictly dependent on the adaptor fashion7. Therefore, LPS-induced activation of Syk was assessed in 10 TRIF , led us to speculate that PGE2–EP4 might be targeting this the presence or absence of the EP4 antagonist. Antagonism of EP4 adaptor pathway specifically. LPS-induced activation of IRF3 was during stimulation with LPS resulted in more-rapid and increased assessed in wild-type and TRIF-deficient (Trif–/–) peritoneal macro- activation of Syk (Fig. 3c), consistent with the results of our TLR4- phages in the presence or absence of antagonists of mPGES or EP4. internalization experiments (Fig. 3b). To further assess the hypoth-

The activation of IRF3 following stimulation with LPS was strictly esis that PGE2–EP4 feedback restricts TRIF signaling specifically, TRIF dependent, whether or not the mPGES antagonist was pres- we stimulated peritoneal macrophages with Pam3Cys, a ligand for ent (Fig. 3a). Notably, the effect of the mPGES and EP4 antagonists TLR2, which is a TLR that engages MyD88 but not TRIF. Addition on LPS-induced re-synthesis of Iκ​Bα​ was completely reversed in of the mPGES antagonist did not affect TLR2-dependent signal- Trif–/– macrophages (Fig. 3a), again in support of the notion that the ing or result in TLR2-dependent activation of IRF3 (Fig. 3d). We

PGE2–EP4 axis specifically targeted the TRIF signaling pathway in did not, however, observe an effect of the EP4 antagonist on the its regulation of TLR4 signaling. induction of IFN-β​ in response to stimulation of TLR3 by its ligand The TRIF pathway is initiated by the CD14-dependent translo- poly(I:C), probably due to the fact that poly(I:C) was a much weaker cation of the TLR4–LPS–MD2 complex to an endosome, where it inducer of the release of PGE2 than was LPS (Supplementary Fig. 4). recruits the adaptors TRAM and TRIF to the TLR4 TIR domain These experiments demonstrated that PGE2 regulated activation of that is exposed to the cytosol6. We monitored the endosomal trans- the TLR4–TRIF pathway through trafficking and that TRIF expres- location of TLR4 by cell-surface staining of TLR4 and flow cytom- sion was required for the regulation of TLR4 signaling by PGE2. etry. Unstimulated cells exhibited no internalization of TLR4 over the time monitored. However, either antagonizing EP4 with the PGE2–EP4 restricts TRIF-dependent activation of caspase-8. EP4 antagonist (Fig. 3b) or treating macrophages with the mPGES While feedback by PGE2 through EP4 is capable of restricting antagonist (Supplementary Fig. 3) resulted in faster and greater TLR4–TRIF–dependent activation of the TBK1–IRF3 pathway, the LPS-induced loss of TLR4 from the cell surface than that of vehicle- TLR4–TRIF complex is also capable of activating caspase-8, which treated (control) macrophages. leads to limited processing of IL-1β​. To determine whether that sec- Little is known of the signaling determinants that permit CD14- ond TRIF-dependent pathway was sensitive to PGE2, we assessed dependent translocation of TLR4 to endosomes, but it has been the effect of the EP4 antagonist on LPS-dependent activation of

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy

abVehicle EP4 antagonist Vehicle EP4 antagonist LPS time (min) 0 30 60 90 120 180 0 30 60 90 120 180 LPS time (h) 02134 5 01 2345

p-IRF3 p-IRF3

p-TBK1 p-TBK1

Total Total IRF3 IRF3 p-IRF3/IRF3 1.1 0.9 0.4 1.2 1.3 2.4 0.6 p-IRF3/IRF3 0.01 0.89 0.1 0.1 1.2 0.8 1.1 0.002 0.001 0.04 0.10 0.06 0.01 0.44 0.007 0.001 0.17

cdVehicle EP4 antagonist

LPS time (min) 0 30 60 90 120 180 0 30 60 90 120 180 Vehicle EP4 antagonist p-Erk LPS time (min) 0 30 60 90 120 180 0 30 60 90 120 180 p-Jnk Total IκBα p-p38 Total IκBα 1 1 0.6 0.3 0.3 0.01 0.09 0.37 0.91 0.380.42 Total 0.008 p38 p-p38/p38 0.1 0.70.08 0.001 0.0130.0010.0010.0010.77 0.0010.0010.001

e f Antagonist

WT Ptger4 –/– Vehicle EP4 EP3 EP2 LPS time (min) 0 60 120 0 60 120 LPS time (min) 0 60 120 0 60 120 0 60 120 0 60 120

p-IRF3 p-IRF3

Total IRF3 Total IRF3 p-IRF3/IRF-3 p-IRF3/IRF3 0.03 0.001 0.001 0.039 0.03 0.02 0.007 0.045 0.001 0.0110.0020.0640.0680.0010.057 0.0010.0310.011

g mPGES h Vehicle antagonist mPGES Vehicle antagonist LPS time (min) 0 30 60 90 120 180 0 30 60 90 120 180

p-IRF3 LPS time (min) 0 30 60 90 120 180 0 30 60 90 120 180

p-TBK1 Total IκBα Total Total IκBα 1 1 0.4 1.2 0.1 0.07 0.42 0.53 0.03 0.06 IRF3 0.003 0.027 p-IRF3/IRF3 1.1 1.8 2.2 2.3 1.3 0.56 0.91 0.21 0.001 0.450.060.001 i Vehicle PGE2

LPS time (min) 0 60 90 120 0 60 90 120 p-IRF3

Total IRF3 p-IRF3/IRF3 0.570.68 0.34 0.08 0.001 0.0010.1130.110

Fig. 2 | PGE2 activation of EP4 receptor restricts IRF3 activation. a,b, Immunoblot analysis of phosphorylated IRF3 and TBK1 and total IRF3 (left margin) in whole-cell lysates of TEPMs pre-treated for 30 min with vehicle (DMSO throughout) or EP4 antagonist (above blots; as in Fig. 1) and then stimulated for 0–180 min a or 0–5 h b (above lanes) with LPS (100 ng/ml throughout). c, Immunoblot analysis of phosphorylated MAPKs Erk, Jnk and p38 and total p38 (left margin) in whole-cell lysates of TEPMs treated as in a (above blots). d, Immunoblot analysis of total Iκ​Bα ​(left margin) in whole-cell lysates of TEPMs treated as in a (above blots). e, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of BMDMs generated from the femurs of Ptger4−/− mice and their littermates (above blots), then stimulated for 0–120 min (above lanes) with LPS. f, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of TEPMs pre-treated for 30 min with vehicle or an EP4-, EP3- or EP2-specific antagonist (50 μ​M) (above blots), then stimulated for 0–120 min (above lanes) with LPS. g,h, Immunoblot analysis of phosphorylated IRF3 and TBK1 and total IRF3 (g) or total Iκ​Bα​ (h) (left margin) in whole-cell lysates of TEPMs pre-treated for 30 min with vehicle or the mPGES antagonist CAY10526 (10 μ​M) (above blots), then stimulated for 0–180 min (above lanes) with LPS. i, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of TEPMs pre-treated for 60 min with vehicle or PGE2 (1 μ​M) (above blots), then stimulated for 0–180 min (above lanes) with LPS. Numbers below lanes, densitometry (comparison or molecule assessed, left margin). Data are from one experiment representative of at least three separate experiments (n >​ 3) with similar results (a–d,g–i) or one experiment representative of two experiments (n = 2)​ with similar results (e).

Nature Immunology | www.nature.com/natureimmunology NATuRE IMMunoLoGy Articles

a –/– WT Trif

Vehicle mPGES antagonist VehiclemPGES antagonist LPS time (min) 0 30 60 90 120 180 0306090 120 180 LPS time (min) 0 30 60 90 120 180 0306090 120 180 p-IRF3 p-IRF3

Total IRF3 Total IRF3

p-IRF3/ p-IRF3/ 0.01 0.03 0.02 0.02 IRF3 0.0010.0010.008 0.0080.004 0.001 0.003 0.011 IRF3 0.0010.0010.0010.001 0.0010.001 0.0010.0010.0010.0010.0010.001

Vehicle mPGES antagonist VehiclemPGES antagonist

LPS time (min) 0306090 120 180 0306090 120 180 LPS time (min) 0 30 60 90 120 180 0306090 120 180

Total IκBα Total IκBα

Total 1 1 1 1 Total 1 1 0.6 0.8 1.2 0.7 0.9 0.5 0.7 1.1 0.2 0.2 0.2 0.4 0.04 0.05 0.47 0.91 0.01 I B IκBα 0.001 κ α

Vehicle EP4 antagonist Vehicle EP4 antagonist

LPS time (min) 0306090 120 180 0306090 120 180 LPS time (min) 0306090 120 180 0306090 120 180

Total IκBα Total IκBα

b c 100 Vehicle Vehicle EP4 antagonist EP4 75 antagonist LPS time (min) 0306090 120 180 0 30 60 90 120 180 * * Unstimulated 50 p-SykTYR525/526

25 Surface TLR4 (% ) Total Syk

0 p-Syk/ 0.1 0.3 0.3 0.6 0.1 0.6 0.8 0.7 1.2 025 50 75 100 Syk 0.18 0.41 0.92 Time (min)

d mPGES Vehicle antagonist

P3C time (min) 0306090 120 180 0 30 60 90 120 180

p-IRF3

p-JnK

Total IκBα

Total IRF3

Total 1 1 1 1 0.6 0.9 IκBα 0.02 0.92 0.98 0.03 0.99 0.95

Fig. 3 | The PGE2–EP4 axis regulates TLR4 signaling via TRIF. a, Immunoblot analysis of phosphorylated and total IRF3 (top) or total Iκ​Bα ​(middle and bottom) (left margin) in whole-cell lysates of wild-type (WT) TEPMs (left) or Trif−/− TEPMs (right) pre-treated for 30 min with vehicle (DMSO) or an antagonist of mPGES or EP4 (above blots), then stimulated for 0–180 min (above lanes) with LPS (100 ng/ml). b, Flow cytometry of wild-type TEPMs left unstimulated or pre-treated for 15 min with vehicle (DMSO) or EP4 antagonist (key), then stimulated for 0–100 min (horizontal axis) with LPS (100 ng/ml), placed on ice and stained for cell-surface TLR4; results are presented as the average mean fluorescent intensity relative to that at the starting point, set as 100%. *P < 0.05​ (Student’s unpaired t-test). c, Immunoblot analysis of Syk phosphorylated at Tyr525 and Tyr526 (p-SykTYR525/526) and total Syk (left margin) in whole-cell lysates of wild-type TEPMs treated as in a, bottom row (above lanes and blots). d, Immunoblot analysis of phosphorylated IRF3 and Jnk and total Iκ​Bα​ and IRF3 (left margin) in whole-cell lysates of TEPMs pre-treated as in (a), then stimulated for 0–180 min (above lanes) with the synthetic TLR2 ligand P3C (300 ng/ml). Data are from one experiment representative of at least three separate experiments (n ≥ 3)​ with similar results (a,c,d) or are from n =​ 4 independent experiments (b; mean ±​ s.e.m.). caspase-8. Blockade of EP4 augmented the TRIF-dependent activa- IRF3, in LPS-stimulated wild-type macrophages relative to that in tion of caspase-8, as indicated by increased abundance of the cas- Trif–/– macrophages (Fig. 4a). It has been reported that the stimula- pase-8 cleavage product p18, as well as enhanced phosphorylated tion of macrophages with LPS alone results in low-level process-

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy abWT Trif –/– c Vehicle LPS + EP4 EP4 EP4 LPS + LPS + Vehicle antagonist Vehicle antagonist antagonist LPS LPS EP4 ATP LPS time LPS time 0 120 210 0 120 210 0 120 210 0 120 210 (min) (min) –/– LPS time (h) – 33666 3 Caspase-8 WT Trif p55 Caspase-8 p55 350 Pro-caspase-1 Caspase-8 300 * p18 Caspase-8 250 p18 200 p-IRF3 p-IRF3 150 Caspase-1 100 p20

IL-1 β (pg/ml) NS p-Erk 50 p-Erk p20 0 0.03 0.12 p18/ p18/ 0.001 0.001 0.0010.0010.001 p55 0.01 p55 0.0010.0070.003 0.0010.028 0.0010.0010.001 0.0010.001 0.001

Fig. 4 | PGE2–EP4 restricts the LPS-induced, TRIF-dependent activation of caspase-8 and atypical processing of IL-1β. a, Immunoblot analysis of the uncleaved caspase-8 pro-form p55 and caspase-8 cleavage product p18, as well as phosphorylated IRF3 and Erk (left margin), in whole-cell lysates of wild-type TEPMs (left) or Trif−/− TEPMs (right) pre-treated for 30 min with vehicle (DMSO) or EP4 antagonist (50 μ​M) (above blots), then stimulated for 0–210 min (above lanes) with LPS (100 ng/ml). b, ELISA of mature IL-1β​ in supernatants of wild-type or Trif−/− TEPMs (above plot) primed for 4 h with Pam3Cys (300 ng/ml), then washed twice with PBS, followed by replacement of the medium and stimulation for an additional 5 h with medium alone (Vehicle) or LPS (100 ng/ml) with or without EP4 antagonist (50 μ​M) (key). NS, not significant (P >​ 0.05); *P = 0.005​ (Student’s unpaired two-tailed t-test). c, Immunoblot analysis of pro-caspase-1 and the active caspase-1 form p20 (left margin) in whole-cell lysates of TEPMs left unstimulated (–) or stimulated for 3 h or 6 h (above lanes) with LPS alone (left) or LPS plus EP4 antagonist (LPS + EP4),​ or with LPS for 3 or 6 h, with ATP (5 mM) added for the final 30 min (LPS +​ ATP). Data are from one experiment representative of three separate experiments (n =​ 3) with similar results (a,c) or are from n =​ 3 separate experiments (b; mean +​ s.e.m.).

ing and release of mature IL-1β​ in a manner that is dependent on PGE2–EP4 regulates infection in vitro and in vivo. To determine 13,15 TRIF and caspase-8 and is independent of caspase-1 . We specu- whether the observed PGE2–EP4 feedback might be important in lated that release of PGE2 might limit this TRIF–caspase-8–depen- restricting the activation of IRF3 during in vitro infection with live dent effect on IL-1β.​ Antagonizing EP4 significantly enhanced the Gram-negative pathogens, we infected peritoneal macrophages LPS-dependent release of IL-1β ​in a TRIF-dependent fashion and with intracellular ST at a low multiplicity of infection (MOI) of 4 in the absence of a second inflammasome-activating stimulus (Fig. in the presence or absence of the mPGES inhibitor and assessed

4b). Because activation of EP4 by exogenous PGE2 has been shown activation of IRF3 6 h later. In the absence of the mPGES inhibi- to inhibit inflammasome NLRP3–dependent activation of cas- tor, no detectable activation of IRF3 was observed after infection pase-1 and processing of IL-1β​ in response to activators of LPS and with ST (Fig. 6a). Inhibition of PGE2 synthesis, however, substan- NLRP334, we assessed caspase-1 directly but found no activation tially increased the activation of IRF3 in response to ST in vitro (Fig. due to the presence of the EP4 antagonist (Fig. 4c). These results 6a). ST is an intracellular pathogen and presumably presents LPS to demonstrated that both of the known signal-transduction pathways TLR4 during the entire course of infection. Therefore, we sought initiated by TLR4–TRIF were negatively regulated by PGE2. to determine whether EP4 was also active in restricting the activa- tion of IRF3 following 60 min of incubation of macrophages with cAMP-dependent effector molecules are regulators of TRIF. either of two extracellular, enteric Escherichia coli strains (enterohe- Having established that signaling through the receptor EP4 was morrhagic E. coli or enteropathogenic E. coli) or the mouse extra- required for the restriction of TRIF-dependent signals downstream cellular enteric pathogen Citrobacter rodentium (MOI =​ 25). As of TLR4, we sought to define EP4-activated pathways necessary for observed for ST, restriction of the production of PGE2 by inhibition this effect. There are two well-defined G protein–activated signal- of mPGES enhanced the activation of IRF3 following infection with transduction cascades downstream of EP4; the first leads to activa- enteropathogenic E. coli, enterohemorrhagic E. coli or C. rodentium tion of phosphatidylinositol-3-OH kinase (PI(3)K) and the kinase (Fig. 6b).

Akt, and the second activates an endogenous adenylate cyclase (AC) To confirm that the PGE2–EP4 axis limited the production of that leads to accumulation of cAMP and activation of the cAMP- type I interferons during in vivo infection with a Gram-negative dependent effector molecules PKA (a kinase) and EPAC (the nucle- pathogen, we pre-treated C57BL/6 J mice with a single intraperi- otide-exchange factor for small GTPases)37 (Fig. 5a). While complete toneal injection of saline or EP4 antagonist (50 μ​g per mouse) inhibition of basal and LPS-inducible phosphorylation of Akt was 60 min before intraperitoneal infection of the mice with ST. At 8 h achieved with the Akt antagonist, that antagonist had no effect on after infection, mice were euthanized and the spleen and liver were the LPS-dependent activation of IRF3 (Fig. 5b). In contrast, inhibi- harvested for gene-expression analysis. Mice that received the EP4 tion of either of the cAMP-dependent effector molecules (PKA or antagonist displayed significantly higher Ifnb expression in the EPAC) led to an increase in the LPS-dependent activation of IRF3 spleen than that of saline-treated mice (Fig. 6c), with a similar trend (Fig. 5c,d), which suggested that the AC-dependent pathway medi- observed in the liver (data not shown). Consistent with our in vitro ated EP4’s effects. To assess directly the ability of cAMP to suppress studies, the effects of the EP4 antagonist displayed gene specificity TLR4-dependent activation of IRF3, we used recombinant AC toxin and did not alter the in vivo expression of Tnf, but we did observe from Bordetella pertussis, which rapidly catalyzes the production of reduced expression of Il1b mRNA (Fig. 6c), in agreement with a 38 39 cytosolic cAMP independently of host enzymes . The addition of published report on a role for PGE2 in Il1b transcription . Analysis B. pertussis AC toxin reduced the TLR4-mediated activation of IRF3 of ST colony-forming units in the spleen did not show substan- in a dose-dependent way (Fig. 5e), in addition to slowing the kinet- tially altered numbers of this bacteria in either group at this early ics with which TLR4 was internalized (Fig. 5f). Thus, the inhibitory time point (Fig. 6d). These results indicated that signaling via the effects of EP4 on the TLR4–TRIF pathway were mediated by cAMP- PGE2–EP4 axis was important in limiting the activation of IRF3 and dependent effector molecules. induction of IFN-β​ during infection with Gram negative bacteria in

Nature Immunology | www.nature.com/natureimmunology NATuRE IMMunoLoGy Articles

a b EP4

AKt Vehicle antagonist LPS time (min) 0 60 120 180 0 60 120 180 Gα s G p-IRF3 AC αi PI(3)K p-AKt

cAMP AKt Total AKt p-IRF3/ 0.5 PKA EPAC 0.61 0.60 0.23 0.66 0.01 IRF3 0.001 0.001

cdEPAC PKA Vehicle antagonist Vehicle antagonist LPS time LPS time (min) 060 120 180 060 120 180 (min) 0 60 120 180 060 120 180 p-IRF3 p-IRF3

Total Total IRF3 IRF3 p-IRF3/ P-IRF3/ 0.3 0.03 0.01 0.08 0.04 0.01 0.04 0.04 0.13 0.04 IRF3 0.001 0.008 0.001 IRF3 0.001 0.005 0.001

ef ACTx (0.5 ng) ACTx (5 ng) 100 EP4 EP4 EP4 Unstimulated Vehicle antagonist antagonist antagonist LPS 75 LPS time * (min) 0060 120 60 120 0 60 120 0 60 120 * ACTx + LPS 50 p-IRF3 25

Total Surface TLR4 (%) IRF3 0 P-IRF3/ 025 50 75 100 IRF3 0.01 0.04 0.03 0.0010.025 0.001 0.035 0.001 0.011 0.001 0.008 0.004 Time (min)

Fig. 5 | cAMP-dependent effector molecules are negative regulators of the TLR4–TRIF pathway. a, Model of signaling pathways downstream of EP4. b–d, Immunoblot analysis of phosphorylated IRF3 (b–d) and Akt (b) and total Akt (b) and IRF3 (c,d) (left margin) in whole-cell lysates of TEPMs pre- treated for 30 min with vehicle (DMSO) or the Akt antagonist AKTi (10 μ​M) (b), the PKA antagonist H89 (10 μ​M) (c) or the EPAC antagonist ESI09 (10 μ​M) (d), then stimulated for 0–180 mine (above lanes) with LPS (100 ng/ml). e, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of TEPMs stimulated for 0–180 min (above lanes) with LPS plus vehicle (far left), EP4 antagonist (middle left), or EP4 antagonist and recombinant purified AC toxin (ACTx) (0.5 ng (middle right) or 5 ng (far right)). f, Flow cytometry analyzing surface TLR4 on TEPMs left unstimulated or stimulated with LPS (100 ng/ml) after no pre-treatment (LPS) or after pre-incubation for 30 min with purified ACTx (2.5 ng/ml) (ACTx +​ LPS) (key). P = 0.01​ (Student’s two-tailed t-test). Data are from one experiment representative of at least three separate experiments (n ≥ 3)​ with similar results (b–e) or are from n =​ 3 individual experiments (f; mean ±​ s.e.m.). vitro and in vivo. The expression of type I interferons during in vivo to restrict both signaling initiated at the cell surface (dependent on ST infection can promote pathogenesis and increase mortality in TIRAP (Mal) and MyD88) and endosomal signaling (dependent on mice17,18,40,41. We therefore hypothesized that in vivo administration TRAM and TRIF). While cell-surface signals dependent on TLR4– of EP4 antagonists during systemic infection with ST would result MyD88 that activate NF-κB​ and MAPK are subject to negative regu- in increased mortality due, at least in part, to elevated production lation by an A20 feedback circuit, A20 does not limit the endosomal of IFN-β​. To determine the effect of EP4 antagonism on the course activation of IRF3 via TLR4 and TRIF24. A negative feedback loop of ST infection, we infected mice intraperitoneally with ST (1 ×​ 103 that restricts the TRIF complex has not been generally described; colony-forming units), followed by twice-daily injection of the EP4 however, there are reports of other mechanisms that limit aspects antagonist or saline (as a control). Mice that received the EP4 antag- of TRIF activity. The deubiquitinase DUBA has been shown to act onist succumbed more rapidly to infection than did saline-treated to inhibit the K63 ubiquitin ligase TRAF3, an activity that limits the 42 mice (Fig. 6e), consistent with a key role for PGE2–EP4 in the host TRIF-dependent activation of IRF3 specifically . Additionally, it defense against ST. Our data are consistent with a model in which has been demonstrated that inflammasome activation of caspase-1

PGE2 is rapidly released in response to ligation of TLR4 by LPS and during some bacterial infections can cleave TRIF and thereby limit selectively suppresses TLR4–TRIF signal transduction by activating TRIF-dependent signaling43; however, such a mechanism would be cAMP-dependent effector molecules to restrict endosomal translo- operative only in the presence of second signal to permit maturation cation of TLR4 (Supplementary Fig. 6). of the inflammasome. Our current observations are both novel and important because Discussion we have described for the first time, to our knowledge, the action

Mechanisms that restrict TLR4-dependent production of cyto- of an autocrine or paracrine PGE2-dependent feedback loop down- kines are critical for limiting inflammatory damage and speeding stream of the activation of TLR4 by LPS that limits both the ampli- the return to host homeostasis. The unique cell biology of TLR4- tude and the duration of TRAM–TRIF–dependent signaling by dependent signaling requires that feedback mechanisms be in place restricting the CD14-dependent trafficking of TLR4–MD2 into

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy

ab

T Mock S Mock EHEC EPEC C. rodentium mPGES mPGES –+ –+ –+ –+ –+ –+ antagonist antagonist

p-IRF3 p-IRF3

Total Total IRF3 IRF3 p-IRF3/ p-IRF3/ 0.1 IRF3 0.09 0.06 0.03 0.22 0.001 0.001 0.001 IRF3 0.001 0.001 0.001 0.008

c 10 * 100 20 * 8 80 15 6 60 10 4 40 II1b (fold) Ifnb (fold) Tnf (fold) 2 5 20 0 0 0 Vehicle EP4 Vehicle EP4 Vehicle EP4 antagonist antagonist antagonist

deCFU 100 105 Saline 80 EP4 4 10 60 antagonist

40 tissue CFU/ g 103 Survival (%) * 20

102 0 Vehicle EP4 antagonist 0246810 Time (d)

Fig. 6 | PGE2 restricts IRF3 activation and IFN-β production in response to Gram negative pathogens in vitro and in vivo. a, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of TEPMs mock infected or infected with ST (MOI = 4)​ for 30 min (above blots), followed by treatment with gentamycin and the addition of an mPGES antagonist (10 μ​M) ( +​ ) or not (–) (above lanes) and incubation for an additional 6 h. b, Immunoblot analysis of phosphorylated and total IRF3 (left margin) in whole-cell lysates of TEPMs mock infected or co-cultured with enterohemorrhagic E. coli (EHEC), enteropathogenic E. coli (EPEC) or C. rodentium (MOI =​ 25) for 60 min (above blots), followed by treatment gentamycin and the mPGES antagonist (or not) and culture as in a (above lanes). c, Expression of Ifnb, Tnf and Il1b in the spleen (presented here) and liver (data not shown) of wild- type C57BL6/J mice (n =​ 5) pre-treated by intraperitoneal injection of saline alone (Vehicle) or with the EP4 antagonist (50 μg​ per mouse) (horizontal axis) 60 min before intraperitoneal infection with ST (1 ×​ 108 per mouse), followed by analysis 8 h later; results were normalized to those of Hprt. *P =​ 0.0017 (Ifnb) or 0.023 (Il1b) (Student’s two-tailed t-test). d, Quantification of viable bacteria (as colony-forming units (CFU)) in the spleen of mice infected and treated as in c, assessed at 8 h by homogenization, followed by serial dilution on streptomycin-containing LB agar plates. e, Survival of mice given a single dose of saline or EP4 antagonist (50 μ​g per mouse) (key) 1 h before intraperitoneal infection with 1 ×​ 103 ST, followed by twice-daily injection of saline or EP4 antagonist (50 μ​g per mouse) each day (key). *P =​ 0.001 (one-sided Mantel-Cox test). Each symbol (c,d) represents an individual mouse; small horizontal lines indicate the mean ( ±​ s.e.m.). Data are from one experiment representative of four (a) or two (b) separate experiments (n =​ 4 (a) or n =​ 2 (b)) with similar results (a,b) or are from one experiment with n =​ 5 mice per group (c,d) or n =​ 10 mice per group (e).

endosomes necessary for the recruitment of TRAM–TRIF to the has an affinity for PGE2 that is about seven- to eight fold greater 46 cytosolic TIR domain. The biochemistry of PGE2 makes it uniquely than that of EP2 , which possibly makes EP4 uniquely suited to suited to serve in a feedback capacity to regulate TLR4 responses sense and respond to the low autocrine or paracrine levels of PGE2 for two reasons. First, PGE2 can be rapidly synthesized and released secreted soon after ligation of TLR4. Notably, while some published within minutes after ligation of TLR4, independently of de novo in vitro studies have shown a broadly suppressive effect of PGE2 25 47–49 transcription and translation . Second, PGE2 is an autocrine-act- on all TLR4-driven cytokine responses , such studies have typi- ing factor with an extremely short life-span in vivo of less than 30 s cally relied on relatively high levels of exogenously added PGE2 that 44 (ref. ), which would permit secreted PGE2 to restrict TRIF sig- might simultaneously ligate multiple distinct receptors for PGE2 naling in a cell-autonomous manner with limited systemic effects, and produce a regulatory outcome distinct from that produced by unlike a relatively long-lived protein cytokine. the autocrine or paracrine loop described here.

While PGE2 can be recognized by four distinct G protein–coupled Our data also support the hypothesis that downstream of EP4, receptors (EP1–EP4)33, we identified EP4 specifically as mediating the activation of endogenous AC enzymes and generation of cAMP the actions of secreted PGE2 directed against the TRIF pathway, and are important for the action of EP4 in restricting TRIF-mediated published reports highlight a growing appreciation of the impor- signaling. The identification of the cAMP-dependent effector mol- tance of EP4 in regulating inflammatory responses in vitro and in ecules PKA and EPAC in limiting TRIF-dependent activation of vivo34,45. While EP4 shares the greatest sequence and signaling simi- IRF3 is intriguing, given the effects of EP4 on the internalization larity (including activating AC) with the PGE2 receptor EP2, EP4 of TLR4. The movement of TLR4 from the cell surface to an early

Nature Immunology | www.nature.com/natureimmunology NATuRE IMMunoLoGy Articles endosome compartment is essential for the recruitment of TRIF and 6. Kagan, J. C. et al. TRAM couples endocytosis of Toll-like receptor 4 to the activation of IRF3. While cell-surface expression of CD14 is abso- induction of interferon-β​. Nat. Immunol. 9, 361–368 (2008). 7. Zanoni, I. et al. CD14 controls the LPS-induced endocytosis of Toll-like lutely required for the internalization of TLR4 in response to LPS, receptor 4. Cell 147, 868–880 (2011). whether the rate of such internalization is constant or is dynamic 8. Jiang, Z. et al. CD14 is required for MyD88-independent LPS signaling. Nat. and subject to regulation has not been investigated. Our work rep- Immunol. 6, 565–570 (2005). resents the first report, to our knowledge, showing that the kinet- 9. Fitzgerald, K. A. et al. IKKepsilon and TBK1 are essential components of the ics with which TLR4 is internalized are subject to regulation by the IRF3 signaling pathway. Nat. Immunol. 4, 491–496 (2003). 10. Fitzgerald, K. A. et al. LPS-TLR4 signaling to IRF-3/7 and NF-κ​B involves the extracellular inflammatory milieu. In this context, it is also notable toll adapters TRAM and TRIF. J. Exp. Med. 198, 1043–1055 (2003). that poly(I:C), a ligand for TLR3, a receptor that does not require 11. Yamamoto, M. et al. Role of adaptor TRIF in the MyD88-independent trafficking to engage TRIF and is not inhibited by the EP4 antag- toll-like receptor signaling pathway. Science 301, 640–643 (2003). 12. Yamamoto, M. et al. TRAM is specifcally involved in the Toll-like receptor onist, elicited much less release of PGE2 than did LPS. While few details are known about the molecular machinery that coordinates 4-mediated MyD88-independent signaling pathway. Nat. Immunol. 4, 1144–1150 (2003). the trafficking of TLR4, TRAM and TRIF to endosomes, it is clear 13. Maelfait, J. et al. Stimulation of Toll-like receptor 3 and 4 induces interleukin- that small GTPases (specifically, Arf6 and Rab11a) are required for 1β​ maturation by caspase-8. J. Exp. Med. 205, 1967–1973 (2008). this50,51. Both PKA and EPAC have been shown to negatively regulate 14. He, S., Liang, Y., Shao, F. & Wang, X. Toll-like receptors activate programmed endosomal trafficking by GTPases of the Rab family and Rab11a in necrosis in macrophages through a receptor-interacting kinase-3-mediated 52–54 pathway. Proc. Natl Acad. Sci. USA 108, 20054–20059 (2011). response to cAMP-generating bacterial toxins . In this context, 15. Moriwaki, K., Bertin, J., Gough, P. J. & Chan, F. K. A. A RIPK3-caspase 8 PKA and EPAC have been shown to act synergistically, targeting dif- complex mediates atypical pro-IL-1β​ processing. J. Immunol. 194, ferent aspects of the Rabll–exocyst complex, and thereby limit the 1938–1944 (2015). kinetics of early recycling endosomes55. Conceivably, a similar role 16. Casanova, J. L., Abel, L. & Quintana-Murci, L. Human TLRs and IL-1Rs in for PKA and EPAC could operate in response to endogenously pro- host defense: natural insights from evolutionary, epidemiological, and clinical genetics. Annu. Rev. Immunol. 29, 447–491 (2011). duced PGE2 to limit assembly of the TLR4–TRIF complex. Further 17. Perkins, D. J. et al. Salmonella Typhimurium co-opts the host type I IFN studies will be needed to elucidate the precise intermolecular inter- system to restrict macrophage innate immune transcriptional responses actions between the cAMP effector molecules and elements of the selectively. J. Immunol. 195, 2461–2471 (2015). TLR4–TRIF pathway; however, a potential implication of this model 18. Deriu, E. et al. Infuenza virus afects intestinal microbiota and secondary is that the rate of CD14-dependent internalization of TLR4 might Salmonella infection in the gut through type I interferons. PLoS Pathog. 12, e1005572 (2016). be responsive to and reflective of the local concentration of PGE2. 19. Henry, T. et al. Type I IFN signaling constrains IL-17A/F secretion by γ​δ​ T Our current work has important implications for understand- cells during bacterial infections. J. Immunol. 184, 3755–3767 (2010). ing how the different modes of TLR4 signaling shape the outcome 20. Mayer-Barber, K. D. et al. Innate and adaptive interferons suppress IL-1α​ and of bacterial infection. In vivo, establishing a balance between pro- IL-1β ​production by distinct pulmonary myeloid subsets during inflammatory cytokines (TNF, IL-1 ​and IL-6) and type I inter- Mycobacterium tuberculosis infection. Immunity 35, 1023–1034 (2011). β 21. Rayamajhi, M., Humann, J., Penheiter, K., Andreasen, K. & Lenz, L. L. ferons is important in pathogenesis. Inflammatory cytokines are Induction of IFN-α​β​ enables Listeria monocytogenes to suppress macrophage clearly needed to restrict bacterial growth and, ultimately, to clear activation by IFN-γ​. J. Exp. Med. 207, 327–337 (2010). microbes16, while some coincident production of type I interferons 22. Heyninck, K. & Beyaert, R. A20 inhibits NF-κ​B activation by dual ubiquitin- can be beneficial by limiting damage due to systemic inflamma- editing functions. Trends Biochem. Sci. 30, 1–4 (2005). tion56. However, the production type I interferons must be carefully 23. Ma, A. & Malynn, B. A. A20: linking a complex regulator of ubiquitylation to immunity and human disease. Nat. Rev. Immunol. 12, 774–785 (2012). regulated, as it has been shown in various other bacterial infection 24. Turer, E. E. et al. Homeostatic MyD88-dependent signals cause lethal models that the suppressive effects of type I interferons promote infammation in the absence of A20. J. Exp. Med. 205, 451–464 (2008). bacterial growth and exacerbate disease17,18,21,29,41,57,58. Therefore, 25. Buczynski, M. W. et al. TLR-4 and sustained calcium agonists synergistically striking a precise balance between outcomes dependent on TLR4– produce eicosanoids independent of protein synthesis in RAW264.7 cells. MyD88 and those dependent on TLR4–TRIF is probably key to the J. Biol. Chem. 282, 22834–22847 (2007). 26. Hessle, C. C., Andersson, B. & Wold, A. E. Gram-negative, but not re-establishment of homeostasis. Our data support the concept that Gram-positive, bacteria elicit strong PGE2 production in human monocytes. the PGE2–EP4 axis is an important conduit of cross-talk between Infammation 27, 329–332 (2003). the two signaling pathways. 27. Moore, R. N., Urbaschek, R., Wahl, L. M. & Mergenhagen, S. E. Prostaglandin regulation of colony-stimulating factor production by Online content lipopolysaccharide-stimulated murine leukocytes. Infect. Immun. 26, 408–414 (1979). Any methods, additional references, Nature Research reporting 28. Bowman, C. C. & Bost, K. L. Cyclooxygenase-2-mediated prostaglandin E2 summaries, source data, statements of data availability and asso- production in mesenteric lymph nodes and in cultured macrophages ciated accession codes are available at https://doi.org/10.1038/ and dendritic cells afer infection with Salmonella. J. Immunol. 172, s41590-018-0243-7. 2469–2475 (2004). 29. Mayer-Barber, K. D. et al. Host-directed therapy of tuberculosis based on interleukin-1 and type I interferon crosstalk. Nature 511, 99–103 (2014). Received: 27 March 2018; Accepted: 14 September 2018; 30. Dennis, E. A. & Norris, P. C. Eicosanoid storm in infection and Published: xx xx xxxx infammation. Nat. Rev. Immunol. 15, 511–523 (2015). 31. von Moltke, J. et al. Rapid induction of infammatory lipid mediators by the References infammasome in vivo. Nature 490, 107–111 (2012). 1. Medzhitov, R., Preston-Hurlburt, P. & Janeway, C. A. Jr. A human homologue 32. Boulet, L. et al. Deletion of microsomal prostaglandin E2 (PGE2) synthase-1 of the Drosophila Toll protein signals activation of adaptive immunity. Nature reduces inducible and basal PGE2 production and alters the gastric 388, 394–397 (1997). prostanoid profle. J. Biol. Chem. 279, 23229–23237 (2004). 2. Poltorak, A. et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr 33. Sugimoto, Y. & Narumiya, S. Prostaglandin E receptors. J. Biol. Chem. 282, mice: mutations in Tlr4 gene. Science 282, 2085–2088 (1998). 11613–11617 (2007). 3. Shimazu, R. et al. MD-2, a molecule that confers lipopolysaccharide 34. Mortimer, L., Moreau, F., MacDonald, J. A. & Chadee, K. NLRP3 responsiveness on Toll-like receptor 4. J. Exp. Med. 189, 1777–1782 (1999). infammasome inhibition is disrupted in a group of auto-infammatory 4. Brubaker, S. W., Bonham, K. S., Zanoni, I. & Kagan, J. C. Innate immune disease CAPS mutations. Nat. Immunol. 17, 1176–1186 (2016). pattern recognition: a cell biological perspective. Annu. Rev. Immunol. 33, 35. Roberts, Z. J. et al. Te chemotherapeutic agent DMXAA potently and 257–290 (2015). specifcally activates the TBK1-IRF-3 signaling axis. J. Exp. Med. 204, 5. Bryant, C. E., Symmons, M. & Gay, N. J. Toll-like receptor signalling 1559–1569 (2007). through macromolecular protein complexes. Mol. Immunol. 63, 36. Hoebe, K. et al. Identifcation of Lps2 as a key transducer of MyD88- 162–165 (2015). independent TIR signalling. Nature 424, 743–748 (2003).

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy

37. Konya, V., Marsche, G., Schuligoi, R. & Heinemann, A. E-type prostanoid 55. Guichard, A. et al. Anthrax edema toxin disrupts distinct steps receptor 4 (EP4) in disease and therapy. Pharmacol. Ter. 138, 485–502 (2013). in Rab11-dependent junctional transport. PLoS Pathog. 13, 38. Perkins, D. J., Gray, M. C., Hewlett, E. L. & Vogel, S. N. Bordetella pertussis e1006603 (2017). adenylate cyclase toxin (ACT) induces cyclooxygenase-2 (COX-2) in murine 56. Castiglia, V. et al. Type I interferon signaling prevents IL-1β​-driven lethal macrophages and is facilitated by ACT interaction with CD11b/CD18 systemic hyperinfammation during invasive bacterial infection of sof tissue. (Mac-1). Mol. Microbiol. 66, 1003–1015 (2007). Cell Host Microbe 19, 375–387 (2016). 39. Zasłona, Z. et al. Te induction of pro-IL-1β​ by lipopolysaccharide requires 57. Auerbuch, V., Brockstedt, D. G., Meyer-Morse, N., O’Riordan, M. & Portnoy, D. A. endogenous prostaglandin E2 production. J. Immunol. 198, 3558–3564 (2017). Mice lacking the type I interferon receptor are resistant to Listeria 40. Robinson, N. et al. Type I interferon induces necroptosis in macrophages monocytogenes. J. Exp. Med. 200, 527–533 (2004). during infection with Salmonella enterica serovar Typhimurium. Nat. 58. Nagarajan, U. M. et al. Type I interferon signaling exacerbates Chlamydia Immunol. 13, 954–962 (2012). muridarum genital infection in a murine model. Infect. Immun. 76, 41. Stefan, K. L., Fink, A., Surana, N. K., Kasper, D. L. & Dasgupta, S. Type I 4642–4648 (2008). interferon signaling restrains IL-10R+ colonic macrophages and dendritic cells and leads to more severe Salmonella colitis. PLoS One 12, e0188600 (2017). Acknowledgements 42. Kayagaki, N. et al. DUBA: a deubiquitinase that regulates type I interferon We thank E. Hewlett (University of Virginia) for purified AC from B. pertussis; L. Wahl production. Science 318, 1628–1632 (2007). (National Institute of Dental and Craniofacial Research, US National Institutes of Health) 43. Jabir, M. S. et al. Caspase-1 cleavage of the TLR adaptor TRIF inhibits for blood from healthy human volunteers at the Department of Transfusion Medicine autophagy and β-interferon​ production during Pseudomonas aeruginosa of the US National Institutes of Health; R. Ernst (University of Maryland, Baltimore) for infection. Cell Host Microbe 15, 214–227 (2014). Salmonella Typhimurium strain SL1344; J.B. Kaper (University of Maryland, Baltimore) 44. Bothwell, W., Verburg, M., Wynalda, M., Daniels, E. G. & Fitzpatrick, F. A. A for enterohemorrhagic and enteropathogenic E. coli and C. rodentium; R. Rajaiah for radioimmunoassay for the unstable pulmonary metabolites of prostaglandin technical assistance with the TLR4-internalization assay; and J.C. Blanco and D. Prantner E1 and E2: an indirect index of their in vivo disposition and for critique of this manuscript. Flow cytometry and microarray analyses were performed pharmacokinetics. J. Pharmacol. Exp. Ter. 220, 229–235 (1982). at the Center for Innovative Biomedical Resources at the University of Maryland, School 45. Dufn, R. et al. Prostaglandin E2 constrains systemic infammation through of Medicine. Supported by the US National Institutes of Health (AI123371 and AI125215 an innate lymphoid cell-IL-22 axis. Science 351, 1333–1338 (2016). to S.N.V.; and AR061491 to B.K.). 46. Fujino, H., West, K. A. & Regan, J. W. Phosphorylation of glycogen synthase kinase-3 and stimulation of T-cell factor signaling following activation of EP2 and EP4 prostanoid receptors by prostaglandin E2. J. Biol. Chem. 277, Author contributions 2614–2619 (2002). D.J.P. performed the majority of the experiments; K.R. performed TLR4-translocation 47. Jing, H., Vassiliou, E. & Ganea, D. Prostaglandin E2 inhibits production of experiments; A.-M.H. performed infection with E. coli and C. rodentium; W.L. performed the infammatory chemokines CCL3 and CCL4 in dendritic cells. J. Leukoc. ELISA of IFN-β ​and analyzed the data; S.N. bred and genotyped all mice with targeted Biol. 74, 868–879 (2003). mutations; B.K. developed, bred, genotyped and isolated femurs from Ptger4–/– mice and 48. Kunkel, S. L. et al. Prostaglandin E2 regulates macrophage-derived tumor their littermates; and D.J.P. and S.N.V. conceived of the study, designed the experiments necrosis factor gene expression. J. Biol. Chem. 263, 5380–5384 (1988). and wrote the manuscript. 49. Xu, X. J., Reichner, J. S., Mastrofrancesco, B., Henry, W. L. Jr. & Albina, J. E. Prostaglandin E2 suppresses lipopolysaccharide-stimulated IFN-β​ production. J. Immunol. 180, 2125–2131 (2008). Competing interests 50. Husebye, H. et al. Te Rab11a GTPase controls Toll-like receptor 4-induced The authors declare no competing interests. activation of interferon regulatory factor-3 on phagosomes. Immunity 33, 583–596 (2010). 51. Van Acker, T. et al. Te small GTPase Arf6 is essential for the Tram/Trif Additional information pathway in TLR4 signaling. J. Biol. Chem. 289, 1364–1376 (2014). Supplementary information is available for this paper at https://doi.org/10.1038/ 52. Guichard, A. et al. Cholera toxin disrupts barrier function by inhibiting s41590-018-0243-7. exocyst-mediated trafcking of host proteins to intestinal cell junctions. Cell Reprints and permissions information is available at www.nature.com/reprints. Host Microbe 14, 294–305 (2013). Correspondence and requests for materials should be addressed to D.J.P. or S.N.V. 53. Guichard, A. et al. Anthrax toxins cooperatively inhibit endocytic recycling by the Rab11/Sec15 exocyst. Nature 467, 854–858 (2010). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 54. Guichard, A., Nizet, V. & Bier, E. RAB11-mediated trafcking in host- published maps and institutional affiliations. pathogen interactions. Nat. Rev. Microbiol. 12, 624–634 (2014). © The Author(s), under exclusive licence to Springer Nature America, Inc. 2018

Nature Immunology | www.nature.com/natureimmunology NATuRE IMMunoLoGy Articles

Methods medium only. The monolayers were incubated for 1 h at 37 °C in 7% CO2. Animals and cells. Animal work performed for this study complied with all The medium was then replaced with 1 ml of fresh medium containing 100 μ​g/ml applicable provisions of the Animal Welfare Act; US Government Principles gentamycin and 10 μ​M CAY10526 (for drug-treated samples), and cells were for the Utilization and Care of Vertebrate Animals Used in Testing, Research, incubated for an additional 6 h. and Training; Public Health Services Policy on the Humane Care and Use of For in vivo ST-infection experiments to measure cytokine expression, wild- Laboratory Animals; and Guide for the Care and Use of Laboratory Animals type C57BL/6 J mice were pre-treated by intraperitoneal injection of saline or (8th edition). Te protocol for this work was approved by the Institutional Animal BGC201531 (50 μ​g) and, after 60 min, mice were infected intraperitoneally as 17 Care and Use Committee of the University of Maryland, School of Medicine. described previously . ST-infected mice were killed 8 h later and the spleen and Primary thioglycollate-elicited peritoneal macrophages (TEPMs) were prepared liver were harvested for gene-expression analysis. For in vivo ST lethality studies, as described previously59. In brief, 3 ml of 3% sterile thioglycollate (Remel) was mice were given a single intraperitoneal injection of BGC201531 (50 μg)​ one hour 3 injected intraperitoneally into 6- to 8-week-old wild-type C57BL/6 J mice (Jackson before intraperitoneal infection with 1 ×​ 10 ST. On each successive day, mice Laboratories). 4 d later, macrophages were harvested by peritoneal lavage with were given injection of BGC20-1531 (50 μ​g) twice daily and were monitored for sterile saline. TRIF-null (Trif–/–) mice backcrossed onto a C57BL/6 J background morbidity and mortality. (n ≥ ​10) were bred in-house as described previously60. Femurs from Ptger4–/– (EP4- 6 defcient) mice and their Ptger4+/+ littermates61,62 were used to generate BMDMs Macrophage stimulation. TEPMs or BMDMs were plated at 2 ×​ 10 cells per well in as described previously63. Te human THP-1 cells were plated at 2 ×​ 106 cells per 12-well plates and were allowed to ‘rest’ overnight. Macrophage cultures were pre- well and were treated with 100 nM phorbol 12-myristate 13 acetate (PMA) for 48 h, treated for 30 min with vehicle or inhibitor, followed by stimulation with LPS (100 ng/ afer which the cells were washed and the PMA was removed. Cells were allowed ml) or Pam3Cys (InvivoGen) (250 ng/ml) for the times indicated in the figures. to ‘rest’ for an additional 24 h before LPS stimulation. Human monocytes were isolated from whole blood by counterfow centrifugal elutriation from PBMCs Measurement of macrophage viability. Macrophage viability was assessed using that were obtained from the blood of healthy human volunteers at the Department the Cell Titer Glo 2 reagent from Promega, according to the manufacturer’s of Transfusion Medicine, National Institutes of Health (NIH; provided by L. instructions. Wahl, National Institute of Dental and Craniofacial Research, NIH). Elutriated monocytes were plated for 2 h in serum-free medium (DMEM plus 1% penicillin- Immunoblot analysis. Whole-cell lysates from macrophage cultures were obtained streptomycin and 1% l-glutamine). Human AB-positive serum (2.5%; Gemini by the addition of lysis buffer (20 mM HEPES, pH 6.8, 1.0% Triton X-100, 0.1% Bio-Products) was added afer 2 h. Monocytes were allowed to ‘rest’ overnight SDS, 150 mM NaCl, 10 mM NaF and 1 mM PMSF) and subsequent incubation before stimulation with LPS (100 ng/ml) without or with EP4 antagonist (50 μ​M) at 4 °C. Cell lysates were separated by denaturing SDS-PAGE and subsequent (Tocris Bioscience). transfer to PVDF membranes. Blots were incubated overnight in relevant primary antibodies (identified above) at 4 °C and were washed three times with PBS, and Antibodies and reagents. Antibodies to cPLA2 (D49A7), phosphorylated then were incubated with the appropriate HRP-conjugated secondary antibody cPLA2 (D4I2A), IRF3 (D83B9), phosphorylated IRF3(4D4G), phosphorylated (Jackson HRP anti-rabbit (catalog number 111-035-003) or Jackson HRP anti- TBK1 (D52C2), phosphorylated JNK (81E11), phosphorylated Erk1/2 mouse (catalog number 115-035-003i); Jackson Immunochemicals). Blots were (D13.14.4E), p38 (D13E1), phosphorylated p38 (D3F9), Iκ​Bα​ (44D4 and developed following incubation in ECL PLUS Western Blotting Detection Reagent 9242), Syk (2712), Syk phosphorylated at Tyr525 and Tyr526 (C87C1), Akt (Amersham Bioscience). Immunoblots were quantitated using Image J software (11E7) and Akt phosphorylated at Ser473 (clone D9E) were obtained from (NIH), and the ratio of specific signal intensity to loading control for each lane is Cell Signaling. Antibody to caspase-1 p20 (clone Casper-1; catalog number provided directly under each blot. AG-20B-0042-C100) was obtained from Adipogen. Protein-free phenol- and water-extracted E. coli K235 LPS was prepared as described previously64. S-[2,3- TLR4-internalization assay. These experiments were carried out in sterile flow-

Bis(palmitoyloxy)-(2-RS)-propyl]-N-palmitoyl-(R)-Cys-Ser-Lys4-OH (Pam3Cys) cytometry tubes (BD). TEPMs were allowed to ‘rest’ overnight, pre-treated with and polyinosinic:polycytidylic acid (poly(I:C)) (VacciGrade) were purchased from medium only, BGC 20-1531, CAY10526 or synthetic PGE2 for 15 min. To stimulate InvivoGen. Purified adenylate cyclase (AC) toxin from Bordetella pertussis was a TLR4 internalization, cells were treated with LPS (100 ng/mL) at 37 °C, with 5% 7 gift from E. Hewlett (University of Virginia) and has been described previously38. CO2. Sample collection and staining were performed as previously described , with 68 The mPGES inhibitor CAY10526 was obtained from Cayman Chemical. PGE2, minor modifications . In brief, at 0, 30, 60 or 90 min after the addition of LPS, EP4 antagonist (BGC 20-1531), EP2 antagonist (PF-04418948), EP3 antagonist cells were rapidly cooled by the addition of two volumes of ice-cold FACS buffer (L-798,106), AKT1&2 inhibitor (AKTi-1/2), PKA inhibitor (H89 dihydrochloride) (0.5% FBS plus 2.0 mM EDTA in PBS), were centrifuged (1200 rpm) in a pre- 6 and EPAC inhibitor (ESI 09) were obtained from Tocris Bioscience. chilled centrifuge and were resuspended in ice-cold medium at 0.5 ml per 1 ×​ 10 cells. Samples were stored on ice until the collection of the final time point. All

ELISA. The ELISA used to detect PGE2 was purchased from ENZO and was subsequent steps were carried out on ice using ice-cold buffers and reagents. Cells used according to the manufacturers’ instructions. The mouse IFN-β ​ELISA was were washed twice in FACS buffer, then were treated with anti-CD16/32 (FcγRII-​ 6 performed as previously described35. blocking antibody (clone 93; catalog number 101302; BioLegend); 2.0 μ​g per 1 ×​ 10 cells) for 20 min and were stained with PE-conjugated anti–mouse CD284 (anti- 6 Bacterial strains and infection. Salmonella Typhimurium (ST) strain SL1344 was TLR4 (clone SA1521; catalog number 145403; BioLegend)); 0.4 μ​g per 1 ×​ 10 cells) a gift from R. Ernst. Enterohemorrhagic E. coli (EHEC), serotype O157:H7, strain or PE-conjugated rat IgG2a κ​-chain isotype-matched control antibody (clone RTK TUV93-0 (EDL933 ∆stx​ ), enteropathogenic E. coli (EPEC), serotype O127:H6, 2758; catalog number 400507; BioLegend) for 30 min in the dark. Cells were washed strain E2348/69, and Citrobacter rodentium strain ICC168 have been previously twice in FACS buffer, and results were ‘read’ within 30 min on a FACSCanto II Flow described65–67, and were provided by J. B. Kaper. In vitro infection of mouse Cytometer (BD) using FACSDiva software (BD) in the University of Maryland peritoneal macrophages with ST was carried out as described previously17. In brief, Flow Cytometry Core Facility, Center for Innovative Biomedical Resources, with 4 a single colony of ST strain SL1344 was inoculated into 5 ml LB medium and was 1 ×​ 10 single cell events collected per tube. To confirm the viability of cells after grown overnight at 37 °C with shaking. The following morning, an additional flow cytometry, a sample of the cells remaining after completion of flow cytometry 10 ml of LB medium was inoculated with 200 μ​l of the overnight culture and was was stained with 0.2% trypan blue solution (Sigma), and live and total cell counts incubated at 37 °C with shaking until the OD600 reached 1.5. Then, 1 ml of culture were determined by counting of unstained cells and all cells, respectively, using a was pelleted by centrifugation and re-suspended into 1 ml sterile PBS. A quantity standard hemocytometer (Reichert). Graphing and statistical analysis of FCS3 files amounting to 400 μ​l bacterial suspension was mixed with 600 μ​l sterile PBS to was carried out in FCS Express 6 Plus (DeNovo Software). Data are presented as the obtain a concentration of ~4 ×​ 106 bacteria per 10 ml culture. An appropriate average mean fluorescence intensity at each time point. volume of this re-suspension was mixed with 37 °C antibiotic-free RPMI 1640 to obtain the desired MOI. Bacteria were added to macrophages in culture plates, Caspase-1- and caspase-8-activation assays. To measure caspase-1 activation, 6 and infections were synchronized by centrifugation for 5 min at 700 r.p.m. Infected TEPMs were plated at 2 ×​ 10 per well in 12-well plates. Cells were then treated cultures were incubated at 37 °C for 30 min, and the infection medium was with vehicle or EP4 antagonist (50 μ​M) for 30 min, followed by LPS (100 ng/ml) removed and replaced with RPMI 1640 containing 50 μ​g/ml gentamicin, followed for 3 or 6 h. For positive control conditions, 5 mM ATP (Sigma Aldrich) was added by incubation for an additional 45 min at 37 °C to kill extracellular bacteria. to cultures for the final 30 min. Whole-cell lysates were probed with antibody Following gentamicin incubation, the medium was removed, the cultures were to caspase-1 (identified above). For caspase-8 activation, macrophages were washed twice with sterile PBS and the medium was replaced with antibiotic-free pre-treated with inhibitors for 30 min before stimulation with LPS for the times RPMI 1640. indicated in the figures. Whole-cell lysates were probed with antibody to caspase-8 For in vitro infection with EHEC, EPEC or C. rodentium, TEPM monolayers p55 (identified above; Enzo) or p18 (identified above; Cell Signaling). (9 ×​ 105 cells per well) were plated 24 h before infection. Bacteria were inoculated directly from glycerol stocks in 3 ml of L-broth medium and were grown overnight Quantitative real-time RT-PCR. Total mRNA was isolated from TEPMs or at 37 °C without shaking. Macrophage monolayers were infected with each of the BMDMs with TRIzol reagent (Invitrogen), according to the manufacturer’s strains in triplicate at an MOI of 25 (2.3 ×​ 107 bacteria per well) in 1 ml RPMI 1640 instructions. A total of 1 μ​g RNA was used in oligo(dT) cDNA synthesis (Bio-Rad medium containing 2% FBS. Control monolayers were given the addition of 1 ml RT system). Quantitative RT-PCR (qRT-PCR) was carried out using an ABI Prism

Nature Immunology | www.nature.com/natureimmunology Articles NATuRE IMMunoLoGy

7900HT Sequence Detection System (Applied Biosystems) using SYBR Green “homotolerance” versus “heterotolerance” on NF-κ​B signaling pathway reagent (Applied Biosystems) and transcript-specific primers. mRNA expression components. J. Immunol. 170, 508–519 (2003). values were normalized to those of the housekeeping gene Hprt in each sample, 60. Shirey, K. A. et al. Te anti-tumor agent, 5,6-dimethylxanthenone-4-acetic and the change in expression (fold values) was calculated by the 2−∆∆Ct method69. acid (DMXAA), induces IFN-β​-mediated antiviral activity in vitro and in vivo. J. Leukoc. Biol. 89, 351–357 (2011). Generation of microarray data and bioinformatics. Wild-type C57BL/6 J and 61. Facemire, C. S. et al. A major role for the EP4 receptor in regulation of renin. Ifnb–/– TEPMs were mock-infected or infected with ST (MOI =​ 4) for 6 h. Total Am. J. Physiol. Renal Physiol. 301, F1035–F1041 (2011). RNA was purified with a Roche HiPure kit according to the instructions. RNA was 62. Nguyen, M. et al. Te prostaglandin receptor EP4 triggers remodelling of the hybridized to Affymetrix GeneChip 2.0 mouse total transcriptome arrays at the cardiovascular system at birth. Nature 390, 78–81 (1997). University of Maryland, Baltimore, Center for innovative Biomedical Resources. 63. Richard, K., Vogel, S. N. & Perkins, D. J. Type I interferon licenses enhanced Genes displaying a difference in expression of greater than twofold on the Ifnb–/– innate recognition and transcriptional responses to Franciscella tularensis live background were used for Upstream Regulator Analytic in the Ingenuity Pathway vaccine strain. Innate Immun. 22, 363–372 (2016). Assist software (Promega). 64. McIntire, F. C., Sievert, H. W., Barlow, G. H., Finley, R. A. & Lee, A. Y. Chemical, physical, biological properties of a lipopolysaccharide from Statistics. All data were analyzed using GraphPad Prism (v.5.0) software. Data are Escherichia coli K-235. Biochemistry 6, 2363–2372 (1967). presented as arithmetic means with error bars, which reflect s.e.m., as indicated. 65. Barthold, S. W., Coleman, G. L., Bhatt, P. N., Osbaldiston, G. W. & Jonas, A. M. Where relevant, the sample size is indicated in the figure legends. Statistical Te etiology of transmissible murine colonic hyperplasia. Lab. Anim. Sci. 26, significance was determined using Student’s t test (two-tailed) except for Fig. 6e, 889–894 (1976). for which a Mantel-Cox test was used. A P value of less than 0.05 was considered 66. Donohue-Rolfe, A., Kondova, I., Oswald, S., Hutto, D. & Tzipori, S. statistically significant. Escherichia coli O157:H7 strains that express Shiga toxin (Stx) 2 alone are more neurotropic for gnotobiotic piglets than are isotypes Reporting Summary. Further information on research design is available in the producing only Stx1 or both Stx1 and Stx2. J. Infect. Dis. 181, Nature Research Reporting Summary linked to this article. 1825–1829 (2000). 67. Levine, M. M. et al. Escherichia coli strains that cause diarrhoea but do not Data availability produce heat-labile or heat-stable enterotoxins and are non-invasive. Lancet 1, The data that support the findings of this study are available from the 1119–1122 (1978). corresponding author upon request. Microarray data associated with Fig. 1 have 68. Rajaiah, R., Perkins, D. J., Ireland, D. D. & Vogel, S. N. CD14 dependence of been deposited win the GEO database with accession code GSE111826. TLR4 endocytosis and TRIF signaling displays ligand specifcity and is dissociable in endotoxin tolerance. Proc. Natl Acad. Sci. USA 112, 8391–8396 (2015). References 69. Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data –∆∆ 59. Dobrovolskaia, M. A. et al. Induction of in vitro reprogramming by Toll-like using real-time quantitative PCR and the 2( CT) method. Methods 25, receptor (TLR)2 and TLR4 agonists in murine macrophages: efects of TLR 402–408 (2001).

Nature Immunology | www.nature.com/natureimmunology nature research | reporting summary

Corresponding author(s): Stefanie Vogel and Darren Perkins

Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.

Statistical parameters When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main text, or Methods section). n/a Confirmed The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one- or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section. A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated

Clearly defined error bars State explicitly what error bars represent (e.g. SD, SE, CI)

Our web collection on statistics for biologists may be useful.

Software and code Policy information about availability of computer code Data collection SDS software v. 2.4 (ABI) was used to acquire and analyze Real time PCR data. FACS DIva software (BD) v.8.0 was used in TLR4 internalization assays.

Data analysis The Transcriptional Regulator Analytic within the Ingenuity Pathway Analysis (IPA) software (Qiagen) current release was used to analyze microarray data. Prism V. 5 software for Mac was used in statistical analyses. Image J (NIH) v.1.52 was used to quantify immunoblots. FCS express v.6 software (DeNovo) was used in analysis of TLR4 internalization data.

For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information. April 2018

1 Data nature research | reporting summary Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: - Accession codes, unique identifiers, or web links for publicly available datasets - A list of figures that have associated raw data - A description of any restrictions on data availability

Micro-array data associated with figure 1 has been deposited within GEO database #GSE111826. All primary data relating to remaining figures will be made available upon request.

Field-specific reporting Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection. Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf

Life sciences study design All studies must disclose on these points even when the disclosure is negative. Sample size No statistical method was used to predetermine sample size. Instead sample size was determined based on the numbers required to generate statistical power in previous similar experimental designs.

Data exclusions No data was excluded

Replication All experiments were replicated a minimum of three independent times. Attempts to replicate all data presented were succesful.

Randomization samples were randomly allocated into experimental groups.

Blinding Investigators were not blinded during data collection, as this was not technically feasible.

Reporting for specific materials, systems and methods

Materials & experimental systems Methods n/a Involved in the study n/a Involved in the study Unique biological materials ChIP-seq Antibodies Flow cytometry Eukaryotic cell lines MRI-based neuroimaging Palaeontology Animals and other organisms Human research participants

Antibodies Antibodies used Antibodies against cPLA2 (clone D49A7) cat no. 5249 , phospho-cPLA2 cat no 53044 (clone D4I2A), IRF3 (clone D83B9) cat no.4302 , phospho-IRF3(clone 4D4G) cat no. 4947, phospho-TBK-1 (clone D52C2)cat no. 5483 , phospho-JNK (clone 81E11) cat no.4668, phospho-ERK1/2 (clone D13.14.4E) cat no.4370, p38 (clone D13E1) cat no. 8690 , phospho-p38 (clone D3F9)cat no.4511, IκBα (clones 44D4 and 9242)cat no. 4812 and 5743 , Syk (clone D3z1E) cat no.13198 , phospho-Syk TYR525/526 (clone April 2018 C87C1) cat no. 2710, AKT (clone 11E7)cat no.4685, and phopho-AKT Ser473 (clone D9E)cat no. 4060, and Caspase 8 p18 (clone D5B2) cat no. 14071, were obtained from Cell Signaling (Danvers, MA). Anti-Caspase-1 p20 was obtained from Adipogen (cat no. AG-20B-0042). Anti- caspase 8 p55 (clone 1G12) cat no. ALX-804-447-C100 was from ENZO.

Validation Each antibody produced by Cell Signaling is independently verified for target specificity and lot to lot variation. Manufacturers statement (https://www.cellsignal.com/contents/our-approach/cst-antibody-validation-principles/ourapproach-validation-

2 principles). Caspase 1 antibody has been verified for specificity by manufacturer using knockout cell lines https://adipogen.com/

ag-20b-0042-anti-caspase-1-p20-mouse-mab-casper-1.html. Caspase-8 p55 antibody has been validated by manufacturer in nature research | reporting summary wild-type and caspase 8 knockout cell lines http://www.enzolifesciences.com/ALX-804-447/caspase-8-mouse-monoclonal- antibody-1g12/

Eukaryotic cell lines Policy information about cell lines Cell line source(s) The human THP-1 cell line was obtained from directly from the American Type Culture Collection

Authentication Cell line was not authenticated

Mycoplasma contamination Cell line tested negative for Mycoplasma

Commonly misidentified lines No commonly misidentified cell lines were used (See ICLAC register)

Animals and other organisms Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research Laboratory animals All animals used were murine C57BL/6J background males and females between 6 and 8 weeks of age, with exception of EP4 null animals which are maintained on a recombinant inbred (RI) background and were a mix of males and females between 6 and 7 weeks of age.

Wild animals N/A

Field-collected samples N/A April 2018

3