The Prostaglandin E2-EP3 Receptor Axis Regulates Anaplasma phagocytophilum-Mediated NLRC4 Inflammasome Activation

Rickettsial agents are sensed by pattern recognition receptors but lack pathogen-associated molecular patterns commonly observed in facultative intracellular bacteria. Due to these molecular features, the order Rickettsiales can be used to uncover broader principles of bacterial immunity. Here, we used the bacterium Anaplasma phagocytophilum, the agent of human granulocytic anaplasmosis, to reveal a novel microbial surveillance system. Mechanistically, we discovered that upon A. phagocytophilum infection, cytosolic phospholipase A2 cleaves arachidonic acid from phospholipids, which is converted to the eicosanoid prostaglandin E2 (PGE2) via cyclooxygenase 2 (COX2) and the membrane associated prostaglandin E synthase-1 (mPGES-1). PGE2-EP3 receptor signaling leads to activation of the NLRC4 inflammasome and secretion of interleukin (IL)-1β and IL-18. Importantly, the receptor-interacting serine/threonine-protein kinase 2 (RIPK2) was identified as a major regulator of the immune response against A. phagocytophilum. Accordingly, mice lacking COX2 were more susceptible to A. phagocytophilum, had a defect in IL-18 secretion and exhibited splenomegaly and damage to the splenic architecture. Remarkably, Salmonella-induced NLRC4 inflammasome activation was not affected by either chemical inhibition or genetic ablation of genes associated with PGE2 biosynthesis and signaling. This divergence in immune circuitry was due to reduced levels of the PGE2-EP3 receptor during Salmonella infection when compared to A. phagocytophilum. Collectively, we reveal the existence of a functionally distinct NLRC4 inflammasome illustrated by the rickettsial agent A. phagocytophilum.


Introduction
Rickettsial diseases are arthropod-borne illnesses caused by obligate intracellular bacteria grouped in the order Rickettsiales [1,2]. They include: (i) rickettsioses due to bacteria of the genus Rickettsia, including the spotted fever and the typhus group; (ii) scrub typhus due to Orientia tsutsugamushi; and (iii) ehrlichioses and anaplasmosis due to bacteria within the family Anaplasmataceae [1,2]. Some aspects of rickettsial recognition by the immune system have been described [1,2]. For instance, Rickettsia spp. have a structurally distinct form of lipopolysaccharide (LPS) that appears identifiable by Toll-like receptor (TLR)4 [2][3][4][5], whereas the TLR2-MyD88 (Myeloid Differentiation Primary Response Protein 88) axis plays a critical role in host defense against ehrlichial infection [6,7]. However, how these organisms are sensed by pattern recognition receptors (PRRs) remains mostly undefined. Bona fide pathogen-associated molecular patterns (PAMPs) are conspicuously absent in some of these microbes when compared to classically-defined bacterial pathogens [2,[8][9][10]. As an example, Anaplasma and Ehrlichia spp. are considered Gram-negative bacteria, but are unable to synthesize LPS or peptidoglycans [8,9,11]. Additionally, O. tsutsugamushi does not carry genes in its genome for producing lipid A and has no LPS [10,12].
These findings suggest that the life style of rickettsial agents induces a mode of immune recognition, which can be exploited for the discovery of unique pathogen-sensing systems. Previously, we discovered that mice deficient in Nlrc4 and Caspase-1/11 are susceptible to A. phagocytophilum infection [25]. We also reported that A. phagocytophilum causes NLRC4 inflammasome activation and caspase-1 autoproteolysis through the phospholipid-binding protein Annexin A2 [27,28]. The mechanistic delineation of how the NLRC4 inflammasome was induced remained elusive. In this article, we show a novel mode of NLRC4 inflammasome circuitry that is dependent on the eicosanoid prostaglandin E 2 (PGE 2 ). Upon A. phagocytophilum infection, cytosolic phospholipase A 2 (cPLA 2 ) cleaves arachidonic acid from phospholipids, which is converted to PGE 2 via cyclooxygenase 2 (COX2) and membrane associated prostaglandin E synthase-1 (mPGES-1), the terminal enzyme that catalyzes the isomerization of prostaglandin H 2 (PGH 2 ) to PGE 2 [29,30]. PGE 2 -EP3 receptor signaling then leads to NLRC4 inflammasome assembly, which induces the release of IL-1β and IL-18. Consistent with our previous reports where mice deficient in RIPK2 are susceptible to A. phagocytophilum infection [14], we identified RIPK2 as a major regulator of the innate immune response against A. phagocytophilum. Ripk2 -/immune cells exhibited a defect in activation for the nuclear factor (NF)-κB and the NLRC4 inflammasome pathways. Altogether, we define the existence of a functionally distinct NLRC4 inflammasome upon microbial infection.

A. phagocytophilum infection stimulates eicosanoid biosynthesis
A. phagocytophilum transiently infects bone-marrow derived macrophages (BMDMs) [27,28] and clinical features in animal models and infected patients suggest classical macrophage activation [31][32][33][34]. To determine which genes are important for host immunity, we infected macrophages with A. phagocytophilum. Deep sequencing analysis [deposited at the Gene Expression Omnibus database (GSE63647)] indicated that the transcription of genes that encode for phospholipase A 2 (pla2g12a, pla2g5 and pla2g2e), COX2 (ptgs2) and PGE synthase (ptges) was increased upon A. phagocytophilum infection (Fig 1A). These genes are critical for prostanoid biosynthesis (Fig 1B) [35] and correlated with elevated enzymatic activities of cytosolic phospholipase A 2 (cPLA 2 ), COX1 and COX2 (Fig 1C-1E), which led to increased levels of arachidonic acid (AA), PGE 2 , prostaglandin D 2 (PGD 2 ) and thromboxane A 2 (TBXA 2 ) ( Fig  1F-1I) upon A. phagocytophilum infection. cPLA 2 promotes activation of the A. phagocytophilum-induced NLRC4 inflammasome Eicosanoids have been associated with NLRC4 inflammasome activation [36] and phospholipase A2 releases arachidonic acid from phospholipids for eicosanoid biosynthesis (Fig 1B) [35]. Therefore, we examined whether cPLA 2 was regulating the A. phagocytophilum-induced NLRC4 inflammasome. Pharmacological inhibition of cPLA 2 , but not other phospholipases [e.g., soluble phospholipase A 2 (sPLA 2 ), phospholipase C (PLC) and phospholipase D (PLD)] reduced the levels of PGE 2 , PGD 2 and TBXA 2 upon A. phagocytophilum infection of macrophages (Fig 2A-2C). We also observed lower levels of IL-1β, IL-18 and caspase-1 activation upon bacterial stimulation of immune cells (Fig 2D, 2E and 2G). Similar results were obtained with macrophages deficient in cPLA 2 at low and high A. phagocytophilum multiplicity of infection (MOI) (Fig 3A-3F and 3H), indicating that pharmacological inhibition of cPLA 2 does not lead to off-target effects and the results obtained occurred independently of bacterial numbers. Importantly, secretion of IL-6 and translation of IL-1β and IL-18 by macrophages, which are not regulated by the inflammasome, remained unaffected during pre-treatment of macrophages with pharmacological inhibitors or in the absence of cPLA 2 (Fig 2F and 2G and Fig 3G and 3H). Surprisingly, chemical inhibition or genetic ablation of cPLA 2 did not affect caspase-1 autoproteolysis and cytokine secretion when macrophages were infected with Salmonella (S1 Fig), a pathogen that stimulates the NLRC4 inflammasome through the T3SS and flagellin [18][19][20][21][22][23][24]. Altogether, these results revealed that although both A. phagocytophilum and Salmonella trigger formation of the NLRC4 inflammasome, the signaling cascades that enable its activation appeared fundamentally different.

RIPK2 elicits NLRC4 inflammasome activity during A. phagocytophilum infection
Next, we performed a kinetics experiment in macrophages to better characterize A. phagocytophilum infection in the context of NLRC4 inflammasome biology. As previously shown, A. phagocytophilum was undetectable inside macrophages at 2-hours post-infection [27]. A small number of bacteria was observed at 6 hours, followed by an increased load at 18 hours and reduction at 48 hours, which led to almost complete elimination after 72 hours of infection in macrophages ( Fig 8A) [27]. Consistently, PGE 2 secretion, caspase-1 activation and IL-1β and IL-18 secretion but not IL-6, peaked at 18 hours, the same time point where the greatest number of A. phagocytophilum was detected inside macrophages ( Fig 8A-8D and S4A Fig).
A. phagocytophilum does not synthesize LPS or peptidoglycans [8,9,11]. Therefore, one interesting immunological question pertains to the host molecule that induces NF-κB activation upon infection. We reasoned that RIPK2 could be this master regulator. This hypothesis rested on four findings. First, RIPK2 activates NF-κB signaling and mitogen activated protein (MAP) kinases upon infection [13]. Second, A. phagocytophilum interacts with the host endoplasmic reticulum (ER) [39], which may exert RIPK2 activity in the absence of peptidoglycans due to cellular stress [40]. Third, COX2 expression is regulated through a signaling cascade that converges at the MAP kinase and the NF-κB pathways [41]. Fourth, mice deficient in RIPK2 are susceptible to A. phagocytophilum infection and secrete reduced levels of IL-18 in the peripheral blood [14]. Accordingly, ripk2 -/macrophages exhibited a defect in NF-κB and MAP kinase signaling, which led to decreased translation of COX2, pro-IL-1β and IL-6 secretion (Fig 8G and 8H). RIPK2 activity also affected PGE 2 release and caspase-1 autoproteolysis upon A. phagocytophilum infection, as indicated by reduced levels of PGE 2, IL-1β, IL-18 and caspase-1 activation in cell culture supernatants of ripk2 -/macrophages (  Collectively, we identified RIPK2 as a major regulator of the innate immune response against A. phagocytophilum.

The PGE 2 -EP3 receptor regulates activation of the NLRC4 inflammasome upon A. phagocytophilum infection
We then blunted the PGE 2 signaling cascade with chemical antagonists that bind covalently to the four PGE 2 receptor subtypes (EP1-EP4) [30] and compared our findings with Salmonella. We observed that inhibition of the PGE 2 -EP3 receptor significantly decreased IL-1β and IL-18 release, and caspase-1 activation, but not IL-6 secretion upon A. phagocytophilum infection (Fig 9C-9E and S5 Fig). The EP3 receptor for PGE 2 is sensitive to pertussis toxin (PT) [42]. Macrophages pre-treated with PT and then stimulated with A. phagocytophilum also resulted in inhibition of the NLRC4 inflammasome (S5A- S5D Fig). Importantly, the catalytically inactive pertussis toxin (PT Ã ), with a two amino acid substitution (9K129G) [43], did not block PGE 2 signaling upon A. phagocytophilum colonization (S5A- S5D Fig). Next, we took advantage of the ep3 -/mice and showed that in the absence of the EP3 receptor molecule, A. phagocytophilum did not induce caspase-1 activation and IL-1β and IL-18 secretion by macrophages (Fig 9F-9I). Conversely, lack of the PGE 2 -EP3 receptor did not affect the NLRC4 inflammasome induced by Salmonella (Fig 9J-9M and S6 Fig).
PGE 2 exerts its actions by acting on G-protein-coupled receptors (GPCRs). PGE 2 binds to the EP3 receptor, which inhibits the membrane associated adenylyl cyclase via Gαi [44]. This signaling relay decreases cytosolic cyclic AMP (cAMP) production, as adenylyl cyclase catalyzes the conversion of adenosine triphosphate (ATP) to cAMP [44] (S8 Fig). We validated Next, we showed that membrane, but not soluble, adenylyl cyclase modulated the A. phagocytophilum-induced NLRC4 inflammasome. Forskolin, a selective inhibitor of the membraneassociated adenylyl cyclase [45], inhibited IL-1β, IL-18 and caspase-1 autoproteolysis during A. phagocytophilum infection of macrophages (S7B-S7E Fig). On the other hand, pre-treatment of macrophages with KH7, a specific pharmacological inhibitor of soluble adenylyl cyclase [45], did not affect NLRC4 inflammasome function during A. phagocytophilum infection (S7F-S7I Fig). Altogether, these findings: (i) indicated that the PGE 2 -EP3 axis is critical for the NLRC4 inflammasome elicited by A. phagocytophilum; and (ii) explained why Salmonella is unable to trigger a similar pathway when compared to A. phagocytophilum. This was likely due to reduced expression of the EP3 receptor during Salmonella infection of macrophages (Fig 9A  and 9B).

Mice deficient in COX2 are susceptible to A. phagocytophilum
To prove that the results obtained in vitro could also be observed in vivo, we then infected mice deficient in COX2 (Ptgs2) with A. phagocytophilum. Ptgs2-deficient animals were more susceptible to A. phagocytophilum infection (Fig 10A) and exhibited reduced levels of IL-18 in the peripheral blood when compared to the wildtype mice ( Fig 10B). As previously seen, no detectable levels of IL-1β were observed in the blood of A. phagocytophilum-infected mice [25]. These findings agreed with our prior publications, showing that IL-18 release mediated by RIPK2 and the NLRC4 inflammasome regulates interferon (IFN)-γ production by CD4 + T cells upon A. phagocytophilum infection [14,25]. COX2 (Ptgs2)-deficient mice infected with A. phagocytophilum also revealed lower levels of PGE 2 , PGD 2 , TBXA 2 and splenomegaly ( Fig  10C-10G). COX2 (Ptgs2)-deficient animals had increased cellular infiltration in the red pulp and damage to the splenic architecture upon A. phagocytophilum infection (Fig 10H). In sum, these results showed that COX2 is critically important for A. phagocytophilum infection in vivo.
We provide unequivocal evidence that two distinct signaling pathways occur for NLRC4 inflammasome activation within the cell: one termed classical (i.e., stimulated by Salmonella) and another referred to as alternative (i.e. described here, responding to A. phagocytophilum). Given how inflammasome biology intersects with a growing number of disciplines, we reason that these findings are conceptually valuable because we reveal that eicosanoid receptors in immune cells activate diverging signaling cascades. For instance, both A. phagocytophilum and Salmonella lead to PGE 2 production by macrophages. However, Salmonella is unable to activate the eicosanoid-dependent NLRC4 inflammasome pathway because it does not induce PGE 2 -EP3 receptor expression. PGE 2 is likely acting in an autocrine/paracrine manner to drive NLRC4 inflammasome activation upon A. phagocytophilum infection. This is based on the evidence that A. phagocytophilum infection upregulates the EP3 receptor, which is known to elicit PGE 2 signaling in a cellintrinsic manner [30,35]. Alternatively, PGE 2 may also affect the function of "bystander" cells in a paracrine manner given that our exogenous PGE 2 "add-back" assays restored NLRC4 inflammasome activity in A. phagocytophilum-infected cells.
Can rickettsial agents be used to uncover broader principles of immune sensing? The answer to this question may have to deal with the biology of these organisms. Rickettsial agents differ greatly in terms of how they invade and replicate within the mammalian host when compared to other bacteria commonly used to study microbial immunity. Their obligate intracellular life style, coupled to the intense selective pressure to survive both in the arthropod vector and the mammalian host [1,2] suggests that these microbes have to employ extreme measures to conceal themselves from the immune system. This reasoning may explain why A. phagocytophilum triggers such a distinct pathogen-recognition mechanism when compared to other bacteria.
In summary, we discovered a novel mode of NLRC4 inflammasome activation triggered by the rickettsial bacterium A. phagocytophilum. We revealed that some microbial pathogens lacking the T3SS and flagellin activate the NLRC4 inflammasome. We also illustrated how this protein scaffold distinguishes bacterial infection within the cell. Altogether, our findings suggest that there are broader yet-to-be discovered principles of microbial sensing in the context of NLRC4 inflammasome biology.

Mice and bacteria
Breeding and experiments were performed in strict compliance with guidelines set forth by the National Institutes of Health (Office of Laboratory Animal Welfare [OLAW] assurance number A3200-01). Procedures were approved by the Institutional Biosafety (IBC:00002247) and Animal Care and Use (IACUC:0413017 and 0216015) committees at the University of Maryland, Baltimore. Ripk2 -/-(007017), C57BL/6 (000664) and Ptgs2 -/-(COX2) mice (008101) were purchased from Jackson Laboratories. Femurs from mPGES1 -/- [29] and Ep3 -/- [46] mice were a gift from Leslie Crofford and Richard Breyer at Vanderbilt University School of Medicine.
Tlr4 -/and cPla2 -/mice were previously described [47,48]. Mice were gender matched and at least 6-10 weeks of age. BMDMs were generated, as previously described [27]. Culturing for the A. phagocytophilum strain HZ and calculations were described elsewhere [27]. Salmonella strain SL1344 was a gift from Dr. Stefanie Vogel at the University of Maryland, Baltimore School of Medicine. Salmonella was grown in HS media at 37°C and enumerated, as previously described [49]. Cell cultures were tested and determined to be Mycoplasma-negative through a commercially available PCR kit (Southern Biotech -13100-01).

Bacterial infection of macrophages
1×10 6 BMDMs were seeded into 24-well plate in 300 μl of media containing 5% fetal bovine serum (FBS) overnight prior to the challenge by either A. phagocytophilum (MOI 10 and 50) or Salmonella (MOI 25) for 1 hour. 50ng/ml of LPS was used for cell priming at 37°C and 5% CO 2 for 30 minutes during Salmonella infection. LPS-primed cells were washed twice extensively followed by the addition of bacteria. In inhibition assays, 1×10 6 WT and genotype-deficient BMDMs were pre-treated with pharmacological inhibitors at indicated time and concentrations followed by the stimulation with A. phagocytophilum (MOI 10 and 50) overnight or Salmonella (MOI 25) for 1 hour. For the Ptgs2 -/and mPGES1 -/-"add-back" experiments, 1×10 6 WT and deficient cells were infected with A. phagocytophilum (MOI 50) for 4 hours followed by the addition of the respective eicosanoid at indicated concentrations for 18 hours. After infection, cultured supernatants and cell lysates collected from each well were used for ELISA and immunoblot assays.
Enzymatic assays 15×10 6 wildtype cells were stimulated with A. phagocytophilum (MOI 25) overnight. Cells were scraped followed by sonication. COX1/2 enzymatic assays were performed with COX activity assay kit (Cayman Chemicals), whereas cPLA 2 activity was measured following instructions by the manufacturer (Abnova). Arachidonic acid levels were measured according to the instructions of the ELISA kit (MyBiosource). cAMP was measured by using the cyclic AMP XP Assay Kit (Cell Signaling Technology).

Illumina sequencing and bioinformatics
BMDMs were grown into 6-well culture plates at 7×10 6 per well. Cells were stimulated with A. phagocytophilum. Uninfected BMDMs were used as controls and the experiment was performed in triplicate. Total RNA was isolated with the PureLink RNA Mini Kit (Invitrogen). Illumina Sequencing was performed at the University of Maryland, Baltimore. Briefly, Illumina RNAseq libraries were prepared with the TruSeq RNA Sample Prep kit (Illumina, San Diego, CA). The indexed libraries were pooled and sequenced using the HiSeq platform (Illumina) for the mouse samples in order to generate 101 base pair reads. The reads were further trimmed due to low quality at the trailing 3' end. These trimmed paired end reads were populated into 2 separate FASTQ format files and the quality of the reads was tested using the FastQC toolkit to ensure quality of the sequencing reads.
The RNA sequencing reads were used as input for the TopHat read alignment tool to be aligned to the mouse genomic reference sequence (Ensembl GRCm38 version) for each of the samples. The reference genomic sequences for the GRCm38 genome build were downloaded from the Ensembl resources. The output from TopHat was obtained as BAM format files. In the alignment phase, we allowed up to two mismatches per 30 base pair segment and removed reads that aligned to more than 20 genomic locations. The BAM alignment files obtained from the TopHat alignment tool was analyzed to generate the alignment statistics for each sample, namely, the total number of reads, the number of mapped reads and the percent of mapped reads.
For the differential gene expression analysis, the alignment BAM files from TopHat were further utilized to compute gene expression levels and test each gene for differential expression. The mouse gene set reference annotation (version GRCm38) in GTF format was downloaded from the Ensembl resources. The number of reads that mapped to each gene described in the Ensembl annotation was calculated using the python package HTSeq-an alignment read count tool. The read count represented the expression of the gene. Differential gene expression analysis was conducted using the DESeq R package (available from Bioconductor). The DESeq analysis resulted in the determination of differentially expressed genes. DESeq utilized the read counts provided by the HTSeq read count tool. The read counts for each sample were normalized for sequencing depth and distortion caused by highly differentially expressed genes. The negative binomial model was used to test the significance of differential expression between two genotypes. The differentially expressed genes were deemed significant if the FDR (False Discovery Rate) was less than 0.01, the gene expression was above the 45th percentile and gene showed greater than 2-fold change difference (over expressed or under expressed) between conditions. Principal component analysis and other clustering methods were used to visualize the clustering of the replicates across samples. Heat maps were generated to illustrate the genes showing significant differences between multiple comparisons of the control and other infection and/or treatment conditions.
In vivo infection C57BL/6 (n = 20) and COX2 (Ptgs2) -/-(n = 10) mice were infected by intraperitoneal injection with A. phagocytophilum strain HZ (1×10 7 cells). Blood samples were collected at days 0, 5 and 10 for the IL-18 ELISA. Spleens were removed, normalized to the body weight, and compared to those of non-infected mice. Spleens were fixed at day 15 post-infection with 10% neutral buffered formalin and embedded in paraffin wax. Sections (5 μm) were obtained and stained with hematoxylin and eosin. Measurement of A. phagocytophilum load was done at day 15 post-infection in the peripheral blood of infected animals using quantitative RT-PCR, as described above.