Prostaglandin E₂ Inhibits the Ability of Neutrophils to Kill Listeria monocytogenes

Inflammatory stimuli can trigger the production of PGE₂, a process that begins with the liberation of arachidonic acid from mammalian cell membranes (1). During Listeria monocytogenes infection, the pore-forming toxin listeriolysin O activates the group IVA cytosolic phospholipase A2, causing an increase in the release of arachidonic acid by macrophages (2). Arachidonic acid is oxidized by one of the cyclooxygenase enzymes, COX-1 or COX-2, to form an intermediate eicosanoid that is further modified by PGE synthase to produce PGE₂. COX-1 and COX-2 vary in their tissue expression patterns, but exposure to microbe-associated molecular patterns greatly increases expression of COX-2 (3–6). Free arachidonic acid can also be metabolized by the lipoxygenase pathway, ultimately resulting in the production of other eicosanoids such as leukotriene B4 (LTB4), a lipid with chemotactic properties for polymorphonuclear neutrophils (PMN) (7, 8).

PGE₂ has long been studied as a proinflammatory mediator; however, it is becoming increasingly clear that PGE₂ can also have anti-inflammatory effects in certain contexts. More than 30 y ago, Hutchison and Myers (9) first identified PGE₂ as a factor in splenocyte culture supernatants that could decrease the phagocytic activity of macrophages. More recently, PGE₂ was shown to inhibit the production of reactive oxygen species (ROS) and the secretion of both IL-12 and TNF-α from monocytes (10–12). Secreted PGE₂ can bind to one of four different G-protein–coupled receptors (EP1, EP2, EP3, and EP4), which differ in their affinity for PGE₂, tissue expression patterns, and downstream signaling cascades activated (13, 14). Ligation of either EP2 or EP4 stimulates an increase in cAMP, and it is these two receptors that are thought to promote many of the immunosuppressive effects of PGE₂ (11, 15–17).

The results of previous studies using various types of myeloid-derived phagocytes suggest several nonredundant, PGE₂-dependent mechanisms that could lead to impaired killing by PMN during bacterial infection. For example, slower migration toward infectious foci or swarming toward chemotactic signals produced by other PMN could decrease the rate of bacterial killing. In support of this idea, chemotaxis of human PMN toward formylated Met-Leu-Phe peptide was strongly inhibited by either application of PGE₂ or direct activation of adenylate cyclase using forskolin (18, 19). Grainger et al. (10) recently showed that PGE₂ produced during murine Toxoplasma gondii infection suppressed neutrophilic inflammation by decreasing ROS production and TNF-α secretion. Finally, PGE₂-mediated inhibition of phagocytosis can decrease uptake of a variety of microbes including L. monocytogenes, Staphylococcus aureus, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Histoplasma capsulatum (9, 17, 20, 21). However, these studies were performed primarily in macrophages, and it is not yet clear if PGE₂ has a similar effect on phagocytosis in PMN.

In this study, we used the murine model of foodborne listeriosis to test whether the anti-inflammatory effects of PGE₂ would cause decreased killing of L. monocytogenes by PMN. We focused on PGE₂ production in the liver and its effects on PMN recruited to hepatic tissue because PMN are thought to play an important role in clearance of L. monocytogenes in the liver (22). Highly susceptible BALB/cByJ mice were compared C57BL/6J mice, which are more innately resistant to L. monocytogenes infection (23, 24). We previously showed that PMN harvested from the bone marrow of naive BALB/cByJ and C57BL/6J mice displayed no intrinsic difference in the ability to kill L. monocytogenes in vitro (25). However, strain differences in PGE₂ production have been reported, with splenic and peritoneal macrophages from BALB/c mice producing more PGE₂ after LPS stimulation (26). Thus, the hypothesis tested in this study was that increased production of PGE₂ in BALB/cByJ mice would create a local microenvironment that inhibited the killing capacity of PMN infiltrating the liver, contributing to the innate susceptibility of BALB/cByJ mice to infection with L. monocytogenes.

BALB/cByJ (BALB; stock no. 001026), C57BL/6J (B6; stock no. 000664), and B6.129S-Cybb^tm1Din/J (gp91^phox−/−; stock no. 002365) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in a specific pathogen-free facility. Both male and female animals were used, with the exception of the data shown in Fig. 4, for which male gp91^phox−/− mice were compared with male B6 mice. Blood was collected from the aorta immediately after euthanasia by cervical dislocation and transferred to a serum separator tube (BD Biosciences), followed by centrifugation for 2 min at 20,000 × g. Serum was stored at 4°C for use the same day or at −80°C for later use. All procedures were approved by the Institutional Animal Care and Use Committee at the University of Kentucky.

Bone marrow was harvested by flushing the femurs and tibias of one mouse with a total of 20 ml of RP-10 medium (RPMI 1640 [no. 21870; Life Technologies] supplemented with 10% FBS [Gemini], 2.5 mM l-glutamine, 10 mM HEPES, 0.1 mM β2-ME, 100 U/ml penicillin, and 0.1 mg/ml streptomycin [Life Technologies]). RBCs were lysed in ammonium chloride buffer, and the remaining cells were washed with PBS (no. 14190; Life Technologies). Cells were plated at a density of 1 × 10⁶ cells/ml in 16 ml of RP-10 in ultra-low attachment, 100-mm dishes (No. 3262; Corning), and rM-CSF (Miltenyi) was added at a final concentration of 20 ng/ml. Cells were harvested on day 4 or 5, washed with RP-10 medium lacking antibiotics [RP-10(−)], and plated in low-adherence, 24-well plates (no. 3473; Corning) at 2.5 × 10⁵ cells/ml.

Mouse-adapted (InlA^m) L. monocytogenes EGDe derivatives SD2000, SD2001 (kanamycin-resistant), and SD2710 (constitutive GFP) were used in these studies (24, 27). Expression of the modified InlA^m surface protein allows the bacteria to efficiently bind murine E-cadherin, promoting invasion of the intestinal epithelium (28). All strains were cultured in brain heart infusion (BHI) broth shaking at 37°C (for in vitro assays) or 30°C (for oral infection of mice). Isopropyl-β-D-thiogalactopyranoside (final concentration 1 mM) and kanamycin (50 μg/ml) were added to BHI agar (Difco Laboratories) for selection of L. monocytogenes SD2001.

Cells were suspended in a buffer consisting of Ca²⁺/Mg²⁺-free HBSS (no. 14175; Life Technologies) 25 mM HEPES, 5 mM EDTA, and 1% FBS; pretreated with Fc block (rat anti-mouse CD16/CD32; BD Biosciences) and then stained using fluorescently conjugated Abs specific for the following molecules: CD64 (clone X-54-517.1), Ly6C (HK1.4), and Ly6G (1A-8) purchased from BioLegend; F4/80 (BM8), GR1 (RB6-8C5), B220 (RA3-6B2), CD3 (145-2C11), CD11c (HL3), and CD11b (M1/70) purchased from eBioscience. FITC-conjugated anti–COX-2 Ab (Cayman Chemical) was used to detect intracellular COX-2 in cells treated with fixation and permeabilization buffer (BD Biosciences). PMN that were stained for sorting and used in a postsort killing assay were not treated with Fc block prior to staining; specificity was confirmed by back gating.

Fluorescence was measured using either a LSR II (BD Biosciences) or Synergy (iCyt) cytometer, and analysis was performed using FlowJo v.10 (Tree Star). Macrophages were defined as CD11c^lo/int and F4/80^hi or CD64^hi; B cells as CD3⁻B220⁺ or Ly6G⁻Ly6C⁻B220⁺; T cells as CD3⁺B220⁻; PMN as Ly6G^hiLy6C^int; Kupffer cells (KC) as Ly6G⁻Ly6C⁻CD64⁺CD11b⁻; and monocytes as Ly6G⁻Ly6C^hi. PMN were sorted into flow cytometry buffer supplemented with 50% FBS and 25 μg/ml gentamicin and washed twice with RP-10(−) prior to use. The sort gating strategy incorporated live cell and singlet gates prior to gating on individual markers; sort purities for PMN ranged from 97.5 to 98.0% for cells harvested from bone marrow and 92.0–95.7% for liver cells.

For in vivo infection, intestinally-passaged L. monocytogenes were grown to early stationary phase in BHI broth (Difco Laboratories) shaking at 30°C, and aliquots were prepared and frozen at −80°C until use as described previously (29). Aliquots were thawed on ice and incubated for 1.5 h in BHI broth standing at 30°C. L. monocytogenes were washed twice with sterile PBS, and the desired dose was suspended in 5 μl of a salted butter (Kroger) plus PBS mixture (3:2 ratio) and then added to a 2–3-mm piece of white bread (Kroger). Mice (6–12 wk old at time of infection) were fasted overnight prior to ingestion of the L. monocytogenes–contaminated bread and were housed on raised wire flooring as described previously (30, 31). Mock-infected mice were fasted and fed an uncontaminated piece of bread.

Bone marrow was harvested from animals 8–16 wk old; no differences in PMN yield per gram of body weight were noted for male and female mice. Marrow was flushed from femurs and tibias with RP10(−) supplemented with 2 mM EDTA and 25 μg/ml gentamicin. Cells were passed through a sterile mesh filter into a 15-ml polyethylene terephthalate tube (two bones per tube; Corning) and pelleted by centrifugation at 400 × g. RBCs were lysed by exposure to 0.2% NaCl for 20 s, followed by addition of an equal volume 1.6% NaCl. Following RBC lysis, PMN were enriched as described previously (25, 32). After gradient centrifugation, the bottom layer of cells was recovered and suspended in either RP-10(−) or flow cytometry buffer.

PGE₂ was measured using the PGE₂ Metabolite (PGEM) Kit (no. 514531; Cayman Chemical), which converts all the metabolites of PGE₂ present in a sample into a single stable derivative prior to measurement. For Fig. 2, liver hematopoietic cells were gradient enriched and cultured overnight at a density of 2.5 × 10⁵ cells per well in a final volume of 0.5 ml in 24-well, low-adherence dishes (Corning). Samples were pooled (total of 4 × 10⁶ cells), and proteins were precipitated using four volumes of ice-cold acetone, followed by incubation at 20°C for at least 2 h. For Figs. 5 and 6, 4 × 10⁶ liver nonparenchymal cells underwent acetone precipitation immediately following gradient enrichment. Samples were dried under nitrogen, and lipids were extracted using ethyl acetate according to the manufacturer’s instructions.

A freshly streaked colony of each L. monocytogenes strain was inoculated into BHI broth and grown shaking at 37°C overnight (16 h). Stationary phase bacteria were washed once with Ca²⁺/Mg²⁺-free HBSS and then pretreated with autologous naive mouse serum (either collected from the same mouse or thawed from −80°C; final concentration of 10% for 30 min at 37°C). PMN were plated at 1 × 10⁵ cells per well in 96-well, tissue culture–treated, flat-bottom plates. Opsonized bacteria were added at the indicated multiplicity of infection (MOI); the plate was centrifuged for 5 min at 300 × g and then incubated at 37°C in 7% CO₂ for 50 min. The percentage of L. monocytogenes killed was determined as described previously (25). Briefly, serial dilutions were prepared in PBS and plated on BHI agar. For each sample, the mean number of CFU for triplicate wells containing PMN was divided by the mean number of CFU observed for triplicate wells of opsonized bacteria incubated without PMN.

Cells were treated with PGE₂ (no. 14010; Cayman Chemical) suspended in ethanol and diluted in RP10(−) for the indicated time periods prior to addition of bacteria and further assessment of functionality. For antagonist experiments, cells were treated concurrently with 1 μM PGE₂ and with EP2 antagonist PF-04418948 (no. 15016, Cayman Chemical) or EP4-antagonist L-161,982 (no. 10011565; Cayman Chemical) suspended in DMSO and diluted in RP10(−) at a final concentration of 100 nM.

Neutrophil chemotaxis was assessed by under-agarose assay (33). Ultra-pure agarose (1%; Invitrogen) was dissolved in Ca²⁺/Mg²⁺-free HBSS and diluted with phenol red-free RPMI 1640 (no. 32404; Life Technologies) containing a final concentration of 2.5% FBS. FBS-coated, 60-mm petri dishes were filled with 4.5 ml of the agarose mixture, and 3-mm holes were cored 1 mm apart in sets of three using a sterilized hollow punch tool with a silicone template as a guide (Supplemental Fig. 1A). LTB4 (100 nm; Cayman Chemical) diluted in Ca²⁺/Mg²⁺-free HBSS was applied to the center well and allowed to diffuse into the gel for 15 min prior to addition of gradient-enriched vehicle or PGE₂ pretreated PMN (2 × 10⁵ cells per well) to the outer wells. Dishes were incubated at 37°C in 7% CO₂ for 2 h. Images were acquired in DIC mode with a Nikon A1R confocal microscope, using a 10× air objective with a numerical aperture of 0.45 and the transmitted light detector. The manual measurement tool in the NIS-Elements software program (Nikon) was used to determine the distance traveled by the cell that had migrated the farthest and is reported as maximum distance. To quantify the relative number of cells that moved toward LTB4, a 15.24- × 18.42-cm box was superimposed on each image, and the number of flattened cells that had migrated under the agarose into that area were counted (Supplemental Fig. 1B). As a control, an image was acquired at the rear edge of each well to ensure that no random migration had taken place in the direction away from the chemoattractant (data not shown).

PMN were sorted using anti-Ly6C/Brilliant Violet 421 and anti-Ly6G/PerCpCy5.5 Abs (BioLegend) and then incubated in RP-10(−) medium with PGE₂ (final concentration 1 μM) or vehicle (ethanol) for 60 min at 37°C in 7% CO₂. L. monocytogenes SD2710 (GFP⁺) was added, and the plate was centrifuged for 5 min at 300 × g, followed by incubation for 5 min at 37°C in 7% CO₂. Polyclonal rabbit anti–Listeria O primary antisera (Difco Laboratories) and goat anti-rabbit IgG–Texas Red secondary Ab (Thermo Fisher Scientific) were used to perform differential “in/out” staining as described previously (27). Cells were visualized using a Zeiss Axio Imager.Z1 with a 100×/1.4 NA PlanApo oil immersion objective and analyzed with AxioVision software. Slides were blinded and examined by two different investigators; average values are reported.

PMN were plated in 96-well, flat-bottom dishes in RP10(−) with 1 μM PGE₂ or vehicle and incubated for 1 h. The plate was then centrifuged at 300 × g, and cells were suspended in Ca²⁺/Mg²⁺-free HBSS with or without 1 μM PGE₂. Dihydrorhodamine 123 (no. 85100; Cayman Chemical) suspended in DMSO and diluted in HBSS was added to wells (final concentration, 2.5 μg/ml) 5 min prior to addition of either L. monocytogenes or 20 nM PMA (no. P1585; Sigma-Aldrich) diluted in HBSS. Plates were centrifuged at 300 × g for 5 min and then incubated at 37°C in 7% CO₂ for 10 min. The reaction was stopped by the addition of cold flow cytometry buffer, and cells were either stained for PMN markers or directly fixed using 10% neutral buffered formalin. Rhodamine 123 fluorescence was analyzed by flow cytometry in the FL-1 channel.

Mice were euthanized, and livers were perfused via the hepatic portal vein with 11 ml of collagenase type IV solution (250 U/ml in HBSS; Worthington). The gallbladder was removed, and perfused tissue was cut into small pieces, transferred to a 50-ml tube containing DNase (10 U/ml; Worthington) and collagenase type IV (100 U/ml) in 10 ml of RP5(−), and incubated for 35-min shaking at 37°C. The digested tissue was gently pushed through a sterile no. 80 wire mesh screen to create a single-cell suspension. One 10th of the volume was removed for CFU determination by centrifuging for 10 min at 20,000 × g, suspending the pellet in 300 μl sterile water, vortexing for 30 s and incubating for 10 min at room temperature. Dilutions were prepared in sterile water, plated on BHI agar, and incubated at 37°C.

Parenchymal cells were removed by allowing 2 min settling time, followed by centrifugation for 1 min at 50 × g at 4°C. The supernatant was collected and centrifuged at 300 × g for 10 min at 4°C. The cell pellet was suspended in 30 ml of RPMI 1640, and the prior two steps were repeated. Enriched hematopoietic cells were suspended in 1.6 ml cold RPMI 1640, followed by addition of 2.4 ml cold 40% HistoDenz (Sigma-Aldrich) in PBS. The suspension was layered under 2 ml of cold RPMI 1640 in 15-ml polypropylene tubes that had been precoated with FBS. Samples were centrifuged for 20 min at 4°C at 1500 × g with no brake. The interface was collected, and the cells were passed through a nylon filter to remove clumps.

Celecoxib (no. 10008672; Cayman Chemical) was suspended in ethanol and stored at −20°C. Aliquots were mixed 1:1 with vanilla extract (Kroger) to increase palatability, and mice were treated at 10 mg/kg orally every 12 h beginning at 48 h postinfection (hpi). Vehicle control mice were given an equal volume of ethanol plus vanilla. For indomethacin challenge experiments, female BALB/cByJ mice were injected i.p. twice per day with 2 mg/kg indomethacin (no. 70270; Cayman Chemical) suspended in DMSO and diluted in sterile HBSS. For PGE₂ challenge experiments, female B6 mice were injected i.p. with 40 μg of 16,16 dM PGE₂ (no. 14750; Cayman Chemical; suspended in DMSO and diluted in sterile HBSS) once daily, beginning immediately postinfection. At 4 d postinfection (dpi), organs were harvested, homogenized, and plated on BHI agar.

Statistical analysis was performed using Prism for Macintosh Version 6.0f (GraphPad). The specific test used for each data set is indicated in the figure legends.

Grainger et al. (10) previously showed that oral infection of C57BL/6J mice with the obligate intracellular parasite T. gondii caused Ly6C^hi monocytes in the intestinal lamina propria to secrete PGE₂. To find out if infection of susceptible BALB/cByJ mice with the facultative intracellular bacterial pathogen L. monocytogenes also induced PGE₂, we first measured COX-2 expression in bone marrow–derived phagocytes in vitro. Culturing BALB/cByJ bone marrow progenitors with rM-CSF generated a characteristic “waterfall of differentiation,” including Ly6C^hiCD64⁽⁻⁾ monocytes, transitioning cells, and Ly6C^loCD64⁺ macrophages (Fig. 1A). The cultured cells were exposed to either live or heat-killed L. monocytogenes for 2 h, and then gentamicin was added to prevent extracellular growth of the bacteria. Intracellular COX-2 levels in monocytes, transitioning cells, and macrophages were assessed the next day by flow cytometry. There were no differences in the percentage of each cell type found in the culture wells after treatment with live or heat-killed L. monocytogenes (data not shown). However, the number of COX-2⁺ cells increased significantly within this time period for both transitioning cells and fully differentiated macrophages (Fig. 1B). Macrophages also displayed an increase in COX-2 median fluorescence intensity after being exposed to L. monocytogenes.

To determine if COX-2 expression was also induced in the liver during L. monocytogenes infection, nonparenchymal cells from infected (3 dpi) or uninfected BALB/cByJ and C57BL/6J mice were isolated, and intracellular staining for COX-2 was performed directly ex vivo. Monocytes, PMN, and KC were defined using the gating strategy shown in Fig. 1C. No significant changes in COX-2 median fluorescence intensity were noted for any of these cell types in infected mice compared with uninfected mice for either mouse strain (Fig. 1D). These data suggested that exposure to L. monocytogenes could rapidly increase COX-2 expression in myeloid-derived phagocytes, but this induction may not be sustained in a majority of cells in the liver at any one time point during the course of infection.

To assess PG production in the livers of susceptible BALB/cByJ and resistant C57BL/6J mice more directly, we next sought to measure total PGE₂. PGE₂ is metabolized very rapidly, making ex vivo measurement in a mixed population of cells difficult (34, 35). PGEM is a stable derivative of PGE₂ that serves as a useful proxy in situations in which PGE₂ is secreted but can also be metabolized by neighboring cells. Groups of mice were fed L. monocytogenes or mock-infected, and livers were harvested 3 dpi. The livers were homogenized, and clarified supernatants from the bulk population of cells were assayed, but PGEM could not be consistently detected in these samples (Fig. 2A). However, PGEM was readily detected when gradient-enriched liver nonparenchymal cells harvested from infected mice were incubated overnight in tissue culture media.

On average, cells from L. monocytogenes–infected BALB/cByJ livers produced 5-fold more PGE₂ than cells from uninfected livers (Fig. 2B). In contrast, cells from infected C57BL/6J mice displayed no induction of PGE₂ expression compared with uninfected cells. This result suggested that BALB/cByJ liver cells were primed during infection to produce more PGE₂ than cells from C57BL/6J liver. However, we have previously shown that the bacterial burden in susceptible BALB/cByJ mice 3 d following foodborne transmission is slightly higher than in C57BL/6J mice (23–25). Thus, the nonparenchymal cells harvested from BALB/cByJ livers may have harbored more intracellular L. monocytogenes, and the PGE₂ induction observed could have been caused solely by secondary in vitro exposure to L. monocytogenes released from infected cells. To rule out this possibility, liver nonparenchymal cells were isolated from uninfected BALB/cByJ mice and cultured for 2 h with either live L. monocytogenes or heat-killed, sonicated bacteria, and then gentamicin was added, and the cells were incubated overnight. As shown in Fig. 2C, in vitro stimulation of naive cells with L. monocytogenes did not induce PGE₂ secretion. For comparison, cells from infected mice were further stimulated in vitro with heat-killed, sonicated L. monocytogenes, and the amount of PGE₂ produced approximately doubled compared with the unstimulated cells (Fig. 2C versus Fig. 2B). Therefore, only cells taken from infected mice were capable of secreting PGE₂ into the culture supernatant, presumably because they had been primed in vivo. Together, these results suggested that there may be a higher concentration of PGE₂ in the livers of infected BALB/cByJ mice compared with C57BL/6J mice.

Having established that PGE₂ was produced in response to L. monocytogenes infection, we next sought to determine if PGE₂ pretreatment directly affected the ability of PMN to kill the bacteria. Gradient-enriched PMN harvested from the bone marrow of naive BALB/cByJ or C57BL/6J mice were plated in tissue culture media, and PGE₂ was added for either 30 or 90 min prior to the addition of opsonized L. monocytogenes. As shown in Fig. 3A, PGE₂ treatment for as little as 30 min resulted in a dose-dependent reduction in killing efficiency compared with vehicle-treated cells. Pretreatment with 1 μM PGE₂ significantly decreased the killing efficiency of both BALB/cByJ and C57BL/6J PMN, an effect that was highly reproducible, regardless of whether cells were harvested from male or female mice (Fig. 3B).

To verify that the decreased killing function we observed was a direct effect of PGE₂ signaling, PMN were treated with EP2 and EP4 receptor antagonists. As shown in Fig. 3C and 3D, blocking either the EP2 or EP4 receptor at least partially rescued the PGE₂-dependent loss of killing. No synergistic effect was observed when the cells were treated with both receptor antagonists concurrently (data not shown). These results suggested that PGE₂ binding to either EP2 or EP4 results in diminished killing capacity for PMN and demonstrated that there were no inherent differences in the ability of PMN from either BALB/cByJ or C57BL/6J mice to respond to PGE₂ treatment.

To assess the effect of PGE₂ pretreatment on PMN chemotaxis, gradient-enriched bone marrow PMN from BALB/cByJ and C57BL/6J mice were allowed to migrate toward LTB4 in an under-agarose assay. The number of cells that moved in the direction of the LTB4 gradient and the maximum distance traveled by the farthest cell within 2 h was measured. As shown in Fig. 4A, PGE₂-treatment reduced the number of migrating cells in the majority of wells examined and had a small, but reproducible, effect on the maximum distance traveled by PMN harvested from BALB/cByJ mice. Because LTB4 is known to be produced by PMN that have infiltrated sites of infection or inflammation (36, 37), these data suggested that PGE₂ production in the liver might inhibit PMN swarming at infectious foci.

Phagocytic capacity was assessed using sort-purified bone marrow PMN from uninfected BALB/cByJ mice. The cells were pretreated with PGE₂ or vehicle and then GFP-expressing L. monocytogenes were added. Based on a previous study, we expected the majority of the cell-associated bacteria to be internalized within 10 min (25). At that time point, the cells were washed and extracellular bacteria were stained with Texas Red–conjugated Ab. Because the cells were not permeabilized and the Abs did not have access to intracellular bacteria, any internalized L. monocytogenes appeared green, whereas extracellular or adherent bacteria appeared orange/yellow in the merged images. As shown in Fig. 4B, PGE₂ pretreatment reproducibly decreased uptake of L. monocytogenes by ∼20%. We previously showed that a small number of internalized L. monocytogenes were not immediately degraded and could be detected as gentamicin-resistant intracellular bacteria (25). To track the fate of internalized L. monocytogenes in PGE₂-treated cells, we performed a gentamicin protection assay at either 25 or 45 min after addition of bacteria. As expected, PGE₂ pretreatment of PMN reduced the number of gentamicin-resistant bacteria recovered at both time points (Supplemental Fig. 1C). Thus, PMN were impaired in their ability to phagocytose L. monocytogenes following PGE₂ exposure.

Because phagocytosis is known to trigger assembly of NADPH oxidase (38, 39), we predicted that PGE₂ pretreatment would also decrease ROS production in murine bone marrow PMN. BALB/cByJ PMN were pretreated with PGE₂ or vehicle, and flow cytometric detection of rhodamine fluorescence was used as a readout for ROS generation. Adding L. monocytogenes to the PMN increased the number of ROS-positive cells within 10 min, and PGE₂ treatment reduced the ROS burst by more than 50% (Fig. 4C). ROS are not absolutely required for murine PMN to kill L. monocytogenes, but we previously showed that generation of an ROS burst can enhance killing (25). If the PGE₂-mediated reduction in killing was dependent on inhibition of NADPH oxidase in PMN, then we would expect to see no decrease in bacterial killing for gp91phox^−/− cells treated with PGE₂. Indeed, both PGE₂ pretreated and vehicle-treated PMN harvested from gp91phox^−/− mice readily killed L. monocytogenes (Fig. 4D), suggesting that the PGE₂-dependent reduction in killing could not be achieved in the absence of ROS production. Together, these data indicated that exposure to PGE₂ diminished the ability of PMN to kill L. monocytogenes through multiple mechanisms, including reduced chemotaxis, phagocytosis, and ROS generation.

PGE₂ had a dose-dependent effect on bacterial killing by PMN in vitro, and innate immune cells isolated from the livers of BALB/cByJ mice produced more PGE₂ than cells from C57BL/6J mice. Therefore, we hypothesized that PMN harvested from the livers of infected BALB/cByJ mice would display impaired killing of L. monocytogenes ex vivo. To test this, BALB/cByJ and C57BL/6J mice were fed L. monocytogenes SD2000, and PMN were sorted from the liver 4 d (96 h) later (Fig. 5A). This time point was chosen because it was previously shown that the number of L. monocytogenes in BALB/cByJ livers steadily increased until at least 5 dpi, whereas bacterial burdens in C57BL/6J mice decreased from 3 to 5 dpi, presumably due to efficient clearance by phagocytes (23). Sorted PMN were then exposed to a kanamycin-resistant strain (L. monocytogenes SD2001) to assess killing capacity directly ex vivo. Surprisingly, no difference in the percentage of L. monocytogenes killed was observed when comparing BALB/cByJ and C57BL/6J PMN (Fig. 5B). This result could indicate that the liver milieu in BALB/cByJ mice did not adversely affect PMN function; however, it was also possible that too much time had elapsed during the processing of the sorted cells (∼4 h), relieving any inhibitory effect of the liver microenvironment.

To assess the stability of PGE₂-dependent inhibition of PMN killing, we collected bone marrow from an uninfected mouse and exposed the bulk population of cells to PGE₂ for 30 min. We then proceeded with the gradient enrichment process, a shorter procedure that normally takes about one fourth of the time required for sorting cells by flow cytometry. Although the purity of gradient-enriched PMN is less than for sorted cells, we previously showed that the few contaminating cell types present in the gradient-enriched population were unable to kill L. monocytogenes directly ex vivo (25). As shown in Fig. 5C, PMN that were exposed to PGE₂ and then processed for 1 h without exposure to PGE₂ killed, on average, 40% less L. monocytogenes than untreated cells. Thus, continuous exposure to PGE₂ was not required to observe defects in PMN killing. To define the sensitivity required to detect impairment in the ex vivo killing assay, PMN from uninfected mice were gradient-enriched, exposed to PGE₂ for 30 min and then mixed in increasing proportions with untreated PMN before addition of L. monocytogenes. The results of this experiment indicated that prior exposure to PGE₂ was required in more than 80% of the PMN to observe a statistically significant reduction in ex vivo killing (Fig. 5D). Together, these data suggested that PMN exposed to PGE₂ in vivo would maintain their impaired killing function during in vitro gradient enrichment; however, the lack of an ex vivo killing phenotype could be the result of testing bulk populations of cells that may vary in their level of PGE₂ exposure during infection of mice.

To reduce the infection-induced production of PGE₂ in BALB/cByJ mice, animals were treated with celecoxib, a nonsteroidal anti-inflammatory drug that selectively inhibits COX-2. Mice were fed L. monocytogenes and starting at 24 hpi, one group of BALB/cByJ mice was given an oral dosage of celecoxib every 12 h (Fig. 6A). The control group of BALB/cByJ mice and a group of C57BL/6J mice for comparison were given oral treatments of vehicle at the same time points. At 92 hpi, livers were perfused, and a portion of gradient-enriched cells was cultured overnight to assay for PGE₂ secretion. The remainder of the cells were used in an ex vivo killing assay. As shown in Fig. 6B, reduced amounts of PGE₂ were detected in the liver cells from the majority of BALB/cByJ mice treated with celecoxib compared with vehicle-treated mice. However, this treatment regimen did not result in a change in the total CFU found in the liver (Fig. 6C). Likewise, neither treatment of C57BL/6J mice with synthetic PGE₂ to increase systemic PGE₂ levels or treatment of BALB/cByJ mice with indomethacin, a nonspecific COX inhibitor, altered bacterial burdens at 72 hpi (Supplemental Fig. 2).

Repeated celecoxib treatment of BALB/cByJ mice also did not result in a detectable difference in ex vivo killing function for gradient-enriched PMN (Fig. 6D). One previous study suggested that long-term exposure to PGE₂ would decrease the presence of both EP2 and EP4 receptors on the cell surface (40). To find out if cells harvested from the livers of mice 4 dpi were still responsive to PGE₂, gradient-enriched PMN were incubated with PGE₂ for 1 h and then in vitro killing function was assessed. Cells harvested from vehicle-treated mice displayed a small, but consistent, decrease in killing efficiency following exposure to PGE₂ in vitro (Fig. 6E). In contrast, PMN harvested from the livers of celecoxib-treated mice displayed a variable response. For three of the seven mice tested, additional PGE₂ treatment decreased the ability of liver PMN to kill L. monocytogenes ex vivo; for the remaining four mice, PGE₂ treatment in vitro slightly increased killing efficiency. Together, these results indicated that increased in vivo production of PGE₂ during L. monocytogenes infection did not directly affect PMN function ex vivo, which suggests that there may be other signals in the inflammatory milieu of the liver that can modulate the effects of PGE₂.

Neutrophil killing is an important component of the innate immune response that leads to clearance of extracellular L. monocytogenes. We show, in this study, that murine PMN exposed to PGE₂ have a decreased ability to kill opsonized L. monocytogenes in ex vivo killing assays. Multiple mechanisms appeared to contribute to this phenotype, as PGE₂ inhibited neutrophil migration, phagocytosis of L. monocytogenes, and production of ROS. Although PMN from both BALB/cByJ and C57BL/6J mice were equally susceptible to the effects of PGE₂, when livers from infected mice were assayed, BALB/cByJ cells produced significantly more PGE₂ than C57BL/6J cells. Together, these results suggested that excess PGE₂ production could help to explain why BALB/cByJ mice are more innately susceptible to listeriosis than C57BL/6J mice. However, treating mice with celecoxib to reduce PGE₂ production during L. monocytogenes infection did not alter CFU burdens in the either the spleen or the liver and had no effect on PMN killing when tested directly ex vivo.

Overall susceptibility to infectious disease is typically multifactorial, and indeed, several genetic loci associated with either susceptibility or resistance or differences in innate immune responses to L. monocytogenes in BALB/c and C57BL/6J mice have already been identified (41–44). In this context, excess PGE₂ production by BALB/cByJ mice could contribute to an enhanced susceptibility phenotype by making PMN killing less efficient. Several factors make it difficult to examine the effects of PGE₂ directly ex vivo. First, the sensitivity of our killing assay was not high, with an in vitro control study indicating that at least 80% of the PMN needed to be exposed to PGE₂ to observe a decrease in bacterial killing by a bulk population of cells. It is unlikely that every PMN recruited to the infected liver of a mouse will be exposed to elevated levels of PGE₂, particularly if PGE₂ secreted near infectious foci in the liver results in decreased swarming of neutrophils toward LTB4 or bacterial products. Another challenge is the lack of certainty regarding the length of time that may have passed since any given PMN was exposed to PGE₂ in vivo.

In previous studies, PGE₂ was shown to inhibit chemotaxis of human lymphocytes toward C5a (45) and human PMN toward the N-formylated peptide fMLF (19) and movement of equine PMN toward LTB4 (46). In agreement with those reports, we found that PGE₂-treated murine PMN were impaired in their ability to migrate toward LTB4. Although ligation of either EP2 or EP4 receptors leads to increased accumulation of cAMP, these two receptors have different affinities for PGE₂ and can also cause divergent downstream signaling in cAMP-independent pathways. For example, Armstrong et al. (19) found that EP2 agonists mediated inhibition of chemotaxis toward fMLF, but an adenylate cyclase inhibitor did not antagonize the inhibition, suggesting that this effect is not cAMP dependent, and may result from selective signaling downstream of EP2. Likewise, activation of adenylate cyclase by forskolin inhibited human PMN chemotaxis toward formylated peptides, but not LTB4 (18, 19). In support of this, we found that simultaneous antagonism of both of these receptors did not lead to a synergistic effect, suggesting that the signaling triggered by each was nonoverlapping. Ligation of EP4 resulted in adenylate cyclase activity after just brief exposure to PGE₂, but the cAMP response waned rapidly (40). The same study showed that signaling through EP2 required significantly more exposure time and was more sensitive to the first metabolite of PGE₂, 15-keto PGE₂. These observations suggest that the EP2 receptor may be responsible for the more long-lasting effects of PGE₂ during infection and may help to explain why we did not uniformly observe an inhibition of chemotaxis in every murine PMN examined.

We found that PGE₂ inhibition of phagocytosis in murine PMN was reproducible, but not as substantial as previously observed in other studies that examined the effects of PGE₂ on macrophages (9, 20). In this study, the L. monocytogenes used in killing assays were opsonized with 10% mouse serum because we previously showed that killing of nonopsonized bacteria by PMN was slower and less efficient (25). Serum opsonization may have resulted in more uptake of bacteria in our PMN via pathways that are less sensitive to the effects of PGE₂. For example, Rossi et al. (47) showed that elevated cAMP decreased the uptake of apoptotic cells by macrophages, but did not alter phagocytosis of IgG-opsonized erythrocytes. In another study, Domingo-Gonzalez et al. (21) found that elevated PGE₂ after bone marrow transplant dysregulated alveolar macrophage expression of scavenger receptors, increasing MARCO and decreasing SR-AI/II. This increased phagocytosis of the Gram-positive bacterium S. aureus, but ultimately resulted in a decrease in intracellular killing.

PGE₂ signaling can trigger both inflammatory and anti-inflammatory responses, so an understanding of the exact timing of PGE₂ secretion, as well as the cellular context within a given tissue, is critical for determining the ultimate role of PGE₂ production in infection. For example, in a study by Rangel Moreno et al. (48), use of a systemic COX-2 inhibitor during the acute phase of murine Mycobacterium tuberculosis infection resulted in increased lung colonization; however, when administered later, during the chronic phase of infection, COX-2 inhibition resulted in improved clearance. These results support of the conclusions of Tripp et al. (49), who suggested that PGE₂-mediated decreases in MHC class II (I-A) expression on macrophages could signal the resolution phase of infection. Additionally, Levy et al. (50) have proposed that accumulation of one eicosanoid can lead to an enzymatic changeover to production of another class and ultimate resolution, a temporal phenomenon they called an eicosanoid switch. Although we did not assay any eicosanoids other than PGE₂ in this study, it is intriguing to speculate that such a mechanism might be involved in the clearance of L. monocytogenes infection and that the overall ratio of eicosanoids present in the liver could vary in BALB/cByJ and C57BL/6J mice.

Dr. Greg Baumann and Jennifer Strange in the U.K. Flow Cytometry and Cell Sorting Core Facility and Dr. Thomas Wilkop in the U.K. Microscopy Core Facility provided technical support for these studies. We are also grateful for the technical assistance of Travis Combs and Jessica Ferrell. We thank Dr. David Nardo for careful reading of this manuscript, Dr. Jeff Rush for providing a nitrogen evaporator for PGEM analysis, and Matt Christensen for suggesting strategies to increase the palatability of celecoxib in mice.

The authors have no financial conflicts of interest.