Introduction

Prostaglandin E2 (PGE2) exhibits diverse hormone-like effects in animals, acting through a wide range of physiological targets and mechanisms, such as neurons, smooth muscles, secretary cells, and immune function [1,2,3]. Immune function of PGE2 in modulating neuro-inflammation has been an active area of research, but the roles of PGE2 as a pro-inflammatory or anti-inflammatory mediator are far from clear. Under normal conditions, a modest level of PGE2 is produced in neurons, microglia, and astroglia; however, a large amount of PGE2 is produced mainly by microglia during neuro-inflammation [4]. Bifunctional effects of PGE2 on inflammatory conditions have been documented. In pathological conditions, PGE2 regulates multiple processes in the nervous system, including fever generation, sickness behavior, pro-inflammatory cytokines production, such as IL-6 and IL-1β, and also contributes to neuropathic pain and neuro-degenerative disorders [5,6,7,8,9,10,11,12,13]. Conversely, PGE2 also exhibits anti-inflammatory functions [14, 15]. Several lines of evidence suggest that PGE2 generated during periods of inflammation is subject to negative feedback in order to downregulate other pro-inflammatory factors [16, 17]. Moreover, accumulating in vivo and in vitro studies reveal that exogenous PGE2 is able to inhibit lipopolysaccharide (LPS)-induced microglial activation and the production of pro-inflammatory factors, including TNF-α, IL-1β, iNOS, IL-6, and IL12 [18,19,20,21,22].

Numerous studies reveal that PGE2 displays functionally opposing actions in distinct cell types or disease models, including pro- and anti-inflammatory effects, or toxic and protective effects to neurons [11, 14, 23,24,25,26,27]. The reason for these opposing reports is not clear. We speculated that the difference in concentration of PGE2 used in various reports could contribute to the controversial conclusions. Most pharmacological studies involve pathological PGE2 concentrations ranging from 100 nM to 10 μM, which is 100- to 1000-fold higher than physiological concentrations of PGE2 in the central nervous system (CNS) (< 10 nM), while very few reports use low concentrations of PGE2 [28,29,30,31,32,33]. The purpose of this study was to gain a further understanding of the physiological function of PGE2 in the modulation of brain immune homeostasis by focusing on the effects of sub-nanomolar concentrations of the prostaglandin. In the present study, we investigated whether low concentrations of PGE2 (0.01 to 10 nM) were capable of eliciting a neuroprotective effect through mediation of pro-inflammatory factors, especially superoxide free radicals produced by the enzyme NADPH oxidase (NOX2), in an LPS-induced neuro-inflammation model. We found that low concentrations of PGE2 displayed both anti-inflammatory and neuroprotective functions, and further investigation revealed that anti-inflammatory effects were independent of prostaglandin E2 receptors (EP). Additionally, we discovered that inhibition of NOX2-generated superoxide production from microglia was the key mechanism underlying the anti-inflammatory function of sub-nanomolar concentrations of PGE2.

Materials and Methods

Animals

Timed-pregnant (gestational day 14) female Fisher 344 rats were purchased from Charles River Laboratories (Raleigh, NC, USA). Timed-pregnant C57BL/6J and B6.129S-Cybb<tm1Din>/J (gp91 deficient) mice were generated by our institute’s animal husbandry staff using breeders obtained from Jackson Laboratories (Ben Harbor, ME, USA). EP2 knockout mice were kindly gifted from Dr. Robert Langenbach at National Institute of Environmental Health Sciences. Rat and mouse dams were housed in polycarbonate cages in animal facilities with controlled environment conditions with a 12-h artificial light-dark cycle and provided fresh deionized water and NIH 31 chow ad libitum. All animal procedures were approved by the Institutional Animal Care and Use Committee and conducted in strict accordance with the National Institutes of Health animal care and use guidelines.

Reagents

Xanthine, xanthine oxidase, SQ 22,536, L798106, poly-d-lysine, superoxide dismutase (SOD), mazindol, 3,3′-diaminobenzidine, and urea-hydrogen peroxide tablets were purchased from Sigma-Aldrich (St. Louis, MO, USA). Water-soluble tetrazolium salt-1 (WST-1) was purchased from Dojindo (Rockville, MD, USA). Lipopolysaccharide (LPS; E. coli strain O111: B4) was purchased from Calbiochem (San Diego, CA, USA). Cell culture ingredients were obtained from Life Technologies (Grand Island, NY, USA). Prostaglandin E2 (PGE2), 17-phenyl trinor prostaglandin E2, butaprost, CAY-10598, SC-19220, AH-6809, and AH-23848 were purchased from Cayman Chemical (Ann Arbor, MI, USA). Anti-tyrosine hydroxylase (TH) was purchased from Chemicon (Billerica, MA, USA) and antibody diluent was purchased from DAKO (Carpinteria, CA, USA). TNF-α ELISA kit was purchased from R&D Systems (Minneapolis, MN, USA). Direct cAMP ELISA kit was purchased from Enzo Life Sciences (Farmingdale, NY, USA). Anti-p47phox antibody was purchased from Millipore (Temecula, CA, USA). Anti-gp91phox antibody was purchased from BD Biosciences (San Jose, CA, USA). Anti-GAPDH antibody was purchased from Abcam (Cambridge, MA, USA). Alexa Fluor 488 goat anti-rabbit IgG was purchased from Invitrogen (Carlsbad, CA, USA). Goat anti-rabbit biotinylated secondary antibody was purchased from Vector Laboratory (Burlingame, CA, USA).

Mesencephalic Neuron-Glia Culture

Rat mesencephalic neuron-glia cultures were prepared following protocols described previously [34, 35]. Briefly, midbrain tissues were dissected from day 14 embryos and then gently triturated into single cell suspension. Cells were then seeded (5 × 105 cells/well) in poly-d-lysine (20 μg/ml) precoated 24-well plates. The cultures were incubated at 37 °C in 5% CO2 for 3 days and then replenished with 500 μl of fresh maintenance media. Cultures were treated 7 days after seeding.

Primary Cortical Mixed Glial Culture

Primary cortical mixed glial cultures were prepared from rat or mouse pup brains at postnatal days 1–3, as previously described [34, 35]. Briefly, the cortices were isolated, the meninges and blood vessels removed, the tissue gently dissociated through trituration, and the single cell suspension plated on either 24-well plates or 96-well plates precoated in poly-d-lysine (20 μg/ml) at 1 × 105 cells/well or 5 × 104 cells/well, respectively. Cells were maintained in DMEM-F12 (1:1) media supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mM l-glutamine, 1 mM sodium pyruvate, 100 μM non-essential amino acids, 50 U/ml penicillin, and 50 μg/ml streptomycin. Media was refreshed every 3 days until they were experimentally treated 7 days after seeding.

Microglia-Enriched Cultures

Microglia-enriched cultures were prepared from primary mixed glial cultures as previously described [34, 35]. Briefly, mixed glia cultures were plated on 150-cm3 flasks precoated in poly-d-lysine (20 μg/ml) at 5 × 107 cells/flask and were maintained in DMEM-F12 media changed every 3 days for 2 weeks. At 2 weeks, microglia were shaken off at 180 rpm for 40 min and re-plated either on 6-well plates precoated in poly-d-lysine (20 μg/ml) at 1 × 106 cells/well for cAMP assay.

Cell Line Culture

Mycoplasma-free rat HAPI microglial cell line was maintained in DMEM supplemented with 10% FBS, 50 U/ml penicillin, and 50 μg/ml streptomycin. Cells were split into T75 flasks and used for experiments at passages 3–5, where they were seeded in 24-well plates at 1 × 105 cells/well and treated 24 h later.

Wild-type and transfected monkey kidney COS7 cells were gifts from Dr. Mary Dinauer (Indiana University, IN, USA) and cultured as described previously [36, 37]. Briefly, wild-type COS7 cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS), 50 U/ml penicillin, and 50 μg/ml streptomycin. COS7 cells stably expressed with gp91phox, p22phox, p47phox, and p67phox (COS-NOX2 cells) were maintained in DMEM as above containing 0.8 mg/ml Geneticin, 2 μg/ml puromycin, and 0.35 mg/ml hygromycin B.

Dopamine Uptake Assay

The [3H] dopamine (DA) uptake assay was performed as described previously [38]. Briefly, the rate of uptake of radiolabeled DA by DAnergic neuron cultures was measured for 21 min at 37 °C. Cells were washed and lysed to release internalized radiolabeled DA and quantified with a liquid scintillation counter (Tri-Carb 4000; Packard, Meriden, CT, USA). Non-specific [3H] DA uptake was accounted for by competitively inhibiting DA uptake with 10 μM of mazindol.

Immunocytochemical and Immunofluorescence Staining

For immunocytochemical staining, neuron-glia cultures were fixed with 3.7% formaldehyde in phosphate-buffered saline (PBS) for 20 min; fixed cultures were treated with 1% hydrogen peroxide for 10 min and incubated for 20 min with blocking solution (BSA 1%/Triton X-100 0.4%/normal goat serum 4% in PBS). Cells were immunostained overnight at 4 °C with rabbit polyclonal antibody against tyrosine hydroxylase (TH; 1:5000) in Antibody Diluent. Cultures were washed for 10 min in PBS (three times) and incubated for 2 h with PBS containing 0.3% Triton X-100 and a biotinylated secondary antibody (goat anti-rabbit antibody, 1:227; Vector Laboratory, Burlingame, CA). After washing (three times) with PBS, the cultures were incubated for 1 h with the Vectastain ABC reagents (Vector Laboratory, Burlingame, CA) diluted in PBS containing 0.3% Triton X-100. To visualize the signal, the cultures were incubated with 3,3′-diaminobenzidine and urea-hydrogen peroxide tablets dissolved in water. TH-positive cells were manually counted under a microscope (Nikon, model DIAPHOT, Garden City, NY, USA) by at least two investigators and the results were averaged.

For immunofluorescence staining, microglia-enriched cultures were fixed with 3.7% formaldehyde in PBS for 20 min and incubated for 20 min in blocking solution (BSA 1%/Triton X-100 0.4%/normal goat serum 4% in PBS) to block non-specific binding. Cells were immunostained overnight at 4 °C with rabbit polyclonal antibody against p47phox (1:1000) diluted in Antibody Diluent. The antibody was detected and visualized using Alexa Fluor 488 goat anti-rabbit IgG (1:750) secondary antibody. The images were acquired using multiphoton laser-scanning microscope Zeiss 710.

Peritoneal Macrophage Isolation

Macrophages were obtained from peritoneal cavities of mice using protocols described previously [39]. Briefly, mice were injected with 1 ml of 3% Brewer thioglycollate medium into the peritoneal cavity. After 4 days, mice were euthanized, and the peritoneal cavity was flushed with 5 ml ice-cold RPMI medium 1640. After being washed twice in RPMI, the cells were preincubated in the no-serum medium for 1 h. The cells were then washed twice to remove non-adherent cells. Adherent macrophages were cultured in DMEM containing 10% FBS, 50 U/ml penicillin, and 50 μg/ml streptomycin at 37 °C in a humidified atmosphere of 5% CO2 and 95% air.

[3H]-Labeled PGE2 Binding Assay

The binding assay was performed according to protocols described previously [40]. Briefly, macrophages from C57BL/6J and CYBB (gp91 deficient) mice, wild-type COS7 cells and COS7 cells stably expressing gp91phox, p47phox, and p67phox (COS7-NOX2) (2 × 106 cells) were incubated with 0.3 nM of 3H-labeled PGE2 in binding buffer (HBSS + 10% fetal bovine serum) for 2 h at 37 °C. After binding, the cells were transferred to glass fiber filter with 1 μm pore size and washed three times using warm HBSS. Then the filters were put into vials containing 15 ml of Ultima Gold scintillation fluid (PerkinElmer, Waltham, MA, USA) and counted for radioactivity. The results are expressed as a percentage of total binding observed with corresponding COS7-WT or C57BL/6 group.

Saturation curves of [3H]-PGE2 binding were performed using COS7-WT and COS7-NOX2 cells. Total binding was measured by the binding of [3H]-labeled PGE2 ranging from 0.03 to 3 nM while non-specific binding was determined in the presence of additional 3 μM of PGE2. Specific binding was calculated by the difference between total binding and non-specific binding.

TNF-α, MCP-1, and Nitrite Measurement

The release of TNF-α and MCP-1 was measured from culture supernatant with a commercial enzyme-linked immunosorbent assay (ELISA) kit from R&D Systems, in accordance with the manufacturer’s instructions. NO was assessed by measuring nitrite in 50 μl of supernatants using Griess reagent as described previously [41]. The culture supernatant was collect at 3 h after treatment for TNF-α and 24 h after treatment for NO and MCP-1 assay. The results were quantified using an SPECTAmax PLUS 384 spectrophotometer (Molecular Devices, Sunnyvale, CA, USA).

cAMP Assay

Primary microglia cells were seeded in 6-well plates (1 × 106 cells/well) and then treated with vehicle, LPS, or PGE2 (1, 10, 50, 100, and 1000 nM) for 30 min. Cells were lysed in 0.1 M HCl for 10 min, and the supernatants were used for cAMP assays following the manufacturer’s protocol of cAMP Enzyme Immunoassay Kit. The absorbance was measured at 405 nm using a colorimetric 96-well plate reader. Data were expressed as picomoles cAMP per 1 million cells.

Superoxide Assay

The production of extracellular superoxide was determined by measuring the superoxide dismutase (SOD)-inhibitable reduction of water-soluble tetrazolium salt-1 (WST-1) [42, 43]. Mixed glial or neuron-glia cultures in 96-well culture plates were incubated at 37 °C for 30 min with vehicle or PGE2. Then the cultures were washed twice with HBSS without phenol red. After washing, the cultures were treated with LPS or LPS + PGE2 in 150 μl phenol red-free, to which 50 μl of WST-1 (1 mM) with and without 800 U/ml SOD was immediately added. The absorbance at 450 nm was read with a SpectraMax Plus microplate spectrophotometer (Molecular Devices, Sunnyvale, CA, USA) and the data at 8–14 min post-treatment were analyzed. The difference in absorbance observed in the presence and absence of SOD was considered to be the amount of superoxide produced, and results were expressed as the percentage of vehicle-treated control cultures. For Fig. 3d, the cultures were pretreated with EP1 to EP4 receptor antagonists for 30 min, and then the steps described above were followed.

Xanthine/Xanthine Oxidase Reaction

Xanthine/xanthine oxidase (X/XO) assays were conducted in the presence of indicated concentrations of PGE2, 0.01 U xanthine oxidase, 50 μM xanthine, and 250 μM partially acetylated WST-1 in 50 mM potassium phosphate buffer (pH 7.6) in a 96-well plate (100 μl/well final volume). Xanthine was added to initiate the reaction, and the absorbance at 450 nm was continuously monitored for 20 min using an SPECTAmax PLUS 384 spectrophotometer (Molecular Devices, Sunnyvale, CA, USA) and the data at 8–14 min post-treatment were analyzed. Superoxide dismutase (SOD) added in the X/XO system served as a positive control.

Plasma Membrane Preparation and Western Blot Analysis

Plasma membrane of HAPI was isolated following a published protocol [44]. Briefly, cells were suspended in isolation buffer (10 mM Tris–Cl, pH 8.0, 0.25 M sucrose, 1 mM EDTA, and protease inhibitor cocktails), and cell membranes were broken by Dounce homogenization. The cell lysates were centrifuged at 6000×g for 10 min at 4 °C to remove unbroken cells, cell debris, and mitochondria; afterward, pellets of membranes were obtained by ultracentrifugation at 100,000×g for 1 h at 4 °C. After being washed with 1 M KCl, membrane pellets were either used fresh or stored at − 80 °C. The pellets were solubilized in 1% Nonidet P-40 hypotonic lysis buffer. Equal amounts of protein were separated by 4 to 12% Bis-Tris Nu-PAGE gel and transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). Membranes were blocked with 5% non-fat milk and incubated with antibodies against p47phox, gp91phox (1:1000 dilution), or GAPDH (1:2500 dilution). The protein bands were developed by incubating with horseradish peroxidase-conjugated secondary antibodies (Vector Laboratories, Burlingame, CA, USA) and an enhanced chemiluminescence substrate kit (Millipore, Billerica, MA, USA). The results were quantified by the AlphaImager 3400 software (Alpha Innotech, Ltd., San Leandro, CA, USA).

Statistics

Data are presented as the mean ± SEM. Comparisons between more than two groups was conducted using one-way ANOVA, followed by Bonferroni’s post hoc multiple comparison test. Comparisons between more than two parameters were conducted by two-way ANOVA analysis, followed by Bonferroni’s post hoc multiple comparison test. Data were analyzed using Prism (v6.00, GraphPad, San Diego, CA). P values less than or equal to 0.05 were considered statistically significant.

Results

Low Nanomolar Concentrations of PGE2 Protect Dopaminergic Neurons Against LPS-Induced Neurotoxicity

To determine the immune regulatory function of PGE2 in concentrations close to those found under normal physiological conditions, mesencephalic neuron-glia cultures were pretreated with PGE2 ranging from 0.01 to 10 nM for 30 min followed by LPS treatment (15 ng/ml). After 7 days, the viability of dopaminergic (DA) neurons was determined by a [3H] DA uptake assay, tyrosine hydroxylase (TH)-positive neuron counting, and morphological analysis. [3H] DA uptake showed that pretreatment of PGE2 protected neurons against LPS-induced neurotoxicity (F(5, 33) = 30, P < 0.0001) in a dose-dependent manner. Post hoc analysis revealed that exposure of LPS reduced DA uptake capacity by 60% compared to vehicle-treated cultures (t = 11.1, P < 0.0001). Pretreatment with PGE2 significantly blocked this reduction at 1 nM (t = 5.028, P < 0.0001) and 10 nM (t = 7.022, P < 0.0001), when compared to DA uptake in LPS-only treatments (Fig. 1a). Consistent with DA uptake results, TH-positive neuron count showed a similar protective effect of PGE2 (F(5, 36) = 11.2, P < 0.0001) (Fig. 1b). Post hoc analysis showed that neuroprotective effects of PGE2 were observed in concentrations as low as 0.1 nM (0.1 nM: t = 3.245, P < 0.05; 1 nM: t = 4.275, P < 0.01; 10 nM: t = 4.838, P < 0.001), compared to the LPS-only treatments. Morphological analysis showed that the neurites of TH-positive neurons in LPS-treated cultures were short and fragmented compared to those in the control cultures. The neurites in cultures pretreated with PGE2 were less affected, compared with neurites in LPS-treated cultures (Fig. 1c). Together, these results indicate that nanomolar concentrations of PGE2 exert potent neuroprotective effects against LPS-induced neurotoxicity.

Fig. 1
figure 1

Reduction of LPS-induced dopamine neuron toxicity in midbrain neuron-glia cultures after 7 days, as seen by a [3H] DA uptake, b TH-positive neuron count, and c immunostaining. Scale bar, 50 μm. Primary rat midbrain neuron-glia cultures were pretreated with PGE2 for 30 min followed by LPS treatment (15 ng/ml). Values are mean ± SEM from three independent experiments, with duplicates for each experiment. Data were analyzed with one-way ANOVA followed by Bonferroni’s post hoc multiple comparison test. *P < 0.05, **P < 0.01, ###, ***P < 0.0001 Bonferroni’s t test compared to vehicle-treated group and LPS-treated group, respectively

Full size image

PGE2 Displays Varying Potency in Inhibiting LPS-Induced Production of Pro-inflammatory Factors in Mixed Glial Cultures

Accumulating evidence indicates that neuro-inflammation caused by over-activated microglia is a critical contributor to neuronal loss in neuro-degeneration. Our previous primary neuron-glia culture studies indicate that microglia are the major source of LPS-stimulated production of pro-inflammatory factors (e.g., cytokines, chemokines, and free radicals), which mediate neuro-inflammation and cause neurotoxicity [8]. To investigate the mechanisms of neuroprotection elicited by nanomolar concentrations of PGE2, we investigated the release of pro-inflammatory factors in rat mixed glial cultures (containing 80% astroglia and 20% microglia). The reason why mixed glia cultures were used in lieu of enriched microglia cultures in this study was microglia can survive longer when cultured together with either neurons or astroglia [45, 46]. The results revealed that PGE2 reduced the production of superoxide, TNF-α, NO, and MCP-1 (superoxide, F(5, 18) = 69.35, P < 0.0001; TNF-α, F(7, 79) = 340.8, P < 0.0001; NO, F(7, 67) = 55.78, P < 0.0001; MCP-1, F(5, 12) = 34.87, P < 0.0001) in a dose-dependent manner (Fig. 2). However, it is interesting to note that the inhibitory potency of PGE2 varies greatly: the half-maximal inhibitory concentration (IC50) value is 0.13 ± 0.08 nM for inhibition of superoxide, 10.14 ± 0.18 nM for TNF-α, and more than 100 nM for both NO and MCP-1 production. These results suggest that PGE2 inhibited production of these pro-inflammatory factors through different mechanisms. Moreover, the similar IC50 values of PGE2 in inhibiting neurotoxicity (0.14 nM) and superoxide production (0.13 nM) suggest a strong likelihood that the inhibition of superoxide production is critical in mediating PGE2-elicited neuroprotective effects. Thus, more detailed studies were conducted to further understand the molecular mechanisms underlying the regulation of superoxide production by nanomolar concentrations of PGE2.

Fig. 2
figure 2

Dose-dependent inhibitory effects of PGE2 on LPS-stimulated production of pro-inflammatory factors a superoxide, b TNF-α, c NO, and d MCP-1 with different IC50 values. Rat mixed glia cultures were pretreated with various concentrations of PGE2 for 30 min followed by LPS treatment. Values are mean ± SEM from three independent experiments, with duplicates for each experiment. *P < 0.05, **P < 0.01, ***P < 0.0001 Bonferroni’s t test compared to LPS-only group

Full size image

Inhibition of Superoxide Production Is Not Due To the Scavenging Effect of PGE2 and Is Not Mediated Through EP Receptors

One of the mechanisms of suppressing superoxide production is scavenging of superoxide by agents of interest. To investigate this possibility, we determined whether nanomolar concentrations of PGE2 possess the ability to scavenge superoxide free radicals. We measured the effect of PGE2 on superoxide production using a xanthine/xanthine oxidase superoxide-producing system. Superoxide production was dramatically inhibited by superoxide dismutase (SOD), but not by PGE2 with different concentrations (0.01–10 nM) (Supplementary Fig. 1).

The lack of a superoxide-scavenging effect of PGE2 prompted us to investigate the potential signaling pathway mediating the inhibitory effects of PGE2 on superoxide production. Using a pharmacological approach, we determined the potency of different agonists selective for EP receptor subtypes in inhibiting LPS-induced superoxide production. Rat mixed glial cultures were pretreated with various agonists, 17-phenyltrinor PGE2 (EP1/3 agonist), butaprost (EP2 agonist), and CAY10598 (EP4 agonist) with concentrations ranging from 0.01 to 10 nM for 30 min prior to LPS (20 ng/ml) stimulation. The results showed that all EP agonists mimicked the inhibitory effects of PGE2 on superoxide production, with IC50 values similar to that of PGE2 (17-phenyl trinor PGE2, F(5, 12) = 60.07, P < 0.0001; butaprost, F(5, 24) = 187.9, P < 0.0001; CAY10598, F(5, 12) = 37.93, P < 0.0001) (Fig. 3a–c). Moreover, simultaneously blocking all four EP receptors by adding receptor antagonists to mixed glial cultures failed to disrupt the inhibitory effect of PGE2 (Fig. 3d). The EP receptor-independent effect of PGE2 was further supported by using EP receptor-deficient mixed glial cultures. Here, we used mixed glial cultures prepared from EP2 receptor knockout mice because the EP2 receptor has been previously implicated to be critical in mediating the anti-inflammatory effect of micromolar concentrations of PGE2 [47,48,49]. PGE2-induced inhibition of superoxide production did not differ between EP2-deficient and control mixed glia cultures, indicating that the EP2 receptor was not involved in mediating nanomolar PGE2-induced inhibition of superoxide (Fig. 3e). Additionally, the EP2 receptor agonist butaprost was still capable of inhibiting superoxide production in EP2-deficient mixed glia cultures (Fig. 3f). Since different signaling pathways mediate each EP receptor subtype and cause distinct cellular effects [3], the similar potencies displayed by these EP receptor agonists suggest that their inhibitory properties may be mediated by a novel, EP receptor-independent mechanism. Taken together, these results demonstrate that inhibition of LPS-induced superoxide is not due to the superoxide-scavenging capacity of PGE2 at low concentrations and is instead mediated through a novel EP receptor-independent mechanism.

Fig. 3
figure 3

Rat mixed glia cultures were pretreated with various concentrations of a 17-phenyltrinor PGE2, b butaprost, and c CAY10598 for 30 min and then exposed to 20 ng/ml of LPS. d Rat mixed glia cultures were pretreated with a combination of 4 EP receptor subtype antagonists, SC-19220 (EP1, 3 μM), AH-6809 (EP2, 3 μM), L-798106 (EP3, 1 μM), and AH-23848 (EP4, 1 μM), for 30 min, followed by 10 nM of PGE2 for additional 30 min, and then exposed to 20 ng/ml of LPS. e Wild-type (C57BL/6) and EP2 receptor KO mixed glia cultures were treated with various concentrations of PGE2 for 30 min and then exposed to 20 ng/ml of LPS. f EP2 receptor KO mixed glia cultures were treated with various concentrations of butaprost for 30 min and then exposed to 20 ng/ml of LPS. The generation of superoxide was measured by monitoring the SOD-inhibitable reduction of WST-1. Values are mean ± SEM from four independent experiments, with duplicates. **P < 0.01, ***P < 0.0001 Bonferroni’s t test compared to LPS-only group

Full size image

NOX2 Plays a Pivotal Role in the Neuroprotective Effects of Low Concentrations of PGE2

As a key superoxide-producing enzyme in activated phagocytes such as microglia, NADPH oxidase (NOX2) is composed of two membrane subunits (gp91phox and p22phox) and several cytosolic subunits (p47phox, p67phox, p40phox, Rac1, and Rap1A). Previous reports indicate that activation of microglial NOX2 is key to the pathogenesis of inflammation-related neuro-degeneration, and its gp91phox subunit is a target for several anti-inflammatory agents [50, 51]. To further investigate whether the neuroprotective effect of low doses of PGE2 is mediated through NOX2, we determined the viability of DA neurons in both wild-type and NOX2-deficient neuron-glial cultures. Genetic ablation of gp91phox abolished neuroprotection by PGE2 (Fig. 4), indicating that NOX2 plays a critical role in mediating the anti-inflammatory and neuroprotective effects of low concentrations of PGE2.

Fig. 4
figure 4

Midbrain neuron-glia cultures from a wild-type and b gp91phox KO mice were pretreated with PGE2 (10 nM) for 30 min and then exposed to LPS (15 ng/ml). After 7 days, the cultures were subjected to DA uptake assay. Values are mean ± SEM from three independent experiments, with triplicates. **P < 0.01, Bonferroni’s t test compared to LPS-only group

Full size image

PGE2 Directly Interacts with NADPH Oxidase Subunit gp91phox

To investigate the possibility that PGE2 suppressed LPS-induced superoxide production by inhibiting NOX2 through binding to the gp91phox subunit, we conducted a binding assay using tritium labeled PGE2. We compared the binding capacity of PGE2 in wild-type COS7 cells and COS7 cells transfected with NOX2 membrane subunits (gp91phox/p22phox) and cytosolic subunits (p47phox and p67phox) (COS7-NOX2). The reason for the co-transfection of gp91phox and p22phox is that p22phox facilitates stable expression of gp91phox in the plasma membrane of COS7 cells [52]. The [3H]-labeled PGE2 binding assay showed that COS7-NOX2 cells have a binding capacity to [3H]-PGE2 that is 2-fold higher than that of wild-type COS7 cells (P < 0.0001, Fig. 5a). We also compared the PGE2 binding capacity in macrophages isolated from wild-type and gp91phox−/− mice. We used peritoneal macrophages instead of microglia in this study for the following reasons: (1) microglia are considered to be the macrophages of the brain, as both cells are derived from bone marrow monocytes, and the protein structure of gp91phox of these two cell types is identical; (2) macrophages have high levels of gp91phox, and it was easy to generate large numbers of primary macrophages required for this study. The [3H]-PGE2 binding capacity was significantly reduced in macrophages from gp91phox−/− mice than in that from wild-type mice (P < 0.01, Fig. 5b). The saturation study was performed using wild-type COS7-WT and transfected COS7-NOX2 cells. The [3H]-PGE2 binding was saturable with a binding affinity at a low nanomolar concentration range (Fig. 5c, d). The calculated Kd value was 0.83 ± 0.04 nM, which is close to the IC50 concentrations for the inhibition of both LPS-induced superoxide production (0.13 nM) and neurotoxicity (0.14 nM). These results give credence to the possibility that PGE2 inhibits LPS-induced superoxide production and subsequent neurotoxicity by binding to the gp91phox subunit.

Fig. 5
figure 5

a Wild-type COS7 cells and COS7 cells stably expressing gp91phox and p22phox, p47phox and p67phox (COS7-NOX2) cells were used to conduct the [3H]-labeled PGE2 binding assay. b Wild-type (gp91phox+/+) and gp91phox-deficient (CYBB) macrophages were used to conduct the [3H]-labeled PGE2 binding assay. Data are shown as a percentage of the wild-type control. Results are presented as means ± SEM from three independent experiments in duplicate. ***P < 0.0001 and **P < 0.01, paired t test. Saturation curves of [3H]-PGE2 binding were conducted using COS7-WT (c) and COS7-NOX2 cells (d). Total binding was measured by the binding of [3H]-labeled PGE2 ranging from 0.03 to 3 nM while non-specific binding was determined in the presence of additional 3 μM of PGE2. Saturation isotherms were analyzed in GraphPad Prism 6. Specific binding was calculated by the difference between total binding and non-specific binding. This experiment was conducted three times with similar results

Full size image

PGE2 Inhibits the LPS-Induced Increase in Translocation of the Cytosolic Subunit p47phox to the Cell Membrane

The activation of NOX2 requires the translocation of phosphorylated cytosolic subunits (p47phox, p67phox, and p40phox) to the cell membrane to form the active enzyme complex by binding to the membrane subunits gp91phox and p22phox. We previously reported that several anti-inflammatory agents reduced the production of superoxide by changing the conformation of the membrane subunit gp91phox to reduce the binding affinity for the cytosolic subunits of NOX2 and prevent their translocation to the membrane [51]. To obtain additional evidence supporting the possibility that NOX2 acts through a novel signaling pathway that mediates the anti-inflammatory action of PGE2, we examined the membrane translocation of the cytosolic subunit p47phox by determining the level of p47phox in cytosol and membrane. Primary microglia were pretreated with/without 10 nM of PGE2 for 30 min following by LPS stimulation for 15 min. Staining results showed that LPS significantly increased the immunofluorescence intensity in the membrane, indicating translocation of the cytosolic subunit p47phox to the membrane. Pretreatment with PGE2 alone did not differ from the control while when combined with LPS, it almost completely prevented the LPS-induced translocation of p47phox to the membrane (Fig. 6a). We also conducted Western blot analysis to verify the result of immunofluorescence staining. In this study, rat microglia cell line HAPI cells were used, due to the large number of high purity cells of this type that can be harvested for Western blot analysis. Consistent with immunostaining results, LPS stimulation significantly increased the band intensity of p47phox in the membrane fraction (control vs. LPS, t = 10.86, P < 0.001). PGE2 pretreatment completely prevented these changes (LPS vs. PGE2 + LPS, t = 10.47, P < 0.001) (Fig. 6b, c). Taken together, these results suggest that PGE2 can bind to the subunit of NOX2 gp91phox to inhibit the LPS-induced translocation of cytosolic p47phox and subsequent superoxide production.

Fig. 6
figure 6

a Primary microglial cells were pretreated with PGE2 or vehicle for 30 min and then stimulated with LPS (15 ng/ml) for 15 min. Cells were immunostained with a rabbit polyclonal antibody against p47phox (green) and DAPI (blue). Scale bar, 10 μm. b, c HAPI cells (rat microglia cell line) were pretreated with PGE2 (10 nM) or vehicle for 30 min, followed by LPS stimulation (15 ng/ml). After LPS stimulation for 15 min, subcellular fractions were isolated to conduct Western blot analysis in order to determine the level of p47phox in the membrane fractions. gp91phox was used as membrane internal control. Values are mean ± SEM from three independent experiments. ***P < 0.0001 Bonferroni’s t test compared to vehicle-treated group; ###P < 0.0001 Bonferroni’s t test compared to LPS-treated group

Full size image

Discussion

The anti-inflammatory effects of PGE2 have been extensively studied under both pharmacological and pathological conditions. However, concentrations of PGE2 used in such studies are in the micromolar range (0.1–10 μM) and are rarely observed under normal physiological conditions. The present study demonstrated that low concentrations of PGE2 (0.01 to 10 nM), which are close to normal brain extracellular concentrations (< 10 nM) [28,29,30,31,32,33], exert potent protection of LPS-induced DA neuron damage via a reduction in the release of pro-inflammatory factors, especially the release of reactive oxygen species (ROS) from microglia. Further, the mechanistic studies conducted here reveal that the anti-inflammatory and neuroprotective effects of low concentrations of PGE2 are not mediated through the conventional EP receptors (EP1-EP4). Instead, we found that PGE2 could directly inhibit the activation of microglial NOX2 by binding to its catalytic membrane subunit gp91phox. Our study also indicates that a reduction in superoxide production plays a major role in PGE2-induced anti-inflammatory neuroprotective effects. These findings revealed a receptor-independent novel mechanism by which physiological concentrations of PGE2 function to maintain brain immune homeostasis.

PGE2 has been reported to exert both pro- and anti-inflammatory effects; however, the reasons behind these opposing effects remain unclear. Use of different concentrations and experimental models could be responsible for some of these bifunctional effects of PGE2. We hypothesized that the difference in concentrations could play a role in these opposite effects of PGE2 on immune function. Therefore, this study focused on elucidating the mechanism(s) mediating the action of PGE2 in low concentrations. The LPS-induced release of pro-inflammatory factors, including TNF-α, MCP-1, nitric oxide, and superoxide-related ROS from microglia, was found to decline in the presence of PGE2 in a dose-dependent manner. However, the IC50 concentrations of PGE2 necessary to suppress these pro-inflammatory factors varied greatly. Among these immune factors, superoxide production was the most critical factor related to the anti-inflammatory and neuroprotective effects of PGE2; 1.0 nM of PGE2 completely suppressed LPS-induced production of superoxide and achieved total neuroprotection. In contrast, higher concentrations of PGE2 were needed to inhibit LPS-induced production of TNF-α, MCP-1, and nitric oxide to exert a neuroprotective effect. These findings prompted us to search in further detail for the mechanism(s) behind inhibition of superoxide production in the presence of low concentrations of PGE2.

The most salient finding of this study is the EP receptor-independent inhibition of superoxide production by low concentrations of PGE2. Most of our knowledge regarding the inhibition of superoxide production by PGE2 arrived from studies on peripheral phagocytes using micromolar concentrations of PGE2 [47,48,49], which indicate that the EP2/cAMP signaling pathway plays a pivotal role in inhibition of superoxide production inhibition by PGE2. In contrast, the lack of suppression of superoxide by either low concentrations of PGE2 or the EP2 agonist butaprost in EP2 receptor-deficient mixed glia cultures, along with a lack of effect of combined pretreatment with EP1-4 receptor antagonists on PGE2-induced inhibition of superoxide, implies the existence of an EP receptor-independent regulatory mechanism in superoxide inhibition by PGE2 in microglia. Moreover, we found that cAMP was undetectable in primary microglial cells treated with 10 nM or less of PGE2 (Supplementary Fig. 2), further supporting the EP receptor-independent effect of PGE2.

Previous studies have shown that NOX2, the main superoxide-producing enzyme on microglia cell membranes, is an important source of ROS [50, 53, 54]. ROS, including the superoxide anion, hydroxyl radical, lipid hydroperoxides, and hydrogen peroxide, are generated from activated microglia and play important roles in amplifying the expression of pro-inflammatory genes in microglia and subsequent neuro-degeneration [53]. Various in vitro and in vivo studies indicate that deficiency in NOX2 or pharmacological inhibition of this enzyme’s activity can significantly reduce the production of pro-inflammatory factors, as well as subsequent neuro-degeneration [50, 55, 56]. Such studies prompted us to speculate that NOX2 could be the target of physiological concentrations of PGE2, and the following evidence strongly supports this hypothesis: (1) all the specific agonists for the different EP receptor subtypes, such as 17-phenyltrinor PGE2 (EP1 and EP3), butaprost (EP2) and CAY10598 (EP4), displayed similar potencies as PGE2 in inhibiting superoxide production (Figs. 2a and 3a–c). In contrast, EP receptor antagonists or lack of an EP2 receptor failed to influence the inhibitory effects of PGE2. These results further indicate that a superoxide-generating enzyme, rather than any of the EP receptors, is a common target for the different PGE2 analogs. (2) The ligand binding studies indicated that PGE2 bound to the membrane catalytic subunit gp91phox with a high affinity (i.e., a Kd value around 10−10 M), which is close to the Ki value for inhibition of superoxide production (Fig. 5c, d). (3) PGE2 failed to reduce superoxide produced by the xanthine/xanthine oxidase system, indicating that the inhibition of superoxide production by PGE2 is not due to its superoxide free radical-scavenging effect (Supplementary Fig. 1). Together, the evidence indicates that NOX2 is a novel target mediating the anti-inflammatory action of nanomolar concentrations of PGE2.

To further understand the nature how PGE2 interacts with the catalytic subunit gp91phox, we conducted a pharmacophore analysis by comparing PGE2 with a series of conventional NOX2 inhibitors. Despite the differences in chemical structures between PGE2 and its analogs used in this study and some of the known NOX2 inhibitors (e.g., diphenyleneiodonium (DPI), apocynin, or naloxone [51]), the analysis showed that several chemical features shared between PGE2, EP receptor agonists, and NOX2 inhibitors, as deduced by the program Molecular Operating Environment (MOE, Chemical Computing Group, Inc., Montreal, Canada), include hydrophobic centroids, hydrogen bond acceptors, and hydrogen bond donors (Supplementary Fig. 3). Thus, it is likely that PGE2 and EP receptor agonists might share similar binding properties to microglial NOX2 as DPI and naloxone, hence eliciting similar biological effects.

Although we have identified microglial gp91phox as a novel binding site for PGE2 in eliciting its anti-inflammatory effect, the underlying mechanism(s) into how PGE2 inhibits NOX2 activation remains to be determined. Our previous study indicated that after binding, the NOX2 inhibitor, may alter the conformation of gp91phox and hinder the binding of cytosolic subunits, including p67phox and p47phox, thus preventing formation of an active enzyme complex [51]. Since the crystal structure of gp91phox is still unavailable, we were not able to demonstrate this actual conformational change after treatment with PGE2. Thus, we used an alternative approach by determining the translocation of cytosolic subunits to the cell membrane and demonstrated that PGE2 reduces the translocation of p47phox to the cell membrane (Fig. 6).

In summary, this study demonstrated that physiological concentrations of PGE2 elicited potent anti-inflammatory/neuroprotective effects. These actions are mediated through a novel EP receptor-independent mechanism by inhibiting activation of microglial NOX2 via direct binding to its catalytic subunit gp91phox. These results provide new insights into the potential physiological function of PGE2 in maintaining CNS immune homeostasis.