Prostaglandin E₂ Receptor Subtype 2 Regulation of Scavenger Receptor CD36 Modulates Microglial Aβ₄₂ Phagocytosis

Abstract

Recent studies underline the potential relevance of microglial innate immune activation in Alzheimer disease. Primary mouse microglia that lack prostaglandin E2 receptor subtype 2 (EP2) show decreased innate immune-mediated neurotoxicity and increased amyloid β (Aβ) peptide phagocytosis, features that were replicated in vivo. Here, we tested the hypothesis that scavenger receptor CD36 is an effector of EP2-regulated Aβ phagocytosis. CD36 expression was 143-fold greater in mouse primary microglia than in primary astrocytes. Three different means of suppressing EP2 signaling increased and an agonist of EP2 decreased CD36 expression in primary wild-type microglia. Activation of Toll-like receptor (TLR) 3, TLR4, and TLR7, but not TLR2 or TLR9, reduced primary microglial CD36 transcription and cell surface CD36 protein and reduced Aβ₄₂ phagocytosis as well. At each step, the effects of innate immune activation on CD36 were reversed by at least 50% by an EP2 antagonist, and this partial rescue of microglia Aβ₄₂ phagocytosis was largely mediated by CD36 activity. Finally, we showed in hippocampus of wild-type mice that innate immune activation suppressed CD36 expression by an EP2-dependent mechanism. Taken together with results of others that found brain clearance of Aβ peptides and behavioral improvements mediated by CD36 in mice, regulation of CD36-mediated Aβ phagocytosis by suppression of EP2 signaling may provide a new approach to suppressing some aspects of Alzheimer disease pathogenesis.

Observational studies in patients and experimental studies that use a variety of model systems have concluded that affected regions of the brain in Alzheimer disease (AD) experience a proinflammatory, pro-oxidative state.¹ Indeed, recent studies have highlighted several genes involved in the innate immune response whose variants are associated with increased risk of AD, especially the TREM2 gene.^2–4 TREM2 encodes the triggering receptor expressed on myeloid cells-2, a protein expressed by myeloid lineage cells, including microglia in brain, which can function as a receptor for phagocytosis,⁵ modulate the innate immune response in brain,^6,7 and diminish immune response through several signaling pathways, including suppression of Toll-like receptor (TLR) signaling.^8,9

The prostaglandin (PG) pathway, a cascade initiated by enzyme-catalyzed synthesis of PGH₂ from arachidonic acid by the cyclooxygenases (COXs), COX-1 or COX-2, is an essential component of the innate immune response. Multiple studies have reproducibly shown that nonselective COX or COX-2–selective inhibition by nonsteroidal anti-inflammatory drugs (NSAIDs) can suppress amyloid β (Aβ) accumulation and associated behavioral changes in transgenic mouse models of AD¹⁰ and can reduce the risk of AD dementia in people by approximately 60%.¹¹ Together these results suggest that NSAIDs may effectively suppress some aspect of what is now appreciated to be early steps in AD pathogenesis.¹² These compelling experimental model and human data formed the rationale for multiple clinical trials. Two clinical trials targeted what is now understood to be relatively advanced AD: one with NSAIDs in AD dementia and one with NSAIDs in mild cognitive impairment, a diagnostic group enriched for prodromal AD; both failed.^13,14 A third trial targeted early-stage AD,¹⁵ similar to the epidemiologic cohorts, but was terminated because of toxicity, likely related to perturbation of prostacyclin and thromboxane metabolism.^16–18 Despite these disappointing therapeutic outcomes, some data suggest pharmacodynamic effects of NSAIDs in the central nervous system of patients with AD.^19,20

We and others have sought to disentangle the potentially beneficial effects of NSAIDs in early AD pathogenesis from toxic effects by focusing on specific PG receptors distal in the signaling cascade rather than proximal COX inhibition.^21–23 PGE₂ is the principal proinflammatory PG, and increased PGE₂ concentrations in cerebrospinal fluid have been found in patients with AD,²⁴ particularly early in the disease process.²⁵ In addition, the inducible form of PGE₂ synthase is increased in AD brain and is a key component in Aβ-mediated neurotoxicity in a mouse model of AD.²⁶ We, therefore, have focused on PGE₂ and its functionally antagonistic receptor subtypes, EP1 to EP4, which are widely expressed in central nervous system cells, including microglia, astrocytes, and neurons.²¹ Primary microglia from mice homozygous deficient for the EP2 gene (PTGER2; EP2^−/−) have the highly desirable dual phenotype of suppressed TLR-activated microglia-mediated neurotoxicity that depends on the adaptor protein DOCK2 and enhanced microglial phagocytosis of Aβ that is DOCK2 independent.^27–29 These studies show that microglial EP2 is necessary for lipopolysaccharide-induced microglia-mediated neurotoxicity in microglia–neuron cocultures through regulation of inducible nitric oxide synthase and COX-2 and subsequent neurotoxic cytokine production, including IL-1β production,²⁷ and indicate EP2-dependent neurotoxicity in EP2^−/− microglia, with reduced paracrine neurotoxicity in response to aggregated Aβ in neuron–microglia cocultures.²⁸ Recently, these findings were confirmed in mice with conditional deletion of EP2 in myeloid lineage cells and in bone marrow transplant experiments.^30,31

Microglial phagocytosis is mediated largely by a class of surface proteins called scavenger receptors. CD36, a type B scavenger receptor originally identified as a receptor for long-chain fatty acids and oxidized low-density lipoprotein, is a microglial binding site for Aβ,³² facilitates intracellular nucleation of Aβ peptides into fibrils,³³ and mediates some of the biological effects of Aβ peptides, including innate immune activation and oxidative stress.^34–36 CD36 is expressed on microglia in AD brains.³⁴ CD36 overexpression in human brain correlates with β-amyloid deposition but not with AD; it is undetectable in healthy brains without Aβ deposition.³⁷ We are unaware of any data to link the specific PG receptors with CD36 function. Here, we tested the hypothesis that EP2 suppression of Aβ phagocytosis is mediated, at least in part, by CD36.

Materials and Methods

Materials

Dulbecco’s modified Eagle’s medium/F-12 medium and fetal bovine serum were purchased from Hyclone Laboratories (Logan, UT). Papain and DNase I were from Worthington Biochemical (Lakewood, NJ). Butaprost, NS-398, 17-phenyl trinor PGE₂, CAY10598, ZK118182, 17-phenyl trinor PGF_2α, carbaprostacyclin, U-46619, AH 6809, and CD36 blocking antibody (Clone JC63.1) were from Cayman Chemical Company (Ann Arbor, MI). Dibutyryl cAMP, forskolin, tribromoethanol (Avertin), and polyinosinic-polycytidylic acid (PIC; TLR3 ligand) were from Sigma-Aldrich (St. Louis, MO). Pam3CSK4 (TLR2 ligand), loxoribine (TLR7 ligand), and CpG (TLR9 ligand) were from Invivogen (San Diego, CA). Lipopolysaccharide (TLR4 ligand) was purchased from Calbiochem (La Jolla, CA). pHrodo fluorogenic dye was from Invitrogen (Carlsbad, CA). β Amyloid 1-42 (Aβ₄₂) was from American Peptide Company (Sunnyvale, CA).

Animals

C57BL/6 mice and CD36^−/− (B6.129S1-Cd36tm1Mfe/J) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). EP2^−/− mice were a gift from Dr. Richard Breyer (Vanderbilt University, Nashville, TN). The animals were maintained in a specific pathogen-free environment. All procedures were approved by the University of Washington Institutional Animal Care and Use Committee.

Stereotactic Intracerebroventricular Injection

Ten-week-old male mice were anesthetized with isoflurane inhalation and placed in a stereotactic instrument. After the skull was exposed through skin incision, a cranial window (2 to 3 mm) on the left lateral ventricle was opened by drilling with a small burr bit. Five microliter of reagent was injected over a period of 5 minutes by using a Hamilton microsyringe; injections were 200 ng of AH 6809, 50 ng of PIC, AH 6809 plus PIC, or vehicle. After injection, the mice were placed on an isothermal pad and continuously observed until recovery. Twenty-four hours after injection, the mice were sacrificed by i.p. injection of tribromoethanol and perfused with ice-cold phosphate-buffered saline, and brain tissues were rapidly harvested for analysis.

Cell Culture and Exposure

Primary microglia and astrocytes were isolated from brains of newborn mice and cultured as previously described.^38,39 Medium concentrations of PGE₂ were measured by ELISA exactly as described previously.³⁸ Cells were incubated with the compounds described at the indicated concentrations and times. We note that, although EP antagonist AH 6809 has nearly equal affinity for EP1 and DP1 in human tissue,⁴⁰ it has been shown to be selective for EP2 in mouse tissue.⁴¹ EP agonist butaprost was used at 20 μmol/L, based on previous studies that showed effects up to 100 μmol/L.^42,43

Quantitative Real-Time PCR

Total RNA was extracted from primary microglia, astrocytes, or brain tissues by using RNeasy Mini Kit (Qiagen, Valencia, CA). Reverse transcription was performed with Omniscript RT Kit (Qiagen). TaqMan probes and primers (Mm01135198_m1 for CD36, fMm99999915_g1 for GAPDH) were purchased from Applied Biosystems (Carlsbad, CA). PCR were performed on an Applied Biosystems ViiA 7 Real-Time PCR System by using the relative quantitative method.

Flow Cytometric Analysis of Cell Surface CD36 in Primary Microglia

Microglia were preincubated with 10 μg/mL of anti-mouse CD16/CD32 antibody (BD Biosciences, San Jose, CA) for 5 minutes on ice to block Fc receptor–mediated binding. The cells were then incubated with 5 μg/mL of phycoerythrin-conjugated CD36 or mouse IgA control (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) for 30 minutes on ice in the dark. Thereafter, microglia were washed twice with phosphate-buffered saline and resuspended in 4% paraformaldehyde for fixation. The fluorescence intensity was determined by BD FACScan flow cytometry.

Aβ₄₂ Phagocytosis in Primary Microglia

Microglia were incubated with 5 μmol/L pHrodo-labeled Aβ₄₂ for 2 hours at 37°C. The nonspecific signal was determined by incubation of microglia with the same concentration of labeled Aβ₄₂ for 2 hours at 4°C. Thereafter, microglia were washed three times in phosphate-buffered saline and fixed in 4% paraformaldehyde. The samples were analyzed by LSRII flow cytometry (BD Biosciences) with a 561-nm excitation laser.

CD36 Expression in CHO Cells

Human CD36 was expressed in Chinese hamster ovary (CHO) cells as described by others.⁴⁴ Briefly, cells were maintained in Ham's F-12 medium that contained 10% fetal bovine serum and penicillin/streptomycin. Human full-length CD36 cDNA was obtained from GE Healthcare (Lafayette, CO) and cloned into the expression vector pIRES2-EGFP at EcoRI and XbaI sites. The recombinant vector or empty vector was transfected into CHO cells by Lipofectamine 2000 according to the manufacturer's protocol. The expression of CD36 was confirmed by quantitative PCR with Taqman expression assay. Two days after transfection, cells were collected for phagocytosis assay and inhibition with sulfosuccinimidyl oleate (SSO) as described by others.⁴⁵

Statistical Analysis

Statistical analyses were performed with GraphPad Prism 5 (GraphPad Inc., San Diego, CA) with α set to 0.05. All P values presented in the text or figures were corrected for multiple comparisons when appropriate with the method of Bonferroni.

Results

Microglia Express Much More CD36 than Astrocytes

Microglia, and less so astrocytes, are cellular effectors of innate immune response in the cerebral cortex. CD36 expression in primary wild-type (WT) cultures of microglia and astrocytes prepared from the same five mice was determined by quantitative real-time PCR with each sample normalized to endogenous control GAPDH. CD36 expression in microglia was on average 143-fold greater in microglial than in astrocytes (P < 0.0001). For this reason, subsequent experiments focused on microglia.

Down-Regulation of CD36 Expression by Specific TLR Activation

Because many endogenous activators of innate immune response, including Aβ,^46–49 act through TLRs and because one function of TREM2 is to regulate TLR signaling,⁸ we measured primary WT mouse microglia CD36 expression after exposure for 18 hours to specific TLR activators that are known to activate microglia.³⁸ Activators of TLR3, TLR4, and TLR7, but not TLR2 or TLR9, suppressed microglia expression of CD36 (Figure 1A). A seven-point time series (Figure 1B) showed no change in CD36 expression after 2 hours of PIC exposure (1.03 ± 0.05 CD36 mRNA compared with time 0), but after 4 hours of exposure CD36 mRNA was reduced to 0.74 ± 0.03 and continued to decrease linearly to 0.16 ± 0.03 at 18 hours (R² = 0.95, P < 0.0001). A similar time course for medium PGE₂ was approximately complementary with 23 ± 2 pg/mL at time 0 and increased to 2199 ± 64 pg/mL at 18 hours (P < 0.0001). Other scavenger receptors also were altered by PIC incubation for 18 hours: microglial expression of SR-A1 was increased 3.1- ± 0.7-fold (P < 0.01) and SR-B1 was decreased 0.25- ± 0.06-fold (P < 0.001) compared with vehicle-exposed cultures.

Figure 1.

Effects of TLR activators on CD36 mRNA expression. A: WT mouse primary microglia were treated with Veh or 1 μg/mL activator of TLR2 (Pam3CSK4), 20 μg/mL TLR3 (PIC), 1 μg/mL TLR4 (lipopolysaccharide), 1 mmol/L TLR7 (loxoribine), or 1 μmol/L TLR9 (CpG) for 18 hours. RNA was extracted for quantitative real-time PCR. CD36 mRNA values in each sample were normalized to endogenous control GAPDH. B: WT mouse primary microglia were treated with 20 μg/mL PIC for the indicated times. CD36 mRNA was assessed by quantitative real-time PCR, and results were normalized to GAPDH. Medium PGE₂ was assayed by ELISA. Data are expressed as means ± SEM. n = 3 per group. ^∗∗∗P < 0.001 for Bonferroni-corrected repeated pair comparisons with vehicle (A) or with value at time 0 (B). PGE2, prostaglandin E2; PIC, polyinosinic-polycytidylic acid; TLR, Toll-like receptor; Veh, vehicle; WT, wild-type.

Regulation of CD36 Expression by EP2

We next surveyed specific agonists to prostanoid receptors EP, DP, FP, IP, and TP to determine which, if any, specific product of COX might contribute to innate immune-mediated reduction in CD36 expression. Agonists of EP2 or EP4 mimicked the effect of innate immune activation by significantly reducing CD36 expression in primary WT mouse microglia, whereas agonists of EP1 or TP had a functionally antagonistic effect; agonists of FP, IP, and DP had no significant effect on CD36 at known effective concentrations (Figure 2A).³⁹ In contrast, a COX-2–selective inhibitor, an EP2 antagonist, and EP2^−/− microglia had progressively increasing microglial CD36 expression. Activation of the innate immune response with agents specific for TLR3, TLR4, or TLR7 each reproducibly decreased microglial CD36 expression, an effect that was mostly reversed for each TLR activator with an EP2 antagonist (AH 6809) (Figure 2B); although reversal by AH 6809 was greatest in the context of TLR3 activation. A selective antagonist of EP4 is not commercially available.^50,51 EP2 and EP4 are stimulatory G-protein-couple receptors, but because EP2 also can activate G-protein–independent second messenger signaling,⁵² we tested the effect of directly manipulating cAMP signaling on CD36 expression; dibutyryl cAMP or forskolin decreased CD36 mRNA values by the same magnitude as butaprost alone (Figure 2C). These data show that multiple eicosanoid products of COX exert tonic control over CD36 expression in mouse primary microglia and that among these EP2 or EP4 activation suppresses CD36 expression to an amount matched by direct activation of cAMP second messenger signaling. Finally, results by using an antagonist show that EP2 signaling is responsible for most reduced microglial CD36 expression with TLR-activated innate immune response.

Figure 2.

Effects of eicosanoid receptors with and without innate immune activation on CD36 mRNA expression by mouse primary microglia. RNA was isolated from cultures prepared from wild-type mice (dark bars) or EP2^−/− mice (light bar) after 18 hours of incubation. Results from quantitative real-time PCR were normalized to endogenous control GAPDH. A: Microglia exposed to vehicle or 5 μmol/L agonists of EP1 (17-phenyl trinor PGE₂), 20 μmol/L EP2 (butaprost), 2.5 μmol/L EP4 (CAY10598), 10 μmol/L DP (ZK118182), 10 μmol/L FP (17-phenyl trinor PGF_2α), 10 μmol/L IP (carbaprostacyclin), or 5 μmol/L TP (U-46619). B: Microglia exposed to vehicle or 20 μmol/L COX-2–selective inhibitor (NS398), 20 μmol/L EP2 antagonist AH, and TLR activators (PIC for TLR3 at 20 μg/mL, LPS for TLR4 at 1 μg/mL, or loxoribine for TLR7 at 0.5 mmol/L) with or without 20 μmol/L AH. C: Microglia exposed to 20 μmol/L butaprost, 250 μmol/L db cAMP, or 10 μmol/L forskolin. Data are expressed as means ± SEM. n = 3 to 12 per group. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001 for Bonferroni-corrected repeated pair comparisons with vehicle; ^†P < 0.05 and ^††P < 0.001 for Bonferroni-corrected paired comparisons between two experimental groups. AH, AH 6809; COX-2, cyclooxygenase 2; db cAMP, dibutyryl cAMP; EP2^−/−, homozygous EP2 deficient; LPS, lipopolysaccharide; PGE₂, prostaglandin E2; PIC, polyinosinic-polycytidylic acid; TLR, Toll-like receptor; Veh, vehicle.

CD36 expression is transcriptionally regulated, but we nevertheless sought to validate the changes observed in microglia CD36 mRNA expression with changes in CD36 protein. Flow cytometric analysis of primary WT mouse microglia labeled with phycoerythrin-conjugated anti-CD36 antibody; 24 hours of exposure to butaprost, an EP2 agonist, significantly reduced CD36 cell surface protein to values that were proportionate to mRNA reduction (Figure 3). Conversely, exposure to AH 6809 increased CD36 cell surface protein values. These results are consistent with the known transcriptional regulation of CD36 protein and indicate that changes in microglia cell surface CD36 protein follow the same pattern as CD36 mRNA in response to an agonist or antagonist of EP2.

Figure 3.

Effects of TLR3 activator and prostaglandin EP2 signaling on CD36 protein. WT microglia were treated with 20 μmol/L butaprost, 20 μmol/L AH, 20 μg/mL PIC, or a combination of PIC and AH for 24 hours. CD36 protein on cell surface was assessed by staining with PE-labeled anti-CD36 antibody or isotype-matched PE-IgA (isotype control). A: Representative histogram from flow cytometric assay of CD36 immunofluorescence. B: Geometric means of fluorescence intensity was normalized to vehicle. Data are expressed as geometric means of fluorescence intensity ± SEMs. n ≥ 3 per group. ^∗∗P < 0.01 and ^∗∗∗P < 0.001 for Bonferroni-corrected repeated paired comparisons with vehicle; ^†P < 0.05 and ^††P < 0.001 for Bonferroni-corrected paired comparisons between two experimental groups. AH, AH 6809; Buta, butaprost; CT, control; EP2, E2 receptor subtype 2; PE, phycoerythrin; PIC, polyinosinic-polycytidylic acid; TLR, Toll-like receptor; WT, wild-type.

Inhibition of CD36-Dependent Phagocytosis by EP2 Activation

Our next series of experiments used primary WT and CD36^−/− mouse microglia to investigate whether CD36 mediated EP2 effects on Aβ₄₂ phagocytosis. We quantified phagocytosis of Aβ₄₂ labeled with a pH-sensitive dye (pHrodo) and under the same conditions that affected CD36 mRNA and cell surface protein values (shown in Figures 2 and 3). Butaprost significantly decreased Aβ₄₂ phagocytosis in WT microglia, validating previous experiments from our laboratory that used EP2^−/− microglia with synthetic Aβ₄₂ and human brain-derived Aβ peptides (Figure 4A).^28,30,53 TLR3 activation from PIC exposure significantly reduced Aβ₄₂ phagocytosis by WT microglia, and this immune suppression was significantly but partially reversed by AH 6809, paralleling the effect of PIC on CD36 expression and indicating that EP2 activation accounted for approximately one-half of PIC-suppressed Aβ₄₂ phagocytosis. CD36 inhibitor SSO reversed this rescue, consistent with CD36-dependent phagocytosis. Because SSO may have actions in addition to inhibiting CD36, we sought to validate this finding with three different experiments. In the first, we used AH 6809 (coincubated with control IgA) to achieve increased Aβ₄₂ phagocytosis (P < 0.01) that paralleled increased CD36 expression (Figure 4B); this prophagocytic effect of AH 6809 was completely blocked by a CD36 neutralizing antibody. Consistent with this, CD36 blocking antibodies did not further inhibit butaprost-treated cells. In the second set of experiments, we compared phagocytosis in microglia prepared from WT and CD36^−/− mice (Figure 4C). CD36^−/− microglia had reduced Aβ₄₂ phagocytosis compared with WT microglia (P < 0.001). Importantly, increased Aβ₄₂ phagocytosis seen with exposure to AH 6809 was ablated in CD36^−/− microglia, whereas butaprost had no significant effect on Aβ₄₂ phagocytosis in these cells. In combination, results from WT microglia with the use of SSO or a CD36-neutralizing antibody and from CD36^−/− microglia support the conclusion that the effects of EP2 activation on Aβ₄₂ phagocytosis are largely mediated by CD36. Finally, we transfected CHO cells with human CD36 as described by others.⁴⁴ Vector-transfected CHO cell uptake of Aβ₄₂ was defined as 100%. Expression of human CD36 significantly increased uptake of Aβ₄₂ by CHO cells to 182% ± 6% of vehicle-transfected control (P < 0.001), similar to what has been reported previously by others.⁴⁴ Also validating the results of others,⁴⁵ exposure to SSO by using the same method as described above for mouse primary microglia completely inhibited Aβ₄₂ uptake by human CD36-expressing CHO cells to 102% ± 9% of vehicle-transfected control; Aβ₄₂ uptake by vehicle-transfected CHO cells was not significantly altered by SSO (94% ± 8% of vector-transfected cells without SSO).

Figure 4.

Regulation of mouse primary microglia phagocytosis of Aβ₄₂ by prostaglandin EP2 receptor signaling. After exposure to agents, phagocytosis was determined by incubation with pHrodo-labeled Aβ₄₂ for 2 hours at 37°C. Fluorescence intensities were determined by flow cytometry, and geometric means of fluorescence intensity were normalized to vehicle-exposed WT cells. A: WT microglia were pretreated with 20 μmol/L EP2 agonist butaprost, 20 μg/mL TLR3 activator PIC, or 20 μg/mL PIC plus 20 μmol/L EP2 antagonist AH 6809 for 24 hours and then incubated with pHrodo-labeled Aβ₄₂. Cultures treated with PIC + AH 6809 were also incubated with 500 μmol/L CD36 inhibitor, SSO, for an additional hour before phagocytosis assay. B: WT microglia were treated with 20 μmol/L AH 6809 or 20 μmol/L agonist butaprost for 24 hours, followed by incubation with 10 μg/mL anti-CD36 IgA or mouse control IgA for 1 hour, and then phagocytosis was determined. C: Primary microglia from WT or CD36^−/− mice were treated with 20 μmol/L EP2 antagonist AH 6809 or 20 μmol/L agonist butaprost for 24 hours, and then phagocytosis was determined. Data are expressed as means ± SEM. n = 3 per group (A and C); n = 3 or 4 per group (B). ^∗P < 0.05, ^∗∗P < 0.01, ^∗∗∗P < 0.001, and ^∗∗∗∗P < 0.0001 for Bonferroni-corrected repeated paired comparisons with vehicle; ^†P < 0.05 and ^††P < 0.01 for Bonferroni-corrected paired comparisons between two experimental groups. Aβ₄₂, β-amyloid 1-42; EP2, E2 receptor subtype 2; PIC, polyinosinic-polycytidylic acid; SSO, sulfosuccinimidyl oleate; TLR, Toll-like receptor; WT, wild-type.

Reduced CD36 Expression by PIC Largely Depends on EP2 in Mouse Brain

Finally, we investigated whether EP2 regulated CD36 expression in a relevant region of WT mouse brain. Hippocampal CD36 mRNA values were quantified 24 hours after 50 ng intracerebroventricular injection of PIC into the left lateral ventricle with or without two different doses of the EP2 antagonist AH 6809 (200 or 50 ng) (Figure 5). Our results showed that direct exposure to TLR3 activator PIC significantly reduced ipsilateral hippocampal CD36 expression to one-half of vehicle (0.51 ± 0.09; P < 0.05 by Bonferroni-corrected posttest) and that this was reversed by simultaneous exposure to AH 6809. Indeed, two-way analysis of variance had P < 0.01 for an interaction term between PIC and AH 6809, and Bonferroni-corrected posttests were not significant for comparison of PIC with PIC plus 200 ng of AH 6809. Similar results were achieved when using 50 ng (0.86 ± 0.21) rather than 200 ng of AH 6809 (0.81 ± 0.22).

Figure 5.

Modulation of CD36 expression in mouse hippocampus by PIC and prostaglandin E2 receptor subtype 2 antagonist AH 6809. Wild-type mice received intracerebroventricular injection into the left lateral ventricle with vehicle, 50 ng PIC, 200 ng AH 6809, or 50 ng PIC plus 200 ng AH 6809. Ipsilateral hippocampus was isolated after 24 hours, mRNA values of CD36 were determined by quantitative real-time PCR, and results were normalized to endogenous control GAPDH. Data are expressed as means ± SEM. n = 6 mice per group. Two-way analysis of variance (df 1,1,1, 20) had ^∗∗P < 0.01 for PIC and interaction between PIC and AH 6809 but was insignificant for AH 6809 alone. Bonferroni-corrected posttests were significant only for the pair with no AH 6809 (^∗∗P < 0.01) but were insignificant for the pair with AH 6809. PIC, polyinosinic-polycytidylic acid; Veh, vehicle.

Discussion

Considerable data from observational studies of human brain and experimental models support the hypothesis that microglial innate immune activation is a feature and possible contributor to AD pathogenesis, a hypothesis recently reinforced by genetic association of AD dementia with several genes that regulate innate immunity, especially TREM2. Results from these studies are further buttressed by compelling epidemiologic and experimental model data that support inhibition of the PG cascade as an effective therapeutic approach early in the pathogenesis of AD before clinical expression or behavioral impairments in transgenic mouse models. Enthusiasm over these observational and experimental data are tempered by failed trials that used NSAIDs to suppress innate immunity in subjects with mild cognitive impairment or AD dementia and the termination of an AD prevention trial because of toxicity. Although the clinical trial results clearly are disappointing, they do not necessarily indict the innate immunity hypothesis for AD but rather may reflect limitations of intervening too late in the course of AD and/or therapeutic shortcomings of proximal inhibition of the PG cascade in people. Together, these results cautiously animate pursuit of components distal in the PG cascade as a potentially more effective means of suppressing microglial innate immunity while averting unwanted side effects.

Previously, with the use of primary mouse microglia, we have shown that suppression of EP2 signaling leads to a therapeutically desirable dual phenotype of reduced innate immune-mediated neurotoxicity and enhanced Aβ phagocytosis, including neuritic plaque material.²⁸ Subsequently, we and others have validated these findings in macrophages and in vivo in mouse brain by using homozygous deficient mice, conditional deletion mice, and bone marrow transplant mice. Although we have made some progress in understanding the molecular mechanisms of EP2 regulation of microglia-mediated neurotoxicity, the mechanisms by which EP2 signaling regulates Aβ₄₂ phagocytosis remained unclear. Here, we showed that scavenger receptor CD36, a key regulator of microglia Aβ₄₂ phagocytosis, is closely regulated by EP2 signaling in primary WT mouse microglia both without further activation or in the context of TLR activators. We focused our work on primary microglia because they express the majority of CD36 among phagocytosis-competent cells in the central nervous system. Primary WT microglial expression of CD36 was increased by an EP2 antagonist and decreased by an EP2 agonist; the latter was associated with proportionately reduced CD36 cell surface protein and Aβ₄₂ phagocytosis. We showed that specific TLR activators led to decreased microglia CD36 transcription and that this translated into reduced cell surface CD36 protein and reduced Aβ₄₂ phagocytosis. At each step in this process, we showed that the effects of innate immune activation on CD36 could be reversed by at least half by suppressing EP2 signaling and that the effects of EP2 activation on Aβ₄₂ phagocytosis are largely mediated by CD36. Together, our results from primary microglia cultures indicated tonic control of CD36 expression by EP2 signaling, and that most innate immune-mediated suppression of Aβ₄₂ phagocytosis depended on EP2, and that this EP2 effect was CD36 dependent.

Current data support reduced clearance as the dominant mechanism for Aβ accumulation in the most common forms of AD (reviewed in Wildsmith et al⁵⁴). Although Aβ may be transported out of brain, it also may be cleared by microglial phagocytosis. Indeed, several lines of evidence point to Aβ phagocytosis as a potentially important process in AD. These include the increased genetic risk of AD with TREM2 variants, reduced expression of the phagocytosis-related protein beclin 1 in diseased brain regions and isolated microglia from patients with AD,^55,56 and reduced CD36 expression on leukocytes in patients with AD dementia or mild cognitive impairment.⁵⁷ It is important to note that, although we have focused on its phagocytic function, the CD36/Aβ complex is pleiotropic and interacts with several intracellular molecules, such as p130Cas, Pyk2, and Paxillin, and initiates signaling pathways that ultimately lead to F-actin polymerization and cytoskeleton reorganization.⁵⁸

Our results from mouse microglia confirmed those from other cell types that have shown decreased expression of CD36 after TLR3 or TLR4 activation.^59,60 We expanded this survey to include TLR7, activation of which by an miRNA can propagate neurodegeneration.⁶¹ Our data also showed that the innate immune-mediated decrease in CD36 was mostly, but not fully, reversed by an EP2 antagonist, suggesting that some other proinflammatory factor(s) also inhibit microglial CD36 expression. One possibility is tumor necrosis factor-α,^62,63 although this is complicated because tumor necrosis factor-α secretion depends in part on EP2.³⁸ Our data also suggest EP4 as another possibility. However, we were limited in pursuing this further because of lack of a selective EP4 antagonist. In contrast to innate immune activation effects on CD36 expression, a variety of anti-inflammatory mediators increase CD36 expression, including IL-4, IL-13, and peroxisome proliferator-activated receptor γ agonist.^64–67 Blocking CD36 with an antibody reduces cytokine and nitric oxide when microglia are incubated with PrP protein.⁶⁸ Although difficult to classify broadly as either proinflammatory or anti-inflammatory, we found that an agonist of EP1 or TP also increased CD36 expression in microglia. Functional antagonism is observed commonly among eicosanoid receptors.²¹ Our results point to a dominant effect of EP2-mediated suppression of CD36 expression in the context of microglial innate immune activation, perhaps related to its higher binding affinity for PGE₂ than for EP1.²¹

Our in vivo experiments showed that CD36 expression in WT mouse hippocampus followed the same mechanisms as microglia. Specifically, activation of innate immune response with PIC significantly decreased CD36 expression, although the magnitude was somewhat less than that observed in primary microglia. The reasons for this difference in magnitude of effect by PIC in vivo versus in vitro are not clear from our experiments but may be related to the influence of other cell types and clearance of the inflammagen. Regardless, our results with a highly selective activator of TLR3 are consistent with those from AD transgenic mice that also observed reduced CD36 expression in the brain,⁶³ a finding confounded by the multiple actions of Aβ peptides. Most importantly, coadministration of AH 6809 largely reversed the effect of PIC, so much so that there was no significant difference between hippocampal CD36 mRNA values in the PIC alone group versus the PIC plus AH 6809 group.

We did not attempt to duplicate behavioral experiments in mice because several other laboratories already have investigated CD36 in the context of mouse behavior and models of immune activation or AD. Indeed, CD36-deficient mice show significant learning impairment as assessed by radial arm maze.⁶⁹ In contrast, increased CD36 expression by peroxisome proliferator-activated receptor γ/retinoid X receptor α agonists increased microglial Aβ phagocytosis and improved cognitive performance in an AD transgenic mouse.⁶⁷ Similarly, increased CD36 expression after brain microinjection of IL-4/IL-13 into another AD transgenic mouse was associated with increased Aβ clearance and improved cognitive performance.⁶⁶

Conclusions

In summary, our data are consistent with an innate immune hypothesis for AD in which age, inherited deficits, or environmental factors impair the clearance of Aβ peptides either by reduced transport and/or by reduced phagocytosis. This promotes accumulation of Aβ peptides and other debris that function as endogenous ligands of TLRs and other pattern recognition receptors that activate an innate immune response; this in turn further lowers phagocytic activity. Our results are consistent with others in supporting a key role for CD36 in Aβ phagocytosis by microglia and highlight EP2 antagonists as an effective means to suppress this toxic reinforcing cycle. Combined with results of others, our data suggest that EP2 antagonists have the potential as a therapeutic approach to suppressing key aspects of AD pathogenesis.