Prostaglandin D2 mediates neuronal protection via the DP1 receptor


Address correspondence and reprint requests to K. Andreasson, 600 N. Wolfe Street, Meyer 5-119B, Baltimore, MD 21205, USA. E-mail:


Cyclo-oxygenases (COXs) catalyze the first committed step in the synthesis of the prostaglandins PGE2, PGD2, PGF, PGI2 and thomboxane A2. Expression and enzymatic activity of COX-2, the inducible isoform of COX, are observed in several neurological diseases and result in significant neuronal injury. The neurotoxic effect of COX-2 is believed to occur through downstream effects of its prostaglandin products. In this study, we examined the function of PGD2 and its two receptors DP1 and chemoattractant receptor-homologous molecule expressed on Th2 cells (CRTH2) (DP2) in neuronal survival. PGD2 is the most abundant prostaglandin in brain and regulates sleep, temperature and nociception. It signals through two distinct G protein-coupled receptors, DP1 and DP2, that have opposing effects on cyclic AMP (cAMP) production. Physiological concentrations of PGD2 potently and unexpectedly rescued neurons in paradigms of glutamate toxicity in cultured hippocampal neurons and organotypic slices. This effect was mimicked by the DP1-selective agonist BW245C but not by the PGD2 metabolite 15d-PGJ2, suggesting that neuroprotection was mediated by the DP1 receptor. Conversely, activation of the DP2 receptor promoted neuronal loss. The protein kinase A inhibitors H89 and KT5720 reversed the protective effect of PGD2, indicating that PGD2-mediated neuroprotection was dependent on cAMP signaling. These studies indicate that activation of the PGD2 DP1 receptor protects against excitotoxic injury in a cAMP-dependent manner, consistent with recent studies of PGE2 receptors that also suggest a neuroprotective effect of prostaglandin receptors. Taken together, these data support an emerging and paradoxical neuroprotective role of prostaglandins in the CNS.

Abbreviations used

cyclic AMP




13,14-dihydro-15-keto-prostaglandin D2


15-deoxy-Δ12,14-prostaglandin J2


lactate dehydrogenase


phosphate-buffered saline




hematopoietic prostaglandin D synthase


lipocalin-type prostaglandin D synthase


propidium iodide


protein kinase A

Cyclo-oxygenase (COX)-2, the inducible isoform of COX, catalyzes the formation of prostaglandin (PG)H2 from arachidonic acid, the substrate for the formation of prostaglandins (Smith et al. 1991). Basal COX-2 expression in neurons is tightly regulated by normal excitatory synaptic activity (Yamagata et al. 1993) and is believed to play a physiological role in synaptic plasticity (Chen et al. 2002). However, levels and enzymatic activity of COX-2 are markedly induced and lead to neuronal injury in stroke (Nogawa et al. 1997; Nakayama et al. 1998; Govoni et al. 2001; Dore et al. 2003), neurodegenerative diseases such as Parkinson's disease (Feng et al. 2002; Klivenyi et al. 2003; Teismann et al. 2003) and amyotrophic lateral sclerosis (Drachman et al. 2002; Pompl et al. 2003; Klivenyi et al. 2004), and possibly in Alzheimer's disease (Stewart et al. 1997; in t'Veld et al. 2001). Thus, increased enzymatic activity of COX-2 represents a common mechanism of neurotoxicity that underlies a wide range of acute and chronic neurological diseases.

The mechanism by which COX-2 activity promotes neurotoxicity is presumed to involve effects of its downstream prostaglandin products PGE2, PGD2, PGF, PGI2 (prostacylin) and thromboxane A2 that effect cellular changes through activation of specific prostaglandin receptors and second messenger systems. These prostanoids bind to classes of G protein-coupled receptors designated EP (for E-prostanoid receptor), FP, DP, IP and TP receptors respectively, which differ in their effects on cyclic AMP (cAMP) and/or phosphoinositol turnover and intracellular Ca2+ mobilization. However, recent studies focusing on the effects of PGE2 receptor activation on neuronal survival have uncovered a paradoxical neuroprotective effect of the PGE2 EP2 (McCullough et al. 2004) and EP3 receptors (Bilak et al. 2004), suggesting that PGE2 in particular may not mediate COX-2 neurotoxicity.

In peripheral tissues, PGD2 promotes vasodilatation and bronchoconstriction, inhibits platelet aggregation, mediates allergic reactions and reduces intraocular pressure (Whittle et al. 1983; Narumiya and Toda 1985; Matsugi et al. 1995; Matsuoka et al. 2000). In brain, it is the most abundant prostaglandin (Abdel-Halim et al. 1977; Narumiya et al. 1982; Hertting and Seregi 1989) and participates in the induction of non-rapid eye movement sleep (Hayaishi et al. 1993; Urade and Hayaishi 1999; Mizoguchi et al. 2001) and regulation of nociception (Eguchi et al. 1999). PGD2 is formed from the product of COX activity, PGH2, by two distinct PGD synthases, the lipocalin-type synthase (L-PGDS) and the glutathione-dependent hematopoietic PGD synthase (H-PGDS). PGD2 elicits its downstream effects by activating two G protein-coupled receptors, the DP and chemoattractant receptor-homologous molecule expressed on Th2 cells (CRTH2) receptors (hereafter termed DP1 and DP2) (Hirata et al. 1994; Abe et al. 1999; Hirai et al. 2001; Monneret et al. 2001). The DP1 and DP2 receptors have divergent effects on cAMP production, and are coupled to Gαs and Gαi respectively (Crider et al. 1999; Hata et al. 2003). PGD2 can also be degraded non-enzymatically to form J-series cyclopentone prostaglandins such as 15-deoxy-Δ12,14-PGJ2 (15d-PGJ2) which at increased concentrations can bind the peroxisome proliferator-activated receptor-γ (Kliewer et al. 1995).

Of the COX-derived eicosanoids, PGD2 levels exhibit marked changes in pathological paradigms characterized by excess glutamate release, including seizures (Seregi et al. 1985, 1990; Naffah-Mazzacoratti et al. 1995) and ischemia (Tegtmeier et al. 1990; Huttemeier et al. 1993). Increases in PGD2 levels in these settings result in large part from increased COX-2 activity and production of PGH2, because increases in PGD2 as well as neuronal injury can be blocked with inhibition of COX-2 (Govoni et al. 2001). In this study, we investigated whether PGD2 was responsible for mediating COX-2 neurotoxicity via its DP1 and DP2 receptors, and assessed the specificity of its effect on neuronal survival by examining the function of its metabolite 15d-PGJ2 in primary neuronal and organotypic hippocampal cultures.

Materials and methods


This study was conducted in accordance with the National Institutes of Health guidelines for the use of experimental animals. Protocols were approved by the Institutional Animal Care and Use Committee at Johns Hopkins University. Female Sprague–Dawley rats were obtained from Charles River (Wilmington, MA, USA) for dispersed neuronal and organotypic cultures.

Prostaglandin receptor agonists and other reagents

Prostaglandin agonists were prepared as a 10-mm stock in 100% ethanol and frozen at −70°C until use. PGD2, the DP1 and DP2 receptor agonists BW245C and 13,14-dihydro-15-keto-PGD2 (DK-PGD2), and 15d-PGJ2 were obtained from Cayman Chemicals (Ann Arbor, MI, USA). Stock solutions were diluted such that the ethanol concentration in experimental and vehicle-treated cultures was 0.1%.

Generation of cRNA probes

For in situ hybridization, cRNA probes for the DP1 and DP2 receptors and L-PGDS and H-PGDS were generated by RT–PCR of total RNA derived from cerebral cortex and hippocampus of adult rat. The RT–PCR was performed using Superscript one-step RT–PCR kit (Invitrogen, Carlsberg, CA, USA) at 50°C for 30 min according to the manufacturer's instructions. Nested primers were designed to generate the open reading frame of DP1, DP2, L-PGDS and H-PGDS using sequences of rat DP1 (accession number XM223897), mouse DP2 (accession number NM009962), rat L-PGDS (accession number NM013015) and rat H-PGDS (accession number D82071). Nested outer and inner primers were designed from divergent regions of DP1, DP2, LPGDS and HPGDs sequences (Gene Works, Intelligenetics, Campbell, CA, USA). BLAST sequence comparison showed no or very low homology of these primer sequences with other genomic sequences. Outer primers were as follows: DP1, sense 5′-CCGCCCTCGGTCTTTTA-3′ and antisense 5′-TGAAGATCCAGGGGTCCA-3′; DP2, sense 5′-TGCGCCAGACAGTGGTC-3′ and antisense 5′-CACTGCCACGCCTCATC-3′, LPGDS, sense 5′-CCTCAGGCTCAGACACC-3′ and antisense 5′-CAAGGCAAAGCTGGAGG-3′; HPGDS, sense 5′-TTCCAGGCTGAACTTTGGA-3′ and antisense 5′-TACACAGCAAGCACAAT-3′. PCR using outer primers was carried out in a volume of 25 µL (1 × reaction buffer, 0.2 mm dNTPs, 2.0 mm MgCl2, 100 nm primers, 1 µL RT product, 0.3 µL RT/Platinium Taq) for 30 cycles (94°C for 40 s, 52°C for 45 s and 72°C for 60 s), followed by a 10-min extension cycle at 72°C. The PCR reaction mix was then diluted 1 : 1000 and used as template for inner nested primers incorporating 23 bases corresponding to the promoter site of SP6 (sense primers) or T7 RNA polymerase (antisense): SP6-DP1, 5′-TATTTAGGTGACACTATAGTTGGGTTAGCCTCGACCTTA- 3′; T7-DP1, 5′-TAATACGACTCACTATAGGGTGCCTGTAGTCTGAGCCTGA-3′ (451 bp); SP6-DP2, 5′−AATTAACCCTCACTAAAGGGCGTGTTCAGAGACACCAT-CC-3′; T7-DP2, 5′-TAATACGACTCACTATAGGGTTGACCACGCTGTTGAAGAA-3′ (438 bp); SP6-LPGDS, 5′-TATTTAGGTGACACTATAGCCTCCAATTCAAGCTGGTTC-3′; T7-LPGDS 5′-TAATACGACTCACTATAGGGTGAATGCACTTATCCGGTTG-3′ (457 bp); SP6-HPGDS, 5′-TATTTAGGTGACACTATAGCACTCGCTCATCACAGAAGC-3′; T7-HPGDS, 5′-TAATACGACTCACTATAGGGCAACAGGTCGGGCTTTAAGA-3′ (631 bp). PCR using inner primers was carried out in a volume of 25 µL (1 × reaction buffer, 0.2 mm dNTP, 2.5 mm MgCl2, 100 nm primers, 10 µL diluted PCR mix and 0.35 µL Taq DNA polymerase) at 95°C for 4 min followed by 94°C for 40 s, 56–58°C for 45 s, 72°C for 1 min, for 30 cycles, and a final extension time of 10 min at 72°C. The identity of all four cDNAs was confirmed by sequencing at the Core Sequencing Facility, Johns Hopkins University.

In situ hybridization

Radiolabelled cRNA probes were generated for DP1, DP2, L-PGDS and H-PGDS with 100 µCi 35S-labelled UTP (NEN life science, Boston, MA, USA) using an in vitro transcription kit (Ambion, Austin, TX, USA) as described previously (Yamagata et al. 1993). Cryostat sections (14 µm thick) mounted on Superfrost Plus slides were equilibrated to room temperature 22°C and fixed for 10 min in 4% paraformaldehyde. Sections were then rinsed 2 × SSC (0.25 m sodium chloride, 0.015 m sodium citrate, pH 7.2) for 2 min, acetylated for 10 min in triethanolamine (0.1 m, pH 8.0, with 0.25% acetic anhydride), dehydrated through graded ethanols and delipidated in chloroform. Sections were rehydrated in 70% ethanol and allowed to dry. Sections were hybridized overnight at 56°C with 1.0 × 106 cpm 35S-labeled cRNA diluted in hybridization medium (50% formamide, 20 mm Tris-HCl, pH 7.5, 1 mm EDTA, 335 mm NaCl, 1 × Denhardt's, 200 g/mL salmon sperm DNA, 150 g/mL yeast tRNA, 20 mm dithiothreitol and 10% dextran sulfate). Slides were rinsed twice each for 5 min in 2 × SSC buffer and treated with Rnase A (100 µg/mL) for 30 min at 37°C. Sections were then washed in 2 × SSC and dehydrated through a graded series of ethanol solutions and air-dried. Slides of sense and antisense probes were exposed to Kodak BioMax X-ray film (Eastman Kodak Company, Rochester, NY, USA) for identical time periods (1–10 days).

Neuronal cultures and assessment of neuronal survival

Pure hippocampal cultures were prepared from E18 rat embryos and plated at a density of 250 000 cells per well in 24-well plates coated with poly-d-lysine (BD Biosciences, Bedford, MA, USA) in Neurobasal medium, 1 × B27, penicillin/streptomycin (Gibco, Carlsbad, CA, USA) as described previously (McCullough et al. 2004). Cultures consisted of 94.6% neurons and 5.2% astrocytes. Excitotoxicity assays were begun on day 10 in vitro. Cell death was quantified 24 h after glutamate stimulation by cell counting and lactose dehydrogenase assay (LDH) of culture medium. For cell counting, quantification of cell death was performed in cultures stained with trypan blue for 30 min at 37°C followed by washing with phosphate-buffered saline (PBS); cells were then fixed with 4% paraformaldehyde and counted with bright-field microscopy. Cells that stained blue were counted as dead, and those that were unstained were counted as viable. Some 100–200 cells in each of four fields were counted for 3–4 wells per condition. For LDH measurements, culture media was diluted 1 : 3 and assayed for LDH (Roche, Indianapolis, IN, USA) in accordance with the manufacturer's instructions. Mean values from three or more experiments were combined.

RT–PCR of DP receptors and PGD2 synthases in culture

Expression of DP1 and DP2 receptors was assayed by RT–PCR, as no antibodies are presently commercially available. Total RNA was extracted from cultured hippocampal neurons with TRIZOL (Invitrogen), and treated with Dnase (Invitrogen) to eliminate any contaminating genomic DNA. The RT reaction was performed using Superscript one-step RT–PCR followed by PCR amplification using Platinium Taq DNA Polymerase (Invitrogen), using the same outer primers as described for in situ hybridization, and the following inner primers: DP1, sense 5′-AGAAGCGCTCATTCTCGGTA-3′ and antisense 5′-GTCTTCCGAGTCTCCGTCAG-3′; DP2, sense 5′-AAGCTACATTCCTCGGTCTTCTT-3′ and antisense 5′-ATTCCAGAGCAGCAAGTTGTAGT-3′; LPGDS sense 5′−TCCGG-GAGAAGAAAGAGCTAC-3′ and antisense 5′-TGGTCCTTGC-TAAAGGTGATG-3′; HPGDS, sense 5′-ATGCCCAACTACAAA-CTGCTT-3′ and antisense 5′-TCTGTCTTCCCAGCCAAATC-3′. Amplification using the inner primers was carried out in a volume of 25 µL (1 × reaction buffer, 0.2 mm dNTPs, 2.5 mm MgCl2, 100 nm primers, 10 µL diluted outer PCR mix and 0.3 µL Taq polymerase) for 30 cycles (94°C for 40 s, 54°C−58°C for 45 s, 72°C for 60 s followed by a 10-min extension at 72°C). The sizes of the DP1, DP2, LPGDS and HPGDS fragments were 338, 258, 346 and 245 bp, consistent with previously published sequences, and the identity of each PCR product was verified by sequencing. Control reactions for outer and inner primers were performed without RT to control for genomic DNA contamination; no bands were observed for these reactions.

Organotypic hippocampal slice model and measurement of neuronal death

Hippocampal organotypic cultures were prepared as described previously (McCullough et al. 2004). Briefly, 350-µm coronal hemispheric slices were derived from postnatal day 7 rat pups and plated on to 30-mm Millicell membrane inserts (0.4 µm; Millipore, Bedford, MA, USA) in six-well plates, with each well containing 1 mL medium (50% minimal essential medium, 25% Hank's balanced salt solution, 25% heat-inactivated horse serum, 6.5 mg/mL d-glucose, 5 units/mL penicillin G and 5 µg/mL streptomycin sulfate). Cultures were maintained in a humidified incubator under 5% CO2 at 37°C for 13 days and the medium was changed twice weekly. At 13 days in culture, the medium was replaced with fresh medium containing propidium iodide (PI) (5 µg/mL; Sigma, St Louis, MO, USA) for 24 h, and basal PI fluorescence was imaged the next day to confirm that slices were healthy and did not show any spontaneous neuronal death (this time point is referred to as ‘basal’ or t = 0 h). Slices were then stimulated with 10 µm NMDA for 1 h in the presence of prostaglandin receptor agonists or vehicle. NMDA was used instead of glutamate because the astrocytes in organotypic slices can clear glutamate via their glutamate transporters. After stimulation, the medium was replaced with fresh medium containing either prostaglandin receptor agonist or vehicle (ethanol 0.1%) and PI (5 µg/mL), and slices were incubated for another 24 h, after which they were imaged again for PI fluorescence (t = 24 h). The medium was then removed and replaced with fresh medium containing a lethal dose of 10 µm NMDA and PI and incubated overnight (time point t = max). Control and NMDA-treated slices were included in each experiment with triplicate wells or 15 slices per condition for each experimental group.

Measurement of neuronal death in organotypic hippocampal slices

Neuronal death was assayed by quantification of mean PI fluorescence in the CA1 and CA3 subregions of each hippocampal slice (Newell et al. 1995). Changes in PI fluorescence in the CA1 and CA3 regions have been correlated with pyramidal neuronal populations (Vornov et al. 1995) and do not reflect changes in astrocytes and microglia, which are more resistant to excitotoxic insults. Mean PI fluorescence intensity represents the mean of the fluorescence intensity values of each pixel in the image of the subregion measured (CA1 and CA3) and is proportional to the number of injured cells. Sequential fluorescence was measured at t = 0 (before NMDA stimulation), at t = 24 h after stimulation with NMDA and at t = max, following a final treatment of slices with a 24-h overnight lethal treatment with 10 µm NMDA. Slices were imaged using an inverted Nikon Diaphot microscope connected to a Cool Snap charge coupled device (CCD) camera (Roper, Tucson, AZ, USA). The same exposure time and gain were used in all experiments. The images were collected and analyzed using Open Lab imaging software (Improvision, Lexington, MA, USA), digitized at 8 bits/pixel, and mean fluorescence values of the CA1 or CA3 were measured for t = 0, t = 24 h and t = max. The background fluorescence measured at t = 0 was subtracted from the t = 24 h and the t = max values, and the percentage neuronal death was calculated by normalizing the t = 24 h value to the t = max value for each slice according to the formula: [(t = 24 h) − (t = 0)]/[(t = max) − (t = 0)]. Each experiment was done in triplicate, with 10–15 sections per condition per experiment. Control slices were processed in parallel with experimental slices, and were subjected to all washes and changes in medium. The absolute level of PI fluorescence in NMDA-treated slices was on average 2–3 fold higher than that in non-NMDA treated controls at t = 24 h. To average experiments done in triplicate, NMDA and experimental (NMDA + agonist) values were normalized by subtracting basal background PI fluorescence from NMDA and experimental values, and expressing experimental conditions as a percentage of the NMDA value.

Immunostaining of L-PGDS and H-PGDS

Hippocampal neurons were cultured on glass coverslips for 14 days and fixed for 30 min in 4% paraformaldehyde/4% sucrose in PBS, rinsed in PBS, and incubated in methanol at 4°C for 20 min. Neurons were permeabilized with 1% Triton for 30 min, blocked in 0.5% Triton, 5% goat serum in PBS for 1 h, and incubated with antibody to L-PGDS (polyclonal, 1 : 100; Cayman Chemicals) or H-PGDS (polyclonal, 1 : 100, Cayman Chemicals) with either monoclonal NeuN (Chemicon, Temecula, CA, USA) or monoclonal anti-glial fibrillary acidic protein (Sigma) with 3% blocking serum overnight at 4°C. Cells were then washed in PBS and incubated with anti-rabbit and anti-mouse Alexa Fluor 488 and 562 secondary antibody (Molecular Probes, Eugene, OR, USA), rinsed and mounted on slides for visualization by microscopy.

Determination of cAMP

Neurons were plated at 1 × 106 cells per well in six-well plates. On day 10 in vitro, cAMP measurements were carried out using the non-acetylated version of a commercial assay kit (Assay Designs, Inc., Ann Arbor, MI, USA). Phosphodiesterase was inhibited by the addition of 3 isobutyl-1-methyl xanthine (50 µm) to the cultures for 30 min. PGD2, the DP1 agonist BW245C or the DP2 agonist DK-PGD2 were added at 100 nm to cultures for 5, 10, 15, 20 and 30 min, after which cells were harvested. Forskolin (100 nm), a direct and potent activator of adenylyl cyclase, was used as a positive control. Cyclic AMP concentrations were standardized to protein content using the Pierce BCA kit (Pierce, Rockville, IL, USA).

Statistical analysis

Statistical analysis was performed by one-way anova, followed by Tukey post-hoc analysis. For cAMP measurements, one-way anova was followed by Neuman-Keuls post-hoc tests. All data are reported as mean ± SEM). p < 0.05 was considered significant.


In situ hybridization of components of the PGD2 system

The expression levels and anatomical distribution of components of the PGD2 system were investigated by in situ hybridization of the DP1 and DP2 receptors and L-PGDS and H-PGDS synthases, which convert COX-derived PGH2 to PGD2(Fig. 1). Both PGD2 synthases were expressed in brain in distinct regions. L-PGDS mRNA was highly expressed in leptomeninges and white matter tracts, consistent with previous studies showing localization to arachnoid membranes and oligodendrocytes (Urade et al. 1993; Beuckmann et al. 2000). In addition, L-PGDS mRNA was abundantly expressed in thalamus, brainstem, cerebellum and spinal cord, with a marked drop-off in levels of expression rostrally in telencephalic regions. H-PGDS, which is expressed in early postnatal development in microglial cells and in cerebellum (Mohri et al. 2003), was highly expressed in adult brain in cerebellum and modestly throughout gray and white matter, with moderate expression in the hippocampal pyramidal neuronal layer.

Figure 1.

In situ hybridization of DP receptors and PGD2 synthases in adult rat brain. In situ hybridization of coronal and saggital sections show mRNA expression of both DP1 and DP2 receptors and both PGD2 synthases. DP1 mRNA is expressed in gray matter, with robust levels in cerebral cortex, hippocampal pyramidal layers and dentate gyrus, thalamus and choroid plexus, and on saggital section is apparent in leptomeninges in a discontinuous fashion. DP2 mRNA is expressed throughout gray and white matter. L-PGDS is highly expressed in leptomeninges, choroid plexus and white matter tracts; it is also robustly expressed caudally in thalamus, brainstem and cerebellum, with lower levels of expression in cerebral cortex, striatum and hippocampus. Unlike L-PGDS, H-PGDS is expressed throughout gray and white matter, with expression in the pyramidal neuronal layer of the hippocampus. Sense cRNA probes (far-right column) do not show a specific signal.

Previous investigations of DP1 receptor expression have demonstrated that it is present in the leptomeninges, choroid plexus and eye tissues (Oida et al. 1997; Gerashchenko et al. 1998; Mizoguchi et al. 2001). Here, the DP1 receptor was expressed in a discontinuous fashion over the surface of the brain, consistent with a leptomeningeal localization, as described previously (Oida et al. 1997; Gerashchenko et al. 1998). In additional, DP1 expression was present in gray matter in hippocampus, cerebral cortex, thalamus and brainstem. DP2 receptor mRNA was also expressed in gray matter, and was prominent in the pyramidal cell layer of the hippocampus. The forebrain expression of DP2 in hippocampus, cerebral cortex, thalamus and hindbrain extends previous findings from northern analyses of human brain (Marchese et al. 1999; Nagata and Hirai 2003). For all constructs, hybridization with sense cRNA probes yielded no specific staining above background (Fig. 1, right column). Thus, both DP1 and DP2 receptors are expressed in cerebral cortex and hippocampus.

Components of the PGD2 system are expressed in cultured hippocampal neurons

RT–PCR revealed expression of DP1 and DP2 receptors and both synthases in hippocampal neuronal cultures (Fig. 2a). The functional activity of both receptors was established with application of either PGD2 (100 nm), the selective DP1 agonist BW245C (100 nm) (Town et al. 1983) or the selective DP2 agonist DK-PGD2 (1 µm) (Hirai et al. 2001; Hata et al. 2003) (Fig. 2c). Application of PGD2 and BW245C led to a rapid increase in cAMP. Conversely, application of the DP2 agonist DK-PGD2 resulted in a decrease in cAMP levels over 120 min. These results are consistent with the positive and negative coupling to cAMP that have previously been shown in non-neuronal cells (Hirai et al. 2001; Hata et al. 2003), and confirm the presence of functional DP1 and DP2 receptors in neurons.

Figure 2.

Cultured hippocampal neurons possess synthase and receptor components of the PGD2 system. (a) RT–PCR of DP1 and DP2 receptors, and L-PGDS and H-PGDS with (+) and without (–) RT shows that mRNA is present for both synthases in hippocampal neuronal cultures. (b) PGD2 synthase-like immunoreactivity for H-PGDS and L-PGDS co-localizes to neurons stained with NeuN. The negative control without primary antibody showed no staining (not shown). (c) Stimulation with PGD2 (100 nm) or the DP1 receptor agonist BW245C (100 nm) elicits an increase in the production of cAMP over 30 min compared to 0 min, whereas application of the DP2 agonist DK-PGD2 (1 µm) leads to a decrease in levels of cAMP over 120 min, indicating that hippocampal neurons express functional DP1 receptors. There is no change in cAMP levels over time with vehicle alone. Values are mean ± SEM. Statistical analysis was by one-way anova was followed by Neuman–Keuls post-hoc test.

We investigated the expression levels of L-PGDS and H-PGDS in hippocampal neurons (Fig. 2b). Strong staining of H-PGDS was found in hippocampal neurons that co-localized with Neu N staining, indicating neuronal expression of H-PGDS. H-PGDS was abundant in neuronal processes and in a perinuclear distribution, consistent with expression in endoplasmic reticulum. L-PGDS was present at lower levels, mainly in cytoplasm and proximal neuronal processes. Thus, consistent with the in vivo expression patterns of the DP1 and DP2 receptors, these data indicate that hippocampal neurons possess functional PGD2 receptor and signaling systems.

Effects of PGD2 in primary hippocampal neurons

The question of whether PGD2 can mediate the neurotoxicity of COX-2 was investigated using a paradigm of glutamate toxicity in which hippocampal neurons were stimulated with glutamate after 10 days in vitro(Fig. 3). Physiologic concentrations of PGD2 (1 nm to 1 µm; Ki = 40 nm; Hirata et al. 1994) had a paradoxical and significant protective effect on neuronal survival as assayed by both LDH release into the media and counting of trypan blue-stained cells. The protective effect of PGD2 was further investigated to determine which of the two DP receptors was responsible for this marked neuroprotection. Activation of the DP1 receptor with its selective agonist BW245C (Ki = 50 nm; Hata et al. 2003) resulted in dose-dependent neuroprotection from 1 nm to 1 µm. Conversely, activation of the DP2 receptor with its selective agonist DK-PGD2 (Ki = 160 nm) resulted in toxicity at higher concentrations (1–10 µm). The divergent effects of the DP1 and DP2 receptors occurred in the context of divergent effects on cAMP production (see Fig. 2). Thus, activation of the DP1 receptor, which is positively coupled to cAMP, resulted in protection of neurons whereas activation of the DP2 receptor, which is negatively coupled to cAMP via a pertussis toxin-sensitive mechanism (Hirai et al. 2001; Sawyer et al. 2002), induced toxicity. Administration of the PGD2 metabolite 15d-PGJ2 did not affect neuronal survival at low concentrations (1 nm to 1 µm; data not shown) and at higher micromolar concentrations (up to 50 µm) (Fig. 3d), indicating that the neuroprotection mediated by PGD2 was not occurring through its degradation to the cyclopentone 15d-PGJ2. PGD2 exerted a net neuroprotective effect, indicating that activation of the DP1 pathway predominated over that of the DP2 pathway in neurons.

Figure 3.

PGD2 elicits a net neuroprotective effect in primary hippocampal neurons subjected to glutamate toxicity (50 µm). After 10 days in vitro, hippocampal neurons were stimulated with glutamate with or without prostaglandin/agonist for 16–20 h, at which time medium was harvested for quantification of LDH (left column) and neurons were stained with trypan blue for counting of dead cells (right column). Basal LDH values and basal levels of dead cells were subtracted from values obtained in the presence of glutamate with or without prostaglandin/agonist. Experimental values (glutamate + prostaglandin/agonist) were then normalized for each experiment with respect to those obtained with glutamate alone (represented as 100%) and averaged; data represent combined values for three separate experiments, 4–6 wells per condition. The concentration of prostaglandin/agonist is shown on the x-axis. (a) PGD2 administration protects neurons at physiologic concentrations (1 nm to 1 µm). (b) The DP1 agonist BW245C rescues neurons at increasing concentrations, with maximal protection at 1 µm. (c) The DP2 agonist DK-PGD2 potentiates neuronal toxicity at micromolar concentrations. (d) The PGD2 metabolite, 15d-PGJ2 has no significant effect on neuronal survival at concentrations up 50 µm. ***p < 0.001; **p < 0.01; *p < 0.05 (one-way anova followed by Tukey post-hoc analysis).

Effects of DP receptor activation in hippocampal organotypic cultures

To model the effects of PGD2 in a system more closely approximating the in vivo state, the effects on neuronal survival of PGD2 were examined in hippocampal organotypic slice cultures (Fig. 4). Hippocampal slices were stimulated at day 14 in vitro with NMDA (10 µm for 1 h) in the presence or absence of either PGD2, BW245C, DK-PGD2 or 15d-PGJ2, and the percentage maximal PI fluorescence was measured for the CA1 and CA3 regions of hippocampus. PGD2 administration again resulted in a marked protective effect on hippocampal pyramidal neuron survival in both CA1 and CA3, and this effect was also seen with selective activation of the DP1 receptor using BW245C. Consistent with the results from dispersed hippocampal cultures, micromolar doses of the DP2 receptor agonist DK-PGD2 resulted in neurotoxicity. Increasing doses of 15d-PGJ2 had no effect on hippocampal neuronal viability, in accordance with the results seen in dispersed hippocampal cultures. Taken together, the data in organotypic cultures are consistent with the findings in primary hippocampal culture, and suggest that although DP1 and DP2 receptors have antagonistic effects on neuronal survival, PGD2 mediates a dominant protective effect through its DP1 receptor. Moreover, these effects are conserved in pure neuronal cultures as well as slices, suggesting that the introduction of non-neuronal cells such as astrocytes and microglia in organotypic slices does not interfere with the PGD2-mediated protective effect on neurons.

Figure 4.

PGD2 mediates a net neuroprotective effect in organotypic hippocampal cultures subjected to NMDA toxicity. (a) Representative panel of sequential fluorescent images from a control slice before stimulation (con), at 24 h after a 1-h stimulation with NMDA 10 µm (24 h), and after a further 24 h treatment with NMDA that results in maximal neuronal death (max); the far-right panel shows a representative Nissl-stained hippocampal slice illustrating the CA1 and CA3 pyramidal neuron subregions. There is no basal neuronal death before stimulation in CA1 or CA3 (con); however, at t = 24 h and t = max, there is a significant increase in PI fluorescence owing to death of pyramidal neurons in CA1 and CA3. (b–e) Neuronal death was measured in hippocampal CA1 and CA3 subregions that were stimulated with NMDA and either PGD2 (b), the DP1 agonist BW245C (c), the DP2 agonist DK-PGD2 (d) or the metabolite 15d-PGJ2 (e) (concentrations on x-axis). The percentage maximal PI fluorescence was calculated for each experiment, and normalized to PI fluorescence induced by NMDA (100%). Values from three separate experiments were averaged (n = 15 slices per condition per experiment). Increasing neuroprotection is observed in CA1 and CA3 regions with PGD2 (10 nm to 1 µm) and the DP1 agonist BW245C (1 nm to 1 µm). Conversely, toxicity is observed with increasing concentrations of the DP2 agonist DK-PGD2 at micromolar concentrations. The DP1 and DP2 receptors have antagonistic effects on cAMP production (Fig. 2) and opposing effects on neuronal survival. The PGD2 metabolite 15d-PGJ2 has no effect on neuronal survival at concentrations up to 10 µm. Significance was calculated by one-way anova followed by Newman–Keuls multiple comparison test.

PGD2 neuroprotection is mediated by the cAMP/protein kinase A (PKA) pathway

There is growing evidence for a protective effect of activation of the cAMP pathway on neuronal survival (Rydel and Greene 1988; D'Mello et al. 1993; Hanson et al. 1998; Walton and Dragunow 2000). The effect of blocking PGD2 neuroprotection by inhibiting PKA activation was tested using two different PKA inhibitors, H89 and KT5720 (Kase et al. 1987; Chijiwa et al. 1990). We tested whether the protective effect of PGD2 depended on an intact cAMP signaling pathway in dispersed hippocampal neurons. Administration of H89 (1 µm) and KT5720 (1 µm) did not affect basal neuronal survival; neither H89 (data not shown) nor KT5720 altered glutamate toxicity. PGD2 at 10 nm and 100 nm induced a neuroprotective effect that was reversed with co-administration of H89 (p < 0.001) (Figs 5a and b). Administration of KT5720 also resulted in a reversal of PGD2 neuroprotection (p < 0.001) (Figs 5c and d). Thus, inhibition of PKA activation using two distinct PKA inhibitors reversed the neuroprotective effect of PGD2 in neurons stimulated with toxic doses of glutamate, indicating that the neuroprotective effect of PGD2 is mediated by cAMP and activation of PKA.

Figure 5.

The neuroprotective effect of PGD2 is reversed by inhibition of PKA activation and dependent on DP1 receptor-mediated cAMP signaling. Primary hippocampal neurons were stimulated with glutamate (Glu; 100 µm) for 16–20 h with or without PGD2, and the PKA inhibitors H89 (1 µm; a and b) and KT5720 (1 µm; c and d). Basal levels of LDH and cell death were subtracted from values obtained with glutamate with or without prostaglandin/agonist, and experimental values were normalized for each experiment with respect to values for no glutamate (represented as 100%) and averaged (three separate experiments, 4–6 wells per condition). Percentage LDH release (a and c) and total numbers of dead cells (b and d) were quantified. **,***Significant neuroprotective effect of PGD2 compared to glutamate. Statistical analysis was by one-way anova followed by Newman–Keuls multiple comparison test.


In this study, we examined whether PGD2 might mediate the toxic effects of COX-2 enzymatic activity. Studies of mRNA localization indicated that both DP receptors are robustly expressed in hippocampus and cerebral cortex, regions that are relevant to transduction of COX-2 neurotoxicity in syndromes such as stroke and neurodegeneration. We focused on the function of the PGD2 DP1 and DP2 receptors on neuronal survival using primary hippocampal neurons and confirmatory studies in organotypic hippocampal slices. In cultured neurons, both DP receptors were expressed, and activation of DP1 or DP2 receptors with selective agonists led to induction and reduction of cAMP levels respectively, indicating the presence of functional DP receptors. In neurons and organotypic hippocampal slices subjected to glutamate toxicity, administration of PGD2 or the DP1 agonist BW245C resulted in a marked dose-dependent neuroprotection. Conversely, the DP2 receptor, which is negatively coupled to cAMP production, promoted neuronal injury in both dispersed and organotypic neurons. The net effect of PGD2 however, was protective, indicating that DP1-mediated signaling was dominant in these culture assays. The dominant effect of DP1-mediated signaling may be due to increased levels of DP1 receptors and downstream signaling proteins relative to the DP2 pathway, and an up-regulation of this protective pathway relative to the toxic DP2 pathway. The neuroprotective effect of PGD2 was blocked by inhibitors of PKA, indicating that PGD2 protection was dependent on downstream effects of increased cAMP. These data demonstrate a novel neuroprotective pathway mediated by PGD2 in brain in which activation of its DP1 receptor can prevent neuronal injury in paradigms of acute excitotoxicity.

Increased COX-2 activity is a hallmark of excitotoxic injury such as cerebral ischemia in which pharmacologic inhibition of COX-2 activity leads to reduction in levels of PGD2 (Govoni et al. 2001) and PGE2 (Nogawa et al. 1997; Nakayama et al. 1998) and rescue of neurons in the penumbra. These data, as well as findings that inhibition of COX-2 protects vulnerable neuronal populations in models of neurodegenerative diseases such as Parkinson's disease and amyotrophic lateral sclerosis (Drachman et al. 2002; Pompl et al. 2003; Teismann et al. 2003) suggest that the downstream prostaglandin products of COX-2 are toxic. In this study, we found that PGD2, the most abundant prostaglandin in brain, instead promoted neuroprotection. This unexpected effect is similar to that of PGE2, which can mediate neuroprotection in glutamate- and lipopolysaccharide-induced neuronal toxicity (Cazevielle et al. 1994; Kim et al. 2002), via its EP2 receptor in a cAMP-dependent fashion in a model of ischemia (McCullough et al. 2004), and its EP2 and EP3 receptors in a model of amyotrophic lateral sclerosis (Bilak et al. 2004). The shared neuroprotection of the DP1 and EP2 receptors, which are both coupled to cAMP, would suggest that other prostaglandin receptors similarly positively coupled to cAMP production, such as the PGI2 IP receptor and the PGE2 EP4 receptor, may also promote neuroprotection.

The findings of positive effects of PGD2 and PGE2 on neuronal survival led to an interesting paradox, in which the inhibition of COX-2 activity resulting in decreased production of prostaglandins is protective, but the administration PGE2 or PGD2 is also protective. This neuroprotection occurs at physiologic concentrations in the low to high nanomolar range, close to the known binding affinities of these receptors (Kiriyama et al. 1997; Breyer et al. 2002). These data suggest at the very least that PGE2 and PGD2 individually do not mediate the toxic effects of COX-2. Possibly one or more of the remaining prostaglandins, namely PGI2, PGF and thromboxane A2, may be important in this process. It is possible that a combination of one or more prostaglandins may lead to a neurotoxic effect, and this should be tested in future studies. In this regard, precedent exists for a combined protective effect of the PGE2 EP2 and EP3 receptors when administered together in a model of chronic neurodegeneration (Bilak et al. 2004).

Another possible candidate for COX-2-mediated neurotoxicity is 15d-PGJ2. PGD2 is very short lived and is normally degraded to form one of several J-series cyclopentone prostaglandins, including 15d-PGJ2, which has been found to bind the peroxisome proliferator-activated receptor-γ at micromolar concentrations (Kliewer et al. 1995). The effects of the 15d-PGJ2 metabolite were tested in part to verify the specificity of the PGD2 neuroprotective effect, and also to examine whether there was an independent effect of 15d-PGJ2 on the survival of dispersed and organotypic hippocampal neurons subjected to glutamate toxicity. Previous investigations of 15d-PGJ2 effects on neuronal survival have yielded divergent results, depending on the neuronal cell type, toxicity paradigm and assay for cell death. 15d-PGJ2 rescues cerebellar granule cells in models of lipopolysaccharide (LPS) inflammation (Heneka et al. 2000) and retinal ganglion cells subjected to glutamate toxicity (Aoun et al. 2003), but promotes toxicity of cortical neurons (Rohn et al. 2001). In our study, 15d-PGJ2 elicited no effect on primary hippocampal neurons or organotypic slices even at micromolar concentrations, confirming that PGD2, which was protective at low nanomolar concentrations, was not eliciting its protective effect through its 15d-PGJ2 metabolite.

Alternative mechanisms for COX-2 neurotoxicity should be explored in future studies, including the potential toxic effect of reactive oxygen species that are produced in the conversion of arachidonic acid to PGH2 (Eling et al. 1990; Smith et al. 1991). However, it is not clear whether the level of free radical generation by cyclo-oxygenase activity reaches that of the nitric oxide pathway or mitochondrial respiration, and in a recent study COX-2-derived free radical species were not found to contribute significantly in a model of stroke injury (Manabe et al. 2004). Another unexplored mechanism of COX-2 neurotoxicity might involve the shunting of arachidonic acid toward the prostaglandin pathway and away from other pathways that employ arachidonic acid as a substrate, including the leukotrienes, hydroxyeicosatetraenoic acids, epoxyeicosatrienic acids and lipoxins. Indeed, recent studies have highlighted potential anti-inflammatory properties of non-prostaglandin eicosanoids (Serhan and Chiang 2002).

In summary, PGD2 mediates an unexpected neuroprotective effect in hippocampal neurons, an effect mediated by the DP1 receptor and dependent on intact cAMP signaling. These findings are consistent with recent studies of PGE2 EP receptors that also indicate a neuroprotective effect of prostaglandin receptors. Taken together, these data support an emerging and paradoxical neuroprotective role for prostaglandins that may present novel therapeutic targets in neurological diseases.


This work was supported by the National Institute of Neurological Disorders and Stroke (NINDS) (KA) and the American Federation for Aging Research (KA).