Localization of Functional Prostaglandin E2 Receptors EP3 and EP4 in the Nuclear Envelope*

The effects of prostaglandin E2are thought to be mediated via G protein-coupled plasma membrane receptors, termed EP. However recent data implied that prostanoids may also act intracellularly. We investigated if the ubiquitous EP3 and the EP4 receptors are localized in nuclear membranes. Radioligand binding studies on isolated nuclear membrane fractions of neonatal porcine brain and adult rat liver revealed the presence of EP3 and EP4. A perinuclear localization of EP3α and EP4receptors was visualized by indirect immunocytofluorescence and confocal microscopy in porcine cerebral microvascular endothelial cells and in transfected HEK 293 cells that stably overexpress these receptors. Immunoelectron microscopy clearly revealed EP3α and EP4 receptors localization in the nuclear envelope of endothelial cells; this is the first demonstration of the nuclear localization of these receptors. Data also reveal that nuclear EP receptors are functional as they affect transcription of genes such as inducible nitric-oxide synthase and intranuclear calcium transients; this appears to involve pertussis toxin-sensitive G proteins. These results define a possible molecular mechanism of action of nuclear EP3 receptors.

are also suggested by other data. For example, a transporter that mediates the influx of prostanoid has been identified (8). Cytosolic phospholipase A 2 undergoes a calcium-dependent translocation to the nuclear envelope (9), and cyclooxygenase-2 has been shown to translocate to the nucleus in response to certain growth factors (10). It is thus possible that prostanoids may exert some of their effects via intracellular EP receptors, to have a direct nuclear action as recently proposed by Goetzl et al. (11), and Morita et al. (12).
It has generally been assumed that the signal transduction cascades are initiated at the plasma membrane and not the nuclear membranes. However, recent studies have disclosed that the nuclear envelope plays a major role in signal transduction cascades (13,14). In fact, a novel nuclear lipid metabolism that is a part of unique nuclear signaling cascade termed NEST (nuclear envelope signal transduction) has been hypothesized (15). Both heterotrimeric and low molecular weight G proteins (15,16), phospholipase C (13), phospholipase D (15), and adenylate cyclase (17) have shown to be localized at the nucleus. The nuclear membranes also have distinct inositide cycles (18) and receptors for 1,4,5-triphosphate and inositol 1,3,4,5-tetrakisphosphate (13). Altogether these data raise the possibility of the presence of nuclear prostanoid receptors. This inference has recently been suggested for EP 1 (19), but whether or not this single observation is specific for this receptor or applies to other prostanoid receptors, especially of the widely distributed EP 3 and EP 4 subtypes, is unknown.
In the present study, we investigated the possible expression of nuclear EP 3 and EP 4 receptors using human embryonic kidney (HEK) 293 cells, porcine microvascular endothelial cells, newborn pig brain, and adult rat liver. We selected these tissues because many high affinity PGE 2 binding sites have been reported in the plasma membranes of pig brain (20) and rat liver (21). We focused on EP 3 receptors that are most ubiquitous of the four EP subtypes (6) and also examined localization of EP 4 receptors. Our data provide novel evidence for the existence of EP 3 and EP 4 receptors in the nuclear envelope and reveal that these receptors are functional, and their actions appear to involve pertussis toxin (PTX)-sensitive G proteins. Animals-Newborn pigs (1-3 days old) were killed with intracardiac pentobarbital under halothane anesthesia, and tissues of interest were removed. Adult Sprague-Dawley male rats (250 -300 g) were decapitated and had livers removed.

Materials
Cell Culture-HEK 293 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum. Primary cultures of porcine cerebral endothelial cells from brain microvessels (20) were established as described previously (22).
Preparation of Subcellular Fractions-All steps were performed at 4°C using solutions containing 1.1 mM acetylsalicylic acid, 1 mM benzamidine, 0.2 mM phenylmethylsulfonyl fluoride and 100 g/ml soybean trypsin inhibitor. Nuclei were isolated from adult rat liver (23) and newborn porcine brain cortex (24). Endoplasmic reticulum (ER) was isolated as described (25). The purity of cellular fractions was ascertained by determining 5Ј-nucleotidase using a Sigma diagnostic kit as a marker for plasma membrane (23), and glucose-6-phosphatase was assayed as a marker for ER (26). Proteins were determined by the Bio-Rad assay. 5Ј-Nucleotidase activity (units/mg protein) was 240 Ϯ 15.4 and 5.3 Ϯ 1.3 in plasma and nuclear membrane fractions, respectively, suggesting that the nuclear membranes were relatively free of contamination by plasma membranes. Glucose-6-phosphatase (an ER marker) specific activity (mmol PO 4 released/mg of protein) was 25.2 Ϯ 2.3 in ER and 22.8 Ϯ 3.1 in nuclear membrane fractions since the outer nuclear membrane is contiguous with the ER (27).
Radioligand Binding-Saturation isotherms of specific binding of [ 3 H]PGE 2 to membrane fractions from newborn porcine brain cortex and displacement of [ 3 H]PGE 2 by receptor isoform-specific ligands on brain cortex and rat liver was performed essentially as previously reported (20). Receptor densities (B max ) and affinity constants (K D ) were determined using Prism Graphpad program (San Diego, CA).
Immunoblotting of EP Receptors-Western blotting of EP 3␣ and EP 4 receptors was conducted as described (28) on newborn brain nuclear and plasma membrane fractions. After immunoblotting using EP 3␣ -or EP 4 -specific polyclonal rabbit antibodies (29) (1:1000), immunoreactive bands were visualized by chemiluminescence (Amersham Pharmacia Biotech) as per manufacturer instructions.
EP 3 and EP 4 Receptor Expression in HEK 293 Cells-The full-length cDNA fragments corresponding to human EP 3␣ (30) and EP 4 (31) were cloned separately into the mammalian expression vector pRC-CMV (Invitrogen). HEK 293 cells (2 ϫ 10 5 ) were transfected with 2 g of plasmid DNA and 8 l of LipofectAMINE in Opti-MEM (Life Technologies, Inc.) according to the manufacturer instructions; Geneticin (1 mg/ml) -resistant clones were selected and maintained in Dulbecco's modified Eagle's medium medium containing Geneticin (0.2 mg/ml).
Immunocytochemical Detection of EP Receptors-The immunolocalization of EP receptors in HEK 293 and porcine cerebrovascular endothelial cells was performed by indirect immunofluorescence (32). Briefly, cells were washed in phosphate-buffered saline (PBS), fixed in acetone-methanol (1:1) for 10 min at Ϫ20°C and incubated for 1 h with specific rabbit anti-EP 3␣ or anti-EP 4 receptor antibodies (29) diluted 1:50 in PBS containing 5% goat serum, 5% fetal calf serum, and 0.1% Triton X-100. After washing, samples were incubated for 1 h with Texas Red-conjugated IgG (BioCan, Mississauga, ON) diluted 1:50 in the above buffer. To detect plasma membrane immunolocalization, permeabilization of cells with 0.1% Triton X-100 was limited to 15 min (to improve preservation of membranes) before incubation with primary antibodies and subsequent steps. As a negative control, either the primary antibody was omitted or primary antibody with its cognate peptide (29) was added. Intracellular membranes (predominantly ER) were stained using 3,3Ј-dihexyloxacarbocyanine iodide (DiOC 6 (3)) and nuclei were stained with DAPI, or Sytox Green according to the instructions of the manufacturer (Molecular Probes, Eugene, OR).
Immunoelectron Microscopy of EP Receptors-Pre-embedding immunogold staining was done as described in detail previously (33,34). Porcine brain endothelial cells were fixed for 30 min at room temperature in 4% paraformaldehyde and 0.25% glutaraldehyde in 0.1 M phosphate buffer, washed with 0.2% Triton X-100 in PBS for 15 min at room temperature, and incubated with anti-EP 3␣ or anti-EP 4 receptor antibodies (1:10) overnight at 4°C in PBS; this limited permeabilization of cells is required for pre-embedding immunoelectron microscopy for adequate ultrastructural preservation. The immunogold reaction (33) was performed overnight at 4°C with goat anti-rabbit IgG coupled to 1 nm of gold (1:50) (Amersham Pharmacia Biotech), the reaction was intensified using an Intense Silver Enhancement kit (Amersham Pharmacia Biotech) according to the instructions of the manufacturer. After osmification in 1% osmium tetroxide and Epon embedding, ultrathin sections were observed using a transmission electron microscope (Philips 410 LS, Netherlands).
RNA Hybridization Studies-Nuclei were isolated from endothelial cells (35), and aliquots (100 g of protein) were incubated in the pres-ence or absence of 0.1 M EP 3 agonist, M&B 28,767, for 60 min at 37°C in a total volume of 40 l (per reaction tube) of 10 mM Tris-HCl buffer (pH 8.0) containing 5 mM MgCl 2 , 300 mM KCl, 0.5 mM each of ATP, CTP, GTP, and UTP, RNase guard (111 units), and DNase (10 units). RNA was extracted as described previously (28). For the isolation of total cytoplasmic RNA, cells were incubated in the presence or absence of test agents for 1 h and washed with ice-cold PBS. Nuclear and total RNA were applied to a nylon membrane using a vacuum filtration apparatus (36). 32 P-Labeled cDNA probes for porcine iNOS (37) and mouse ␤-actin (Ambion) were prepared using an oligolabeling kit (Amersham Pharmacia Biotech); unincorporated nucleotides were removed by G-25 column chromatography. Membranes were hybridized to the radiolabeled probes and washed (36). The bands were visualized and quantified by Phosphorimaging (Molecular Dynamics).
Nuclear Calcium Signals and Uptake-The uptake of 45 Ca 2ϩ in isolated nuclei was determined as described previously (14). Briefly, nuclei were resuspended in buffer A (125 mM KCl, 2 mM K 2 HPO 4 , 25 mM Hepes, 4 mM MgCl 2 , and 400 nM CaCl 2 , pH 7.0). 45 Ca 2ϩ (2 Ci/ml) was added, and samples were incubated in the presence or absence of test agents at 37°C for different time periods. The reaction was terminated with ice-cold buffer containing 50 mM Tris-HCl and 150 mM KCl (pH 7.0), filtered under vacuum on glass fiber filters (GF/B, Whatman). The radioactivity on filters was counted on a beta-counter (Beckman LS 7500). The 45 Ca 2ϩ transient was defined as the radioactivity at a given time minus the radioactivity at time zero.
Effects of test agents on calcium transients were measured by fura-2/AM fluorometry as described (14,22) with some modifications. Isolated liver nuclei were resuspended in buffer A and preloaded with 7.5 M fura-2/AM for 45 min at 4°C. The nuclei were washed and stimulated (ϳ2 ϫ 10 6 nuclei/ml) with various test agents. The intranuclear calcium concentration was measured with a spectrofluorometer (LS 50, Perkin Elmer, Beaconsfield, UK) and fluorescent signal calibrated (22).
To assess the role of PTX-sensitive G proteins in nuclear Ca 2ϩ transients, isolated rat liver nuclei were treated with PTX as described (38). Prior to treatment, the toxin was preactivated by incubating at 37°C for 10 min in 50 mM Tris-HCl (pH 7.5) containing 100 mM dithiothreitol and 0.1 mM ATP. The isolated nuclei were incubated with preactivated PTX (20 g/ml) at 25°C for 20 min in buffer A containing 1 mM NAD and 50 M GDP. The treated nuclei were washed with buffer A and then the effects of prostaglandin analogs on intranuclear calcium levels were measured by fura-2/AM fluorometry as above. 2 Binding to Subcellular Fractions-The maximal specific binding of [ 3 H]PGE 2 to newborn pig brain plasma membrane, ER, and nuclear membrane fractions was comparable (Table I); in adult rat liver, the B max for [ 3 H]PGE 2 on plasma and nuclear membranes was also similar. The affinity constant (K D ) of PGE 2 binding did not significantly differ between tissues and membrane fractions. Subtypes of PGE 2 receptors were studied by displacement of bound [ 3 H]PGE 2 with PGE 2 , AH6809 (EP 1 receptor antagonist), butaprost (EP 2 agonist), M&B 28,767 (EP 3 subtype agonist), and AH23848B (EP 4 antagonist) (6). Neonatal porcine brain plasma and nuclear membranes contained mostly EP 3 receptors (nearly 100 and 45%, respectively); on nuclear membranes, the balance of EP receptors was equally divided among EP 1 , EP 2 , and EP 4 (Fig. 1,  a and b). In adult liver plasma membranes, all four EP receptors were detected in equal proportions (25%); in liver nuclear membranes, EP 3 was most abundant (40%), and EP 1 , EP 2 , and EP 4 consisted 15, 20, and 30% of EP receptors, respectively (Fig. 1, c and d).  ability of specific anti-EP 3␣ and anti-EP 4 receptor antibodies (29) led us to focus our investigation on the cellular localization of EP 3␣ and EP 4 receptors, and the remaining studies concentrated on these two receptor subtypes. Expression of EP 3␣ and EP 4 Immunoreactive Protein in Newborn Brain Subcellular Fractions-Immunoblot analysis revealed immunoreactive bands in plasma and nuclear fractions of similar molecular masses (EP 3␣ , 60 kDa; EP 4 , 63 kDa) (Fig. 2).

RESULTS AND DISCUSSION
Indirect Immunofluorescence of EP 3␣ and EP 4 Receptors in Porcine Cerebral Vessel Endothelial Cells-Because cerebral microvessels contain a number of high affinity PGE 2 binding sites (20), primary cultures of newborn pig brain microvascular endothelial cells were used to study the intracellular distribution of EP 3␣ and EP 4 receptors by confocal microscopy. No fluorescence was detected in the absence of the primary antibodies (Fig. 3a). Immunoreactivity for both receptor subtypes was detected in the plasma membrane (Figs. 3b and 4a), in the cytoplasm, and at the nucleus (Figs. 3c and 4b). EP 3␣ specific fluorescence in the nuclear envelope appeared as a perinuclear halo (Fig. 3c); the latter was more prominent than that of EP 4 receptors (Fig. 4b). The cells were stained with DiOC 6 (3) to identify intracellular membranes, mainly ER (Figs. 3d and 4c). Merging the images from EP 3␣ or EP 4 specific red immunofluorescence with DiOC 6 (3) green staining revealed that EP receptors colocalized on intracellular membranes as indicated by the bright yellow-orange fluorescence (Fig. 3e and 4d); however, the stains did not fully converge, suggesting distinct sites particularly evident in the perinuclear region. Cells were also stained with a nuclear stain, Sytox Green (Fig. 3g). A transverse section (Z-section) of this image superimposed with that of the EP 3␣ immunoreative staining (Fig. 3h) in the same cell revealed that the immunoreactivity was perinuclear and not intranuclear (Fig. 3i); similar results were obtained using the EP 4 antibody (data not shown). No immunofluorescence was detected when the antibodies were preincubated with their cognate peptides (Figs. 3f and 4e). EP 3␣ and EP 4 immunoreactivity in co-localization with DiOC 6 (3) staining apparently in the nucleoli of a few endothelial cells (Figs. 3, c-e and 4, b-d) was not consistently observed (Fig. 3h) and remains unexplained at this point.
Indirect Immunofluorescence of EP 3␣ and EP 4 Receptors in

FIG. 2. Immunoblot of EP 3␣ and EP 4 receptor proteins (see arrows) in plasma membrane (P) and nuclear membrane (N) fractions from newborn pig brain.
Top and bottom arrows point to EP 4 and EP 3␣ bands, respectively; only one band was detected in the range of interest (50 -65 kDa).

FIG. 3. Confocal microscopy of porcine cerebral microvascular endothelial cells immunofluorescently stained for EP 3␣ receptors.
Cells were subjected to indirect immunofluorescent staining using affinity purified rabbit anti-peptide EP 3␣ antibodies followed by Texas Red-conjugated anti-rabbit IgG. a, Texas Red-conjugated IgG alone; note absence of immunofluorescence. EP 3␣ immunoreactivity on plasma membrane (b) and nuclear membrane and cytoplasm (c); note perinuclear halo in more extensively permeabilized cells (see "Experimental Procedures"). d, DiOC 6 (3), intracellular membranes (mostly endoplasmic reticulum) stain; e, superimposed images of panels c and d; note red (EP 3␣ immunoreactivity) perinuclear halo. f, anti-EP 3␣ in the presence of cognate peptide, (10 g/ml), note lack of immunofluorescence; g, Sytox Green nucleus stain; h, EP 3␣ immunoreactivity; i, Z section of superimposed images of panels g and h.
HEK 293 Cells-To assess whether this perinuclear distribution of EP 3␣ and EP 4 receptors applies generally to cells, the localization of these receptors was studied after transfection of cDNA for EP 3␣ and EP 4 into HEK 293 cells that do not normally express prostanoid receptors (39); ectopically expressed EP receptors in HEK 293 cells bind PGE 2 and are functional (39,40). Immunoreactivity for EP 3␣ and EP 4 receptors was seen on the plasma membrane (Fig. 5, c and g) and perinuclear area, which are in proximity to each other in the transfected HEK 293 cells (Fig. 5d, h), which are relatively small and contain limited cytoplasm compared with endothelial cells. As expected, no immunofluorescence was detected in the wild-type cells (Fig. 5a) or after preincubation of the antibodies with their cognate peptide epitopes (Fig. 5 f and j).
Immunogold Labeling of EP 3␣ and EP 4 Receptors-Thus far, indirect immunofluorescence studies revealed a perinuclear localization of EP 3␣ and EP 4 receptors. To distinguish the nuclear envelope, immunoelectron microscopy of porcine cerebrovascular endothelial cells was performed and confirmed that EP 3␣ and EP 4 immunoreactivity was indeed at the nuclear envelope (Fig. 6, c and f). As expected, these receptors were detected on plasma membranes (Fig. 6 b and e) and Golgi vesicles (Fig. 6d). No immunogold staining was observed in the absence of the primary antibodies (Fig. 6a) or in the nucleoli of cells (data not shown). EP 3␣ and EP 4 nuclear envelope immu-nogold staining was detected in the majority of cells observed (95% of cells, over 100 cells observed in each case).

Effects of Stimulation of Porcine Cerebrovascular Endothelial Cells Nuclear EP Receptors on iNOS Gene Transcription-
Recent studies have shown that endogenous PGE 2 has a stimulatory effect on inducible nitric-oxide synthase (iNOS) (41,42). We tested whether the stimulation of nuclear EP receptors by prostaglandin analogs may affect iNOS transcription, as determined by dot hybridization of RNA studies. Stimulation of intact nuclei isolated from primary cultures of porcine brain endothelial cells with EP 3 receptor agonist M&B 28,767 (0.1 M) increased transcription of iNOS (Fig. 7a) to a greater extent than after stimulation of whole cells.
Effects of Stimulation of Rat Liver Nuclear EP Receptors on Ca 2ϩ Transients-The nuclear envelope contains distinct nuclear calcium pools that play crucial roles in major nuclear FIG. 4. Confocal microscopy of porcine cerebral microvascular endothelial cells immunofluorescently stained for EP 4 receptors. Cells were subjected to indirect immunofluorescent staining using affinity purified rabbit anti-peptide EP 4 antibodies followed by Texas Red-conjugated anti-rabbit IgG. EP 4 immunoreactivity on plasma membrane (a), and nuclear membrane and cytoplasm (b); perinuclear halo is noted in more extensively permeabilized cells (see "Experimental Procedures"). c, DiOC 6 (3), intracellular membranes (mostly endoplasmic reticulum) stain; d, superimposed images of panels b and c, note red (EP 4 immunoreactivity) perinuclear halo; e, anti-EP 4 in the presence of cognate peptide (10 g/ml), note absent immunofluorescence.
FIG. 5. EP 3␣ and EP 4 immunofluorescent staining of overexpressing clones of human EP 3␣ or human EP 4 receptors in HEK 293 cells. a, cells transfected with vector alone, note absence of immunofluorescence; b, nuclear stain (DAPI) of cells from panel a. EP 3␣ immunoreactivity on plasma membrane (c) and nuclear membrane (d); note the perinuclear halo in more permeabilized cells (see "Experimental Procedures"). e, nuclear stain (DAPI) of cells from panel d. f, anti-EP 3␣ in the presence of cognate peptide (10 g/ml), note the lack of immunofluorescence. EP 4 immunoreactivity on plasma membrane (g) and nuclear membrane (h), note the perinuclear halo in more permeabilized cells (see "Experimental Procedures"). i, nuclear stain (DAPI) of cells from panel h; j, anti-EP 4 in the presence of cognate peptide (10 g/ml).
functions including gene transcription (13). The amplitude and duration of calcium signals have been shown to control differential activation of transcription factors (43). In addition, Ca 2ϩ can activate iNOS independent of protein kinases C and A (44). We tested whether stimulation of nuclear EP 3 receptors with prostaglandin analogs could affect calcium concentrations in isolated nuclei of liver; stimulation of EP 4 receptors could not be performed because of lack of availability of specific agonists. Application of M&B 28,767 (1 M, an EP 3 -selective agonist) to intact isolated nuclei caused rapid nuclear uptake of 45 Ca 2ϩ (Fig. 7b). In addition, this EP 3 agonist produced a dose-dependent increase in rat liver nuclear calcium as determined by fura-2/AM, a fluorescent dye which localizes in the nuclear envelope space (45) (Fig. 7c); M&B 28,767 (1 M) was nearly as effective as the nonselective EP agonist 16,16-dimethyl PGE 2 (1 M).
We determined whether the nuclear calcium uptake evoked by the EP 3 agonist M&B 28,767 was dependent on G proteins. EP 3 couples to G i or G o (46), which are known to affect Ca 2ϩ mobilization (46,47); such G proteins are detected in rat liver nuclei (38). Because PTX causes these G proteins to lose their ability to associate with receptors, we tested the effects of PTX on M&B 28,767-induced Ca 2ϩ transients. Pretreatment of isolated nuclei with PTX markedly attenuated the stimulatory effect of M&B 28,767 (1 M) on intranuclear calcium levels, suggesting the involvement of a PTX-sensitive G protein in mediating the effects of nuclear EP 3 receptors (Fig. 7d). In contrast, M&B 28,767 did not inhibit forskolin-stimulated cAMP formation (data not shown). These findings are consistent with coupling of nuclear EP 3 receptors to G proteins which may directly control Ca 2ϩ channels independently of second messengers, as mostly reported for G i (47,48).
In conclusion, the data presented provides the first clear evidence for the presence of the G protein-coupled receptors, EP 3␣ and EP 4 at nuclear membranes of native tissues as well as primary and transfected cells. Furthermore, these receptors seem to be functional as revealed by increased iNOS transcription and nuclear calcium by EP 3 agonist, M&B 28,767, which also appears to involve PTX-sensitive G proteins. The plasma and nuclear membrane EP 3␣ as well as EP 4 receptors appear to be related because they had similar molecular weights, binding kinetics, ligand binding properties, and immunoreactivity. Also, the comparable distribution of ectopically expressed EP 3␣ and EP 4 receptors in HEK 293 cells suggested that the plasma and nuclear membrane EP receptors may be alike. Radioligand binding studies have identified the presence of two other classes of G protein-coupled receptors at the nuclear membrane, the muscarinic acetylcholine (49) and angiotensin II receptors, AT 1 (23); but the AT 1 receptor can be detected in the nucleus only after stimulation by angiotensin II (50). Other prostanoids, namely PGD 2 , its metabolite PGJ 2 , and PGI 2 can activate the peroxisome proliferator-activated receptors (PPARs) that are members of the nuclear receptor superfamily, but PPARs are not responsive to PGE 2 (51,52). However, the presence of EP 1 receptors at the nuclear membranes has recently been suggested (19) albeit its mechanism of action is not clear.
In the newborn brain and cerebral microvasculature, PGE 2 receptors and associated functions at the plasma membrane are down-regulated (2,20). On the other hand, PGE 2 plays a role in neuroprotection by acting on EP 2 and perhaps EP 4 (53). PGE 2 also increases the expression of nitric-oxide synthase via stimulation of EP 3 receptors in the neonate (54). In addition, the perinuclear cyclooxygenase-1 and -2 (7) can produce prostanoids that can act at the nuclear level (11,12) and modulate transcription of genes, as had been speculated for iNOS (41). The present discovery of nuclear EP 3 and EP 4 receptors proposes new avenues for the intracellular actions of prostanoids, which may also explain certain effects of PGE 2 especially when plasma membrane EP receptors are barely detectable. Further studies are needed to clarify the detailed signal transduction mechanisms involved in this action of prostaglandins via nuclear EP receptors.