Prostaglandin E2 Receptors EP2 and EP4 Are Down-regulated during Differentiation of Mouse Osteoclasts from Their Precursors*

Prostaglandin E2 (PGE2) has been proposed to be a potent stimulator of bone resorption. However, PGE2 itself has been shown to directly inhibit bone-resorbing activity of osteoclasts. We examined the role of PGE2 in the function of mouse osteoclasts formed in vitro. Bone marrow macrophage osteoclast precursors expressed PGE2 receptors EP1, EP2, EP3β, and EP4, and the expression of EP2 and EP4 was down-regulated during osteoclastic differentiation induced by receptor activator of NF-κB ligand and macrophage colony-stimulating factor. In contrast, functional EP1 was continuously expressed in mature osteoclasts. PGE2 as well as calcitonin caused intracellular Ca2+ influx in osteoclasts. However, PGE2 and 17-phenyltrinol-PGE2 (an EP1 agonist) failed to inhibit actin-ring formation and pit formation by osteoclasts cultured on dentine slices. When EP4 was expressed in osteoclasts using an adenovirus carrying EP4 cDNA, both actin-ring and pit-forming activities of osteoclasts were inhibited in an infectious unit-dependent manner. Treatment of EP4-expressing osteoclasts with PGE2 further inhibited their actin-ring and pit-forming activities. Such inhibitory effects of EP4-mediated signals on osteoclast function are similar to those that are calcitonin receptor-mediated. Thus, osteoclast precursors down-regulate their own EP2 and EP4 levels during their differentiation into osteoclasts to escape inhibitory effects of PGE2 on bone resorption.

Bone-resorbing osteoclasts form unique cellular structures such as clear zones and ruffled borders toward the bone surface (12). The clear zone, also called the "actin ring," consists of a ring-like alignment of filamentous actin (F-actin) dots and surrounds the ruffled border, from which protons and lysosomal enzymes are secreted into the resorption lacunae (13). Calcitonin (CT), a bone resorption-inhibiting hormone, disrupts actin rings and inhibits the pit-forming activity of osteoclasts cultured on bone or dentine slices (14,15). Osteoclasts express abundant CT receptors, which are coupled to G␣ s and G␣ q proteins (16). We have previously shown that signals mediated by adenosine 3Ј,5Ј-cyclic monophosphate (cAMP)-dependent protein kinase A (PKA) play important roles in CT-induced inhibition of pit formation and actin ring formation by osteoclasts cultured on dentine slices (17).
Prostaglandin E 2 (PGE 2 ) has diverse biological activities in a variety of tissues (18). The actions of PGE 2 in the target cells are mediated by four different G protein-coupled receptor subtypes, EP1, EP2, EP3, and EP4 (19). The EP subtypes differ in tissue distribution, ligand binding affinity, and coupling to intracellular signaling pathways. The signal of EP1 predominantly increases intracellular Ca 2ϩ and activates protein kinase C (PKC) (19). EP2 and EP4 activate G␣ s followed by increases in adenylate cyclase activity, adenosine cAMP production, and PKA activity in the target cells. In contrast, EP3 acts via G␣ i to inhibit cAMP generation. PGE 2 has been proposed to be a potent stimulator of bone resorption involved in inflammatory diseases such as rheumatoid arthritis (20 -23). However, the mechanism of PGE 2 -induced bone resorption has not yet been clearly explained. Like other osteotropic factors, PGE 2 stimulates expression of RANKL in osteoblasts (21,24). Among PGE 2 receptor subtypes, EP4 has been shown to mainly mediate PGE 2 -induced RANKL expression in osteoblasts (24). In addition, PGE 2 directly and synergistically with RANKL and M-CSF stimulates the differentiation of mouse bone marrow-derived macrophages (BMM), osteoclast precursors, into osteoclasts (25,26). We have recently shown that the synergistic effect of PGE 2 on RANKL-induced osteoclast differentiation is mediated through EP2 and EP4. Transforming growth factor-␤-activated kinase 1 (TAK1) acts as an adapter molecule linking PKA-induced signals and RANKL-induced signals in osteoclast precursors (27). Thus, PGE 2 stimulates osteoclastic bone resorption through two pathways, the induction of RANKL expression by osteoblasts and direct enhancement of RANKL-induced osteoclast differentiation of the precursors.
Paradoxically, PGE 2 has been shown to inhibit bone-resorbing activity of osteoclasts when added to osteoclasts cultured on bone or dentine slices (28,29). Lerner et al. (30) also reported that PGE 2 transiently inhibited bone resorption and the release of lysosomal enzymes in mouse calvarial cultures. This inhibitory effect of PGE 2 on bone resorption may be due to a direct activation of adenylate cyclase in mature osteoclasts, thereby mimicking the effect of CT. Thus, the role of PGE 2 in the function of mature osteoclasts is still a matter of controversy.
In the present study we explored the role of PGE 2 in osteoclast function. Osteoclast precursors of BMM expressed EP1, EP2, EP3␤, and EP4, but mature osteoclasts expressed only EP1. PGE 2 affected neither the actin ring-forming nor the resorption pit-forming activity of osteoclasts cultured on dentine slices. Forced expression of EP4 in osteoclasts suppressed both activities of osteoclasts. The inhibitory effects of the EP4 expression on osteoclast function were quite similar to those of CT. These results suggest that the down-regulation of EP2 and EP4 in osteoclast precursors during osteoclastic differentiation is an important event for escaping the inhibitory effect of PGE 2 on bone resorbing activity. Cultures of Bone Marrow-derived Macrophages-BMM were prepared as osteoclast precursors as described previously (31). Briefly, bone marrow cells were obtained from tibiae of 5-8-week-old male ddY mice (Shizuoka Laboratories Animal Center, Shizuoka, Japan). All procedures for animal care were approved by the Animal Management Committee of Matsumoto Dental University. Bone marrow cells were suspended in ␣-modified minimum essential medium (␣-MEM, Sigma) supplemented with 10% fetal bovine serum (FBS) (JRH Biosciences, Lenexa, KS) on 60-mm diameter dishes for 16 h in the presence of M-CSF (100 ng/ml). Then, nonadherent cells were harvested and further cultured for 2 days with M-CSF (50 ng/ml). The adherent cells, most of which expressed macrophage-specific antigens such as Mac-1, Moma-2, and F4/80, were used as BMM.

Chemicals-Recombinant
Survival Assay of Purified Osteoclasts-Osteoclasts were generated in co-cultures of mouse primary osteoblasts and bone marrow cells in collagen gel-coated dishes as described previously (32). Primary osteoblasts and bone marrow cells were co-cultured in ␣-MEM supplemented with 10% FBS in 100-mm tissue culture dishes (Corning Inc., Corning, NY) precoated with type I-collagen gel (Nitta Gelatin, Osaka, Japan) in the presence of 1␣,25(OH) 2 D 3 (10 Ϫ8 M). After the cells were cultured for 7 days, all cells were recovered from the dishes by treatment with 0.2% collagenase. The purity of osteoclasts in this preparation was about 5%. To purify osteoclasts the crude osteoclast preparation was plated in 100-mm tissue culture dishes. After the cells were cultured for 6 h, osteoblasts were removed by treatment of cells with phosphate-buffered saline (PBS(Ϫ)) containing 0.001% Pronase E and 0.02% EDTA for 5 min. The purity of osteoclasts in this preparation was about 95%. For the osteoclast survival assay, purified osteoclasts were further incubated for the indicated periods in the presence or absence of test chemicals and stained for tartrate-resistant acid phosphatase (TRAP, a marker enzyme of osteoclasts) as described. TRAP-positive multinu-clear cells containing more than three nuclei were counted as viable osteoclasts.
Actin Ring and Pit Formation by Osteoclasts Cultured on Dentine Slices-The crude osteoclast preparation recovered from the co-cultures was placed on dentine slices (4 mm-diameter) for 2 h in 96-well plates. Dentine slices were transferred into 48-well plates and further cultured for 46 h in the presence or absence of calcitonin or PGE 2 or 17-phenyltrinol-PGE 2 . For F-actin staining, cells cultured on dentine slices were fixed with 3.7% formalin diluted with phosphate-buffered saline and rinsed with phosphate-buffered saline containing 0.1% Triton X-100 (Sigma) for 1 min. F-actin in the cells was labeled with rhodamineconjugated phalloidin (3 units/ml, Molecular Probes, Inc.) for 2 h and observed using a fluorescence microscope. For the pit formation assay, dentine slices were recovered from the culture. The surface of the dentine slices was rubbed strongly with a cotton bud to remove all cells on the slices. Ten microliters of Mayer's hematoxylin (Wako Pure Chemical Industries) was placed on the surface using surface tension for 35-45 s to visualize resorption pits. The percentage of resorbed area relative to the total area of dentine slices was determined using an image analysis system. The results were expressed as the mean Ϯ S.D. of triplicate cultures.
Real-time PCR-cDNA (20 ng) was utilized for 40-cycle 2-step PCR in an DNA Engine Opticon system (MJ Japan Ltd., Tokyo) using a SYBR Green qPCR kit (Finnzymes oy, Espoo, Finland) and 300 nM primers. The amplicon size and reaction specificity was confirmed by agarose gel electrophoresis. Each transcript in the samples was assayed three times, and the fold-change ratios between experimental and control samples for each gene used in the analysis were calculated. The results were expressed as the mean Ϯ S.D. of three assays. The sequences of primers for EP receptor isoforms were as follows: EP2, 5Ј-ATGCTCCT-GCTGCTTATCGT-3Ј (forward) and 5Ј-TAATGGCCAGGAGAAT-GAGG-3Ј (reverse); EP4, 5Ј-CCATCGCCACATACATGAAG-3Ј (forward) and 5Ј-TGCATAGATGGCGAAGAGTG-3Ј (reverse).
Assay of cAMP Production-To measure the amount of cAMP produced, cells cultured in 48-well plates (BMM, 8 ϫ 10 4 /well; purified osteoclasts, 2 ϫ 10 4 cells/well) were preincubated for 5 min at 37°C in ␣-MEM containing 1 mM IBMX, then incubated for 5 min at 37°C with elcatonin (10 Ϫ10 M) or PGE 2 (10 Ϫ6 M). Cells were washed with ice-cold phosphate-buffered saline containing 1 mM IBMX and then lysed. The content of intracellular cAMP was determined using a cAMP enzyme immunoassay kit (Amersham Biosciences) according to the manufacturer's instructions. The results were expressed as the mean Ϯ S.D. of quadruplicate cultures.
Measurement of Intracellular Ca 2ϩ Influx-The effects of PGE 2 and calcitonin on intracellular Ca 2ϩ influx in mature osteoclasts and in BMM were measured using a confocal laser scanning microscope (LSM510, Carl Zeiss, Jena, Germany) according to the methods described previously (27). Osteoclasts or BMM were incubated in a glass-bottom dish (ASAHI TECHNO GLASS Corp., Tokyo) for 6 h. Cells were incubated in the presence of 5 M fluo-4 AM, 5 M Fura Red AM, and 0.05% pluronic F127 for 30 min in serum-free Dulbecco's modified Eagle's medium (Sigma). Cells loaded with these dyes were washed twice with ␣-MEM and postincubated in ␣-MEM containing 10% FBS. Cells were further washed three times with Hanks' balanced salt solution and then excited at 488 nm, and emission at 505-530 nm for fluo-4 and 600 -680 nm for Fura Red were acquired simultaneously at 2-s intervals. The ratio of the fluorescence intensity of the fluo-4 to Fura Red was calculated to estimate intracellular Ca 2ϩ concentrations in single cells.
Construction of Adenovirus-A pBluescript vector containing the entire open reading frame of EP4 was a kind gift of Dr. S. Narumiya (Kyoto University, Japan). The EP4 cDNA was inserted into the ApaI-XbaI site of a p-shuttle vector (BD Clontech Laboratories, Inc. Palo Alto, CA), and the DNA fragment was inserted into the PI-SceI/I-CeuI site of a pAdenoX vector (BD Clontech Laboratories, Inc.). Two micrograms of the pAdenoX vector containing EP4 cDNA linearized by PacI was transfected into HEK293 cells (ATCC, Manassas, VA) using TransFast transfection reagents (Promega Corp., Madison, WI). Amplified crude viral stocks were purified by CsCl gradient ultracentrifugation and used for the infection. Adenovirus carrying ␤-galactosidase cDNA (Ad-LacZ) was also generated and used for the control infection. The infectious units (ifu) were determined using an Adeno-X rapid titer kit (BD Clontech Laboratories, Inc.).
Infection of Osteoclasts with Adenovirus Carrying EP4 cDNA-Infection of osteoclasts with Ad-LacZ or adenovirus carrying EP4 cDNA (Ad-EP4) was carried out according to the methods described previously (33). Primary osteoblasts and bone marrow cells were co-cultured for 4 days in collagen gel-precoated dishes (100-mm diameter). Co-cultures were then incubated for 1 h with 2 ml of ␣-MEM containing the indicated ifu of recombinant adenoviruses and 10% FBS and further incubated with 15 ml of ␣-MEM with 10% FBS in the presence of 1␣, 25(OH) 2 D 3 (10 Ϫ8 M) for an additional 48 h.
Statistical Analysis-For the determination of cAMP production, data were expressed as the mean Ϯ S.D. of quadruplicate cultures. For the real-time PCR analysis, osteoclast survival assay, and pit forming assay, representative data were expressed as the mean Ϯ S.D. of three determinations. All experiments were independently repeated three times, and similar results were obtained. Statistical analyses were performed using Student's t test.

Expression of PGE 2 Receptor Subtypes in Osteoclast Precursors
and Mature Osteoclasts-We first analyzed the expression of PGE 2 receptor subtypes in BMM, osteoclast precursors, and mature osteoclasts using RT-PCR. BMM were prepared from bone marrow cultures treated with M-CSF (see Fig. 1C). Purified mature osteoclasts were prepared from collagen-gel co-cultures of calvarial osteoblasts and bone marrow cells treated with 1␣,25(OH) 2 D 3 . BMM expressed EP1, EP2, EP3␤, and EP4 mRNAs, whereas purified osteoclasts expressed only EP1 mRNA (Fig. 1A). CT receptor mRNA was expressed in osteoclasts but not in the precursors. In our experiments co cultures were treated for 7 days with 1␣,25(OH) 2 D 3 to induce formation of osteoclasts. Ikegami et al. (34) showed that treatment of mouse peritoneal macrophages with PGE 2 down-regulated EP4 mRNA expression within 12 h in a cAMP/PKA-dependent manner. These results raise the possibility that 1␣,25(OH) 2 D 3 or endogenously produced PGE 2 may down-regulate EP4 and EP2 expression in osteoclast precursors. Next, we examined the effects of PGE 2 and 1␣,25(OH) 2 D 3 on the expression of EP2 and EP4 mRNAs in BMM in the presence of M-CSF (Fig. 1B). PGE 2 (10 Ϫ5 M) and 1␣,25(OH) 2 D 3 (10 Ϫ8 M) significantly decreased the expression of EP4 mRNA but not EP2 mRNA in BMM (Fig. 1B). Treatment of BMM with RANKL induced osteoclastic differentiation of BMM in the presence of M-CSF. Expression of both EP2 and EP4 mRNAs was significantly down-regulated in BMM treated with RANKL for 24 h (Fig. 1B). Parallel experiments showed that the purified osteoclast preparation and the BMM culture treated with RANKL for 72 h contained more than 95 and 65% TRAP-positive cells, respectively (Fig. 1C). The BMM culture treated with RANKL for 24 h contained less than 2% TRAP-positive cells (data not shown). These results suggest that EP2 and EP4 are down-regulated in osteoclasts during osteoclastic differentiation.
Function of EP Subtypes Expressed in Osteoclast Precursors and Mature Osteoclasts-We next examined the effects of PGE 2 on cAMP production in BMM and mature osteoclasts in com- parison with those of CT. PGE 2 (10 Ϫ6 M) but not CT (10 Ϫ10 M) significantly stimulated cAMP production in BMM ( Fig. 2A,  upper panel). In contrast, CT (10 Ϫ10 M) but not PGE 2 (10 Ϫ6 M) stimulated cAMP production in purified osteoclasts ( Fig. 2A,  lower panel). These results suggest that functional EP2 and EP4 are expressed in osteoclast precursors but not in mature osteoclasts.
EP1 has been shown to be coupled to Ca 2ϩ signals, which induce the formation of inositol 1,4,5-triphosphate and diacylglycerol (18). CT receptors activate not only G␣ s but also G␣ q , the signals of which cause an increase in cytosolic Ca 2ϩ . We next examined whether EP1 is functional in mature osteoclasts and BMM. Treatment of osteoclasts with PGE 2 at 10 Ϫ5 M sharply increased the intracellular Ca 2ϩ influx in osteoclasts (Fig. 2, B and C). PGE 2 at 10 Ϫ6 M, a concentration that was sufficient to activate adenylate cyclase in BMM, did not induce clear Ca 2ϩ influx in osteoclasts (data not shown). Ca 2ϩ influx in osteoclasts was also induced by the addition of CT (10 Ϫ10 M) (Fig. 2D). PGE 2 at 10 Ϫ5 M increased the Ca 2ϩ influx in BMM as well (Fig. 2E). These results suggest that functional EP1 is expressed in both mature osteoclasts and BMM.
Role of EP1 in Osteoclast Function-We previously showed that purified osteoclasts spontaneously died within 48 h due to apoptosis (32). Not only RANKL but also CT was shown to stimulate the survival of osteoclasts (35). We next examined the effects of PGE 2 on the survival of osteoclasts in comparison with those of RANKL and CT (Fig. 3). Purified osteoclasts gradually died during the incubation for 36 h, and RANKL (100 ng/ml) and CT (10 Ϫ9 M) significantly promoted the survival of osteoclasts (Fig. 3, A and B). However, PGE 2 (10 Ϫ6 M) showed no effect on the survival (Fig. 3, A and B). The effect of a higher concentration of PGE 2 (10 Ϫ5 M) on the survival of osteoclasts was also examined because PGE 2 at the concentration of 10 Ϫ5 M increased the intracellular Ca 2ϩ influx in osteoclasts and BMM (Fig. 3C). The survival of osteoclasts was not supported The ratio of the fluorescence intensity of the fluo-4 to Fura Red was calculated to estimate intracellular Ca 2ϩ influx in single osteoclasts treated with PGE 2 (C) and CT (D) and in single BMM treated with PGE 2 (E). The results represent calcium signals in five to six single cells treated with PGE 2 or CT. by either PGE 2 or 17-phenyltrinol-PGE 2 (an EP1 agonist) even at 10 Ϫ5 M. These results suggest that EP1-mediated signals do not affect the osteoclast survival.
When osteoclasts were cultured for 48 h on dentine slices, the osteoclasts formed actin rings and resorption pits on the slices (Fig. 4, A and B). Treatment of osteoclasts with CT (10 Ϫ10 M) strongly inhibited their actin ring-and pit-forming activities. In contrast, neither PGE 2 (10 Ϫ7 -10 Ϫ5 M) nor 17-phenyltrinol-PGE 2 (an EP1 agonist, 10 Ϫ5 M) affected actin ring formation or pit formation by osteoclasts cultured on dentine slices (Fig. 4, A  and B). Thus, mature osteoclasts expressed functional EP1, but EP1-mediated signals did not appear to induce appreciable effects on osteoclast function in our assay systems.
Inhibition of Osteoclast Function by Forced Expression of EP4 -Our final question was why osteoclast precursors lose EP2 and EP4 during the differentiation into osteoclasts; that is, what are the functional consequences of the disappearance of EP2 and EP4 in osteoclasts. To address this question, EP4 was expressed in osteoclasts using Ad-EP4 (Fig. 5). RT-PCR analysis showed that osteoclasts purified from the co-cultures that had been infected with Ad-EP4 at 10 9 ifu expressed high levels of EP4 mRNA, but those infected with Ad-LacZ did not (Fig. 5A). To examine whether EP4 expressed in the infected osteoclasts is functional, the concentration of intracellular cAMP was determined in Ad-EP4-and Ad-LacZ-infected osteoclasts in the presence or absence of exogenous PGE 2 (Fig. 5B). Intracellular cAMP levels were significantly higher in osteoclasts infected with Ad-EP4 than in those infected with Ad-LacZ even in the absence of exogenously added PGE 2 (Fig. 5B). PGE 2 (10 Ϫ6 M) significantly increased cAMP production in Ad-EP4-infected osteoclasts but not in Ad-LacZ-infected osteoclasts. These results suggest that the EP4 expressed in osteoclasts is functionally active.
We finally examined the effects of the forced expression of EP4 on the function of osteoclasts (Fig. 6). Crude osteoclast preparations from co-cultures infected with Ad-EP4 or Ad-LacZ were placed on dentine slices, and the cells were further cultured for 48 h in the presence or absence of PGE 2 or CT. Similar numbers of TRAP-positive cells were observed on the slices cultured with cells expressing Ad-EP4 and Ad-LacZ (Fig. 6A). However, the number of actin rings formed in EP4-expressing osteoclasts was markedly decreased. When PGE 2 (10 Ϫ6 M) was added to the culture, actin rings in osteoclasts expressing Ad-LacZ remained unchanged, but those in osteoclasts expressing Ad-EP4 disappeared completely (Fig. 6A). The pit-forming activity of osteoclasts was not affected by Ad-LacZ but was markedly inhibited by Ad-EP4 infection in an ifu-dependent manner (Fig. 6, B and C). The addition of PGE 2 (10 Ϫ6 M) to the cultures of osteoclasts infected with Ad-LacZ had no inhibitory effects on the pit-forming activity (Fig. 6, B and C). However, PGE 2 significantly suppressed the pit-forming activity of osteoclasts infected with Ad-EP4 at 1 ϫ 10 8 ifu. CT (10 Ϫ10 M) completely inhibited the pit-forming activity of osteoclasts infected with either Ad-LacZ or Ad-EP4 (Fig. 6C). Thus, EP4-mediated signals were similar to those mediated by CT receptors in osteoclasts.

DISCUSSION
In the present study we examined the role of PGE 2 in the function of mouse osteoclasts formed in vitro. BMM osteoclast precursors expressed EP1, EP2, EP3␤, and EP4, and expression of EP2 and EP4 was down-regulated during the differentiation of BMM into osteoclasts. In contrast, functional EP1 was continuously expressed in mature osteoclasts. PGE 2 as well as CT induced intracellular Ca 2ϩ influx in osteoclasts. However, CT but not PGE 2 inhibited formation of actin rings and resorption pits by osteoclasts cultured on dentine slices. When EP4 was expressed in osteoclasts using Ad-EP4, the actin ring-and pit-forming activities of osteoclasts were strongly inhibited in an ifu-dependent manner. PGE 2 further inhibited both of these activities in osteoclasts infected with Ad-EP4. Thus, mouse osteoclasts formed in vitro expressed functionally active EP1 but not EP2 and EP4. However, EP1mediated signals did not appear to regulate the bone-resorbing activity of osteoclasts.
RT-PCR analysis showed that the level of expression of EP2 and EP4 mRNAs in mature osteoclasts was much lower than that in BMM (Fig. 1). It was reported that PGE 2 suppressed After cells were cultured for 46 h the dentine slices were recovered. A, some dentine slices were processed for F-actin staining (upper panels, bar ϭ 25 m). The other slices were processed for Mayer's hematoxylin staining to visualize resorption pits (lower panels, bar ϭ 100 m). B, percentages of resorbed area on dentine slices. Percentages of resorbed area relative to the total surface were determined using an image analysis system. The values were expressed as the mean Ϯ S.D. of triplicate cultures. *, Significantly different from the control cultures; p Ͻ 0.01. the expression of EP4 mRNA in mouse peritoneal macrophages in a cAMP/PKA-dependent manner (34). In our experiments PGE 2 and 1␣,25(OH) 2 D 3 significantly decreased the expression of EP4 mRNA but not EP2 mRNA in BMM within 24 h (Fig.  1). On the other hand expression of both EP2 and EP4 mRNAs was down-regulated in BMM treated with RANKL for 24 h. TRAP-positive osteoclasts appeared on day 3 but not day 1 in BMM cultures treated with RANKL and M-CSF. These results suggest that osteotropic factors can modulate the expression of EP subtypes in osteoclast precursors, but down-regulation of EP2 and EP4 in osteoclasts is a consequence of the decision to differentiate into osteoclasts, probably for other appropriate purposes.
Osteoclasts expressed functionally active EP1 as well as CT receptors. Treatment of osteoclasts with PGE 2 at 10 Ϫ5 M or with CT sharply increased the intracellular Ca 2ϩ influx to a similar extent (Fig. 2). Several studies have shown that intracellular calcium is involved in calcitonin-induced inhibition of osteoclast function (36 -38). Zhang et al. (39) reported that the increase in cytosolic calcium induced by CT decreased tyrosine phosphorylation of Pyk2, which in turn disrupted actin rings in osteoclasts. Activation of protein kinase C but not PKA was also shown to mediate the inhibitory effect of CT on human osteoclast function (40). In our experiments actin ring formation and pit formation by osteoclasts were strongly inhibited by calcitonin but not by PGE 2 or 17-phenyltrinol-PGE 2 , an EP1 agonist (Fig. 4). Survival of osteoclasts was also supported by CT but not by PGE 2 or 17-phenyltrinol-PGE 2 (Fig. 3). These results suggest that osteoclasts express functionally active EP1, but EP1-mediated signals do not play appreciable roles in osteoclast function, at least in our assay systems.
CT receptors are coupled to G␣ s and G␣ q proteins that activate cAMP/PKA and Ca 2ϩ /PKC signals, respectively. Nicholson et al. (41) reported that forskolin, an activator of adenylate cyclase, induced a synergistic effect with CT in stimulating cAMP production in isolated rat osteoclasts and augmented the hypocalcemic response to CT in vivo. We also showed that cAMP/PKA signals were involved in the CT-induced inhibition of actin ring formation and pit formation by osteoclasts (17). Kanaoka et al. (42) reported that CT inhibited nitric oxideinduced apoptosis of osteoclasts in vitro through cAMP/PKA signals. Thus, cAMP/PKA signals play important roles in the inhibition of osteoclast function and the promotion of osteoclast survival induced by CT.
One important question we addressed was why osteoclast progenitors lose EP2 and EP4 in the process of the differentiation into osteoclasts. Forced expression of EP4 in osteoclasts strongly inhibited actin ring formation and pit formation even in the absence of exogenous addition of PGE 2 , suggesting that EP4 expressed in osteoclasts may respond to PGE 2 produced endogenously. In fact, the cAMP concentration in osteoclasts infected with Ad-EP4 was significantly higher than that in osteoclasts infected with Ad-LacZ (Fig. 5). The addition of PGE 2 to the cultures of EP4-expressing osteoclasts enhanced cAMP production and inhibited the actin ring-forming activity and pit-forming activity of those osteoclasts. The EP4-induced inhibitory effects on the osteoclast function were similar to CT receptor-induced ones. These results further support the notion that cAMP/PKA but not Ca 2ϩ /PKC signals inhibit osteoclast function. Down-regulation of EP2 and EP4 in osteoclast precursors during osteoclastic differentiation appears to be an important event for escaping the inhibitory effect of PGE 2 on the bone-resorbing activity (Fig. 7).
Several reports have indicated that PGE 2 acts as a direct inhibitor of bone resorption in isolated rat, chick, and rabbit FIG. 6. Effect of forced expression of EP4 in osteoclasts on their actin ring-and pit-forming activities. Co-cultures grown on collagen gel-coated dishes (100-mm diameter) were infected with either Ad-LacZ or Ad-EP4 on day 4 at doses of 10 8 or 10 9 ifu. Osteoclast preparations were cultured on dentine slices for 48 h in the presence or absence of PGE 2 (10 Ϫ6 M) or CT (10 Ϫ10 M). A, some dentine slices were processed for TRAP and F-actin staining. Bars, 25 m. B and C, the other slices were processed for Mayer's hematoxylin staining to visualize resorption pits. B, resorption pits on dentine slices recovered from Ad-LacZ-or Ad-EP4-infected cultures in the presence or absence of PGE 2 . Bar, 100 m. C, percentages of resorbed area on dentine slices. Percent resorbed area relative to the total surface was determined using an image analysis system. The values were expressed as the mean Ϯ S.D. of triplicate cultures. *, Significantly different from the control culture infected with Ad-LacZ; p Ͻ 0.01. #, Significantly different from the control culture infected with Ad-EP4 (10 8 ifu); p Ͻ 0.05. FIG. 7. Importance of disappearance of EP2 and EP4 in mature osteoclasts. Osteoclast precursors express functionally active EP1, EP2, and EP4 PGE 2 receptors. PGE 2 synergistically stimulates RANKL-induced osteoclastic differentiation of the precursors through EP2-and EP4-mediated signals. The precursors down-regulate the expression of their own EP2 and EP4 during their differentiation into osteoclasts. Osteoclasts continuously express EP1, but the role of EP1 in osteoclast function is not known. When EP4 is expressed in osteoclasts, PGE 2 directly inhibits the pit-forming activity of the osteoclasts through a cAMP/PKA-dependent mechanism. Thus, the EP4-mediated signals in osteoclasts are similar to those mediated by CT receptors. Thus, PGE 2 becomes a potent bone resorption-stimulating factor as a result of the disappearance of EP2 and EP4 from mature osteoclasts. The down-regulation of EP2 and EP4 in osteoclast precursors during osteoclastic differentiation is an important event for escaping the inhibitory effect of PGE 2 on bone resorption. CTR, calcitonin receptor. osteoclasts (28,29,43). A pharmacological study using specific agonists for each EP subtype showed that isolated rat osteoclasts possess EP2 (44). Rabbit osteoclasts have been shown to express EP4 (29). We used mouse osteoclasts formed in vitro, but not authentic osteoclasts, in this study. Therefore, it is possible that authentic mouse osteoclasts may express EP2 and EP4 and decrease their function in response to PGE 2 . However, even if osteoclasts express EP2 and EP4, the expression level must be extremely low. PGE 2 does not inhibit but rather stimulates bone resorption in a mouse organ culture system using fetal or newborn calvariae and long bones (45)(46)(47)(48). Unlike PGE 2 , calcitonin completely inhibits not only osteotropic factor-induced bone resorption but also spontaneously occurring bone resorption in such organ culture systems. These results further support our conclusion that osteoclast precursors lose EP2 and EP4 to escape the inhibitory effects of PGE 2 on bone resorption (Fig. 7).
We and others have shown that PGE 2 synergistically enhanced RANKL-induced osteoclastic differentiation of BMM through EP2 and EP4 (25)(26)(27). Recently, we found that TAK1 (transforming growth factor-␤-activated kinase 1), a signal adapter molecule, plays important roles in cross-talk between cAMP-PKA signals and RANKL-induced signals in PGE 2 enhancement of RANKL-induced osteoclast differentiation (27). Our results suggest that osteoclast precursors express EP2 and EP4, through which PGE 2 synergistically enhances RANKLinduced osteoclast differentiation (Fig. 7). After the differentiation into osteoclasts is decided, osteoclasts would lose EP2 and EP4 to escape the inhibitory effects of PGE 2 . In addition, PGE 2 strongly induces RANKL expression in osteoblasts through EP2/EP4-mediated signals. Thus, PGE 2 is a potent bone resorption factor with multiple effects on osteoclast precursors and osteoblasts.
In contrast to the above in vitro studies, many in vivo studies have shown that the PGE series has stimulatory effects on bone formation (49 -57). Sibonga et al. (52) and Keila et al. (53) independently reported that administration of PGE 2 even into aged rats increased bone mass due to an increase in osteoblasts over osteoclasts. It was also demonstrated that PGE 2 and bipedal stance exercise synergistically prevented cancellous bone loss induced by ovariectomy in aged rats (54). Local administration of PGE 1 has been shown to increase alveolar bone thickness and bone formation in beagle dogs (55). Paralkar et al. (56) reported that a selective EP2 agonist mimics the anabolic effects of PGE 2 on bone when directly injected into the bone marrow of rats. By infusing PGE 2 to mice lacking each of four PGE receptor subtypes, Yoshida et al. (57) have identified EP4 as the receptor that mediates bone formation in response to this agent. These results suggest that the anabolic action as well as catabolic action of PGE 2 in bone is linked to an elevated level of cAMP. In normal bone remodeling, osteoblastic bone formation follows osteoclastic bone resorption and occurs in a precise and quantitative manner. PGE 2 -induced bone resorption may directly induce bone formation. The findings obtained from in vivo studies suggest that EP4 and EP2 are targets for therapeutic intervention in metabolic bone diseases. Further studies will be necessary to elucidate the molecular mechanism of PGE 2 -induced bone formation in vivo.