15-Keto-PGE2 acts as a biased/partial agonist to terminate PGE2-evoked signaling

Suzu Endo, Akiko Suganami, Keijo Fukushima, Kanaho Senoo, Yumi Araki, John W. Regan, Masato Mashimo , Yutaka Tamura*, and Hiromichi Fujino* From the Department of Pharmacology for Life Sciences, Graduate School of Pharmaceutical Sciences and Graduate School of Biomedical Sciences, Tokushima University, Tokushima, Japan, the Department of Bioinformatics, Graduate School of Medicine, Chiba University, Chiba, Japan, the Department of Pharmacology and Toxicology, College of Pharmacy, University of Arizona, Tucson, Arizona, USA, and the Laboratory of Pharmacology, Faculty of Pharmaceutical Sciences, Doshisha Women’s College of Liberal Arts, Kyotanabe, Kyoto, Japan

E-type prostanoid (EP) receptors are known as cognates for prostaglandin E 2 (PGE 2 ) that have four main subtypes: EP1, EP2, EP3, and EP4 (1). Prostanoid receptors have been reported to be activated not only by their cognate ligands but also noncognate prostanoids as biased ligands (2)(3)(4). For example, we previously found that PGE 1 and PGE 3 are able to induce cAMP formation effectively as full agonist like PGE 2 , but they only partially activate b-catenin/T-cell factor (TCF)-mediated signaling as partial agonists, approximately half-maximal effects/ responses compared with those induced by PGE 2 (4). Because b-catenin/TCF-mediated signaling is well-known colorectal cancer-related signaling (5), PGE 1 and PGE 3 act as negative biased agonists for EP4 receptors to mediate anti-cancer effects by selectively not fully activating b-catenin/TCF-mediated signaling. These biased activities are possibly due to the different receptor conformations caused by the numbers and patterns of hydrogen-bonding formation between EP4 receptors and each PGE (4). Therefore, depending on slight structural differences, such as the numbers of double bonds, each prostanoid may activate distinct signaling as biased agonist via the same receptors. PGE 2 is well-known as an endogenous proinflammatory prostanoid, which is synthesized from arachidonic acid by the activation of cyclooxygenase-2 (COX-2) (6). After PGE 2 has evoked sufficient inflammation, it is metabolized to 15-keto-PGE 2 by the action of 15-hydroxyprostaglandin dehydrogenase. 15-Keto-PGE 2 is widely considered as an inactive form of PGE 2 (7,8). On the other hand, 15-keto-PGE 2 has also been previously shown to activate and produce cAMP via Ga s -proteincoupled EP2 and EP4 receptors, although the potencies and/or efficacies are weaker than those of PGE 2 (9), because this prostanoid is not able to effectively activate EP receptors (10). The EP2 and EP4 receptors are, however, currently known to activate not only Ga s -protein-coupled cAMP-mediated signaling but also b-catenin/TCF-mediated signaling (6,11). Because 15keto-PGE 2 is a one-hydrogen-removed reduced form of PGE 2 , this prostanoid will plausibly show the activity of b-catenin/ TCF-mediated signaling besides Ga s -protein-coupled cAMPmediated signaling and could act as a biased ligand for EP2 and EP4 receptors to regulate their diverged signaling pathways.
The expression levels of EP4 receptors have been reported to be higher in normal than cancerous tissues (12,13), so EP4 receptors have been considered to play important roles in the maintenance of normal colorectal homeostasis (14). Normal colorectal epithelial cell proliferation and differentiation have also been reported to be regulated by b-catenin/TCF-mediated signaling. Indeed, this signaling is considered to have key roles in maintaining colorectal homeostasis (15). On the other hand, PGE 2 is also well-known to be involved in colorectal cancer development (1, 16). Thus, increases in the levels of PGE 2 and COX-2 are biomarkers for the early stage of colorectal cancer development. Because the up-regulation of COX-2 expression has been reported to be associated with the activation of EP4 ‡ These authors contributed equally to this work. * For correspondence: Hiromichi Fujino, hfujino@tokushima-u.ac.jp; Yutaka Tamura, yutaka_tamura@faculty.chiba-u.jp.
receptors, the signaling pathways and expression mechanisms of EP4 receptors have been extensively investigated (12,(17)(18)(19). As described, b-catenin/TCF-mediated signaling is a biomarker for colorectal cancer (5), and it is evoked via activation of EP4 receptors (11). Therefore, if the EP4 receptor-mediated b-catenin/TCF-regulated homeostasis systems collapse, it will lead to the early stage of colorectal cancer, possibly due to the aberrant expression of EP4 receptors followed by unrestrained induction of COX-2 and excess amounts of de novo synthesis of PGE 2 (19,20). Meanwhile, b-catenin/TCF-mediated signaling is not only activated by EP4 receptors but also EP2 receptors (11). In collaboration with knockout mouse studies, EP2 receptors have been considered to be involved in colorectal cancer development as well (21). However, the detailed roles of EP2 receptors in colorectal cancer development as well as in normal homeostasis are not clearly understood.
Because 15-keto-PGE 2 is able to bind to and has lower efficacies for EP2 and EP4 receptors than PGE 2 , at least in terms of cAMP formation (9), there is a possibility that 15-keto-PGE 2 may take over the PGE 2 -evoked responses after PGE 2 has been metabolized. Thus, it is plausible that 15-keto-PGE 2 may play the role of attenuating and/or terminating PGE 2 -evoked functions. In the present study, to further estimate the role of 15keto-PGE 2 in attenuating PGE 2 -evoked signaling pathways, we examined the effects of 15-keto-PGE 2 on each diverse signaling pathway as well as the binding affinity of EP2 or EP4 receptors. Because it is very difficult to monitor/measure the ratios of PGE 2 and 15-keto-PGE 2 as well as each ligand-induced signaling in regular cultured cell-based assay methods, operational model calculation, in silico analysis, and computer simulation were applied along with cell-based experiments using actual data, such as E max values and EC 50 values that were obtained by pharmacological experiments as well as IC 50 values of the binding assay as provided in this study. Here we show that PGE 2activated EP4 receptor-mediated signaling may evoke the primary initiating reaction of the cells, which would take over the 15-keto-PGE 2 -activated EP2 receptor-mediated signaling after PGE 2 is metabolized to 15-keto-PGE 2 . In other words, 15keto-PGE 2 may not be just an inactive metabolite of PGE 2 but could act as a "switched agonist" of EP2 receptors from PGE 2activated EP4 receptors, which would mildly restore/terminate the PGE 2 /EP4 receptor-induced signaling for inflammatory reactions, and/or for maintaining homeostasis for colorectal cell functions. Therefore, once the EP2 receptor-mediated signaling is relatively weaker and/or the EP4 receptor-mediated signaling is persistently activated, the 15-keto-PGE 2 -mediated restoration/termination of signaling may not be started. The continuous PGE 2 -mediated signaling would evoke aberrant EP4 receptor-dominated signaling that would disrupt homeostasis and eventually lead to cancerous transformation.
Next, to examine the binding affinity of 15-keto-PGE 2 for each EP2 or EP4 receptor, a competitive whole-cell radioligand binding of [ 3 H]PGE 2 assay was then performed. As shown in Fig. 2B, PGE 2 and 15-keto-PGE 2 caused the concentration-dependent inhibition of [ 3 H]PGE 2 binding to both HEK-EP2 and HEK-EP4 cells. The IC 50 values of EP2 receptors for PGE 2 and 15-keto-PGE 2 were ;2.94 nM (95% confidence interval: 1.23-7.00 nM) and 118 nM (95% confidence interval: 82.2-170 nM), respectively. The IC 50 values of EP4 receptors for both prostanoids were about 434 pM (95% confidence interval: 282-668 pM) and 2.82 mM (95% confidence interval: 1.66-4.80 mM), respectively. Thus, the differences in IC 50 values between PGE 2 and 15-keto-PGE 2 are around 100 times in HEK-EP2 cells but around 10,000 times in HEK-EP4 cells. Therefore, PGE 2 may tend to bind EP4 receptors rather than EP2 receptors, whereas 15-keto-PGE 2 may more easily bind to EP2 receptors than EP4 receptors. The lower potencies and efficacies of 15-keto-PGE 2 for EP4 receptors in terms of cAMP formation, phosphorylation of ERKs, and b-catenin/TCF-mediated transcriptional activity could be due, at least in part, to the lower binding affinity of this prostanoid for EP4 receptors. In addition, 15-keto-PGE 2 is also known to act through the activation of peroxisome proliferator-activated receptor (PPAR) (8); however, treatment with 15-keto-PGE 2 of parent HEK-293 cells did not evoke PPAR-response element-luciferase reporter gene activity at any concentrations (data not shown), so the effects of this prostanoid may be mediated through EP2 or EP4 receptors, but not via PPAR.
We previously showed that tyrosine 196 on transmembrane 5 (TM5) and glutamate 288 on TM6 in human EP2 receptors, as well as lysine 82 on TM2, arginine 291 on TM6, and serine 307 on TM7 in human EP4 receptors, are key amino acids for ligand binding (3,4). Thus, to further examine the interactions between PGE 2 or 15-keto-PGE 2 and either EP2 or EP4 receptors, in silico simulations were performed. As shown in Fig. 3A, 15-keto-PGE 2 was retained to form a nonclassical CH-p hydrogen bond at position 10 with the phenol ring structure of Tyr-196 in EP2 receptors, similar to PGE 2 . However, unlike PGE 2 , because position 15 of the hydroxyl functional group was oxidized to a carbonyl functional group, the hydrogen bond was not formed to 15-keto-PGE 2 with Glu-288 in EP2 receptors. As we discussed previously, PGE 2 forms a CH-p hydrogen bond with the EP2 receptor, resulting in a stable cognate receptor for PGE 2 because the CH-p bonds do not form with PGD 2 or PGF 2a (3). Thus, the higher potencies and efficacies of 15-keto-PGE 2 for EP2 receptors than EP4 receptors in terms of cAMP formation and b-catenin/TCF-mediated activity could be due to the higher affinity of this prostanoid for EP2 receptors through CH-p bond formation, at least to some extent.
In the case of EP4 receptors, as shown in Fig. 3B, 15-keto-PGE 2 was retained to form a hydrogen bond at position 1 with Lys-82 in EP4 receptors, similar to PGE 2 . As we reported previously, PGE 2 formed a single hydrogen bond at position 9 to Arg-291 and two hydrogen bonds at position 15 to Ser-307 in EP4 receptors (4). 15-Keto-PGE 2 formed two hydrogen bonds at position 9 to Arg-291, but no bond was formed with Ser-307, possibly because position 15 of the hydroxyl functional group was oxidized to a carbonyl functional group. These two hydrogen bonds formed at position 9 of the cyclopentane ring of 15keto-PGE 2 with Arg-291 of EP4 receptors were also formed by PGE 1 as well as PGE 3 , which was discussed previously and considered as one reason why they are not able to transform the receptor conformation fully to activate b-catenin/TCF-mediated transcriptional activity of EP4 receptors (4). Interestingly, PGE 1 and PGE 3 are able to form one hydrogen bond at position 15 to Ser-307 of EP4 receptors (4); however, no bond was formed at position 15 of 15-keto-PGE 2 with Ser-307 of EP4 receptors, as shown in Fig. 3B. It is considered that this hydrogen bond between position 15 of prostanoids and Ser-307 of EP4 receptors may be a key factor in biased activity for cAMP formation. Thus, hydrogen bond-forming PGE 1 and PGE 3 are able to act as full agonists in the pathway to cAMP formation, whereas 15- 15-keto-PGE 2 acts as switched agonist of EP receptors keto-PGE 2 may not be able to act as a full agonist but as a partial agonist for that pathway, because the key position 15 of the hydroxyl functional group was oxidized to a carbonyl functional group.
As described in the Introduction, 15-keto-PGE 2 has been considered as an inactive metabolite of PGE 2 (7,8,10). However, to summarize the points so far, 15-keto-PGE 2 acted as a full agonist for EP2 receptors in terms of the pathway to cAMP formation and as a partial agonist of b-catenin/TCF signaling. In terms of EP4 receptors, 15-keto-PGE 2 acted as a partial agonist of all the tested EP4 receptor-mediated signaling. There-fore, it is considered that 15-keto-PGE 2 is not an inactive form of PGE 2 but that this prostanoid itself may have the potential to evoke and take over the signaling even after PGE 2 has been degraded by the action of 15-hydroxyprostaglandin dehydrogenase to 15-keto-PGE 2 . Moreover, 15-keto-PGE 2 showed a predilection for the EP2 receptor and its signaling rather than for the EP4 receptor and its signaling. However, it is very difficult to monitor/measure the ratios of PGE 2 and 15-keto-PGE 2 as well as each ligand-induced signaling in regular cultured cell-based assay methods. Thus, to examine the roles of 15keto-PGE 2 after PGE 2 degradation, we applied the actual data, 15-keto-PGE 2 acts as switched agonist of EP receptors such as IC 50 values of the binding assay and E max values obtained by PGE 2 stimulations and EC 50 values that were obtained in Figs. 1 (B and C) and 2A, to the Black/Leff operational model (24)(25)(26)(27)(28)(29) and to in silico computer simulations.
From the experimental results obtained in Figs. 1 (B and C) and 2A, the best-fit curves of cAMP formation, phosphorylation of ERKs, and b-catenin/TCF-mediated activity evoked by either PGE 2 or 15-keto PGE 2 in HEK-EP2 or HEK-EP4 cells were regressed.
The regressed PGE 2 concentration-response curve was plotted (solid wine-red line), and then the reverse style of the 15keto-PGE 2 concentration-response curve was plotted (solid blue-gray line), as shown in Fig. 4A (a). When PGE 2 reached a maximal concentration of 10 25 M, then all the metabolized PGE 2 would eventually become 15-keto-PGE 2 . Thus, the concentration of PGE 2 would be considered to decrease in reverse increments ( Fig. 4A (a), dashed wine-red line). Along with the decrease in the concentration of PGE 2 , metabolized 15-keto-PGE 2 was increased (dashed blue-gray line). For example, when PGE 2 was 10 26 M, the corresponding metabolized 15keto-PGE 2 was considered to be ;10 25.05 M using the formula, [15-keto-PGE 2 ] = 10 25 2 [PGE 2 ]. When all the PGE 2 had been completely metabolized to 15-keto-PGE 2 , then the maximal concentration of 15-keto-PGE 2 was considered to reduce the concentration as shown in Fig. 4A (a) (solid blue-gray line).
When PGE 2 was metabolized to 15-keto-PGE 2 , however, those prostanoids could compete with each other at each receptor, because both prostanoids concomitantly exist in the same environment. Thus, the area in which PGE 2 and 15-keto-PGE 2 co-existed as shown in Fig. 4A (a) was named as the Schild area.
In the case of cAMP formation response curves in EP2 receptors, as shown in Fig. 1B, 15-keto-PGE 2 was able to activate EP2 receptors to a similar level of PGE 2 as full agonist. Therefore, apparent concentrations of PGE 2and 15-keto-PGE 2 -competing EP2 receptors were calculated using Schild regression analysis, as described under "Experimental procedures" and were replotted using the calculated results (Schild area). Schild regression analysis is known to quantify the affinity of competitive antagonists (i.e. if the concentration of the antagonist is known, the affinity of the antagonist can be calculated) (24). Here, we have utilized Schild regression analysis for estimating the apparent concentrations of the agonist by using the affinity value of the antagonist. Thus, using IC 50 values of PGE 2 and 15keto-PGE 2 for EP2 receptors obtained in Fig. 2B, either 15keto-PGE 2 or PGE 2 was regarded as a competing antagonist, and apparent concentrations of both prostanoids were calculated using Schild regression analysis (Fig. 4, A and C).
On the other hand, 15-keto-PGE 2 was not able to activate both EP2 and EP4 receptors to the levels of PGE 2 , a full agonist, in each signaling pathway except for cAMP formation response in EP2 receptors, as described above. Therefore, 15-keto-PGE 2 acted as partial agonist for phosphorylation of ERKs, b-catenin/TCF-mediated signaling in both EP2 and EP4 receptors, and cAMP formation response in EP4 receptors (Figs. 1 (B and C) and 2A).
Although the IC 50 (the index of binding affinity) values of 15keto-PGE 2 to EP2 and EP4 receptors are shown in Fig. 2B, these values may not be appropriate for directly estimating the apparent concentrations of partial agonist (i.e. 15-keto-PGE 2 ) by Schild regression analysis. Because the concentration-response curve of partial agonist is determined by the efficacy of that partial agonist and the sensitivity of the system of each signaling pathway or the surrounding environment where the receptors are expressing, it may not always reflect the experimental IC 50 value of that partial agonist (29). Therefore, to determine the apparent affinity of partial agonist, the Black/Leff operational model was utilized (24)(25)(26)(27)(28)(29). The Black/Leff operational model is used to accommodate the fitting of experimental results and the occurrence of ligand-stimulated response cooperatively, because experimental concentration-response curves do not always reflect the stimulus-response processes (24). In this particular case, the maximum possible effect/response (E max value) experimentally obtained by each concentration-response curve of each signaling pathway of PGE 2 (full agonist) and each EC 50 value as well as each E max value of 15-keto-PGE 2 experimentally obtained by each concentration-response curve of each signaling pathway (Figs. 1 (B and C) and 2A) can determine each signaling pathway's specific apparent affinity of 15-keto-PGE 2 by  , and it reached the maximal concentration when all PGE 2 had been completely metabolized to 15-keto-PGE 2 (dashed blue-gray line), followed by the reduction (solid blue-gray line). Because those prostanoids could compete with each other at each receptor during the period when PGE 2 was metabolized to 15-keto-PGE 2 (dashed lines), because both prostanoids concomitantly exist in the same environment, the apparent concentrations of each prostanoid competing with either EP2 or EP4 receptors were calculated by Schild regression analysis as described under "Experimental procedures" (Schild Area). b, apparent concentrations of PGE 2 (C E2 ) and 15-keto-PGE 2 (C 15 ) were estimated by Schild regression using actual results of Figs. 1 (B and C) and 2A with IC 50 values of PGE 2 and 15-keto-PGE 2 obtained in Fig. 2B, and each apparent affinity value of 15-keto-PGE 2 was determined by the Black/Leff operational model calculation (K 15 ). c, the effect/response of PGE 2 (E E2 ) for PGE 2 at the concentration of C E2 and effect/response of 15-keto-PGE 2 (E 15 ) for 15-keto-PGE 2 at the concentration of C 15 were estimated from each regressed best-fit curve obtained in Figs. 1 (B and C) and 2A. Total effect/response in the Schild area, which was combined E E2 and E 15 , was plotted to the corresponding concentration of C E2 and C 15 (pink-violet line). B, schema of the increase or decrease of PGE 2 and 15-keto-PGE 2 , where the area in which both prostanoids exist is named the Schild area, from ;10 26 to 10 211 M PGE 2 and corresponding 15-keto-PGE 2 . Shown are the simulated total amounts of cAMP formation (C and D), phosphorylated ERKs (E and F), and TCF transcriptional activities (G and H) of EP2 and EP4 receptors with PGE 2 followed by 15-keto-PGE 2 . Wine-red line, PGE 2 alone; blue-gray line, 15-keto-PGE 2 alone; pink-violet line, PGE 2 with 15-keto-PGE 2 . DTC represents each signaling TC -TC of cAMP signaling (set as the standard).

15-keto-PGE 2 acts as switched agonist of EP receptors
Black/Leff operational model calculation. Note that the apparent responses should be directly calculated by the Black/Leff operational model without using the Schild regression analysis. However, we utilized this operational model just for calculating the apparent affinities of the partial agonistic activity of 15keto-PGE 2 , because it would be difficult to calculate the apparent affinities and efficacies of full agonist (15-keto-PGE 2 , which fully agonistically stimulated cAMP signaling via EP2 receptors) directly by this model. Therefore, apparent concentrations of PGE 2 (C E2 ) and 15keto-PGE 2 (C 15 ) competing with either EP2 or EP4 receptors in the Schild area were estimated by Schild regression using IC 50 values as well as apparent affinities of either PGE 2 or 15-keto-PGE 2 calculated by the Black/Leff operational model (Fig. 4A (b)).
Next, using the regressed best-fit concentration-response curves obtained from Figs. 1 (B and C) and 2A, the apparent effect/response of PGE 2 (E E2 ) at the concentration of C E2 and apparent effect/response of 15-keto-PGE 2 (E 15 ) at the concentration of C 15 were estimated. Finally, the total effect/ response, which was combined E E2 and E 15 , was replotted back to the corresponding concentrations of C E2 1 C 15 , as shown in Fig. 4A (c).
As presented in Fig. 4B, the PGE 2 concentration will simply increase from 0 to 10 25 M, with the characters from a to h shown in wine-red color. When PGE 2 reached 10 25 M, the maximal concentration, then there was a decrease in increment. With a decrease in the concentration of PGE 2 , 15-keto-PGE 2 will increase, because all the metabolized PGE 2 is considered to become 15-keto-PGE 2 . Thus, again, when PGE 2 was metabolized to 10 26 M, the corresponding 15-keto-PGE 2 was estimated as ;10 25.05 M, which was calculated from the formula, [15keto-PGE 2 ] = 10 25 2 [PGE 2 ]. The Schild area, where both prostanoids exist, is indicated with the characters from i to n in pink-violet color (from ;10 26 to 10 211 M PGE 2 and corresponding 15-keto-PGE 2 ). Then, after PGE 2 has been completely metabolized to 15-keto-PGE 2 , it will simply be decreased, as indicated with the characters from o to v in blue-gray color (from about 10 26 to 0 M).
Based on the Schild regression analysis using IC 50 values and/or apparent affinities calculated by the Black/Leff operational model, we estimated the total effects/responses of each receptor-activated cAMP formation, phosphorylation of ERKs, and b-catenin/TCF-mediated transcriptional activities. In the Schild area, when apparent concentration of PGE 2 to the EP2 receptors was 10 26.11 M, the corresponding apparent concentration of 15-keto-PGE 2 was 10 27.58 M. The effects/responses evoked by each prostanoid using regressed best-fit curves of cAMP were 23.13 pmol for PGE 2 and 3.720 pmol for 15-keto-PGE 2 , so that the total amount of cAMP would be 26.92 pmol; although the E max value of cAMP formation caused by PGE 2 was 23.15 pmol, total amounts higher than 23.15 pmol were plotted as 23.15 pmol (Fig. 4C). Similar simulations were performed in terms of cAMP in EP4 receptors as well as phosphorylation of ERKs and b-catenin/TCF-mediated transcriptional activities in both EP2 and EP4 receptors (Fig. 4, D-H).
As shown in Fig. 4C, the simulated EP2 receptor-stimulated total cAMP formed by PGE 2 and 15-keto-PGE 2 (pink-violet line) reached the E max at 10 28 M PGE 2 (e) and continued to the highest levels in the Schild area at 10 26 M 15-keto-PGE 2 (p) and then gradually decreased to 10 29 M prostanoid (s). Although, as shown in Fig. 4D, the simulated EP4 receptor-stimulated total cAMP formed by PGE 2 and 15-keto-PGE 2 (pink-violet line) reached the E max at 10 29 M PGE 2 (d), the effect/response was less than half when compared with EP2 receptor-stimulated cAMP formation shown in Fig. 4C. The E max level continued in the Schild area at around 10 29 M PGE 2 with ;10 25 M 15-keto-PGE 2 (l) and was markedly reduced during the last half of the Schild area and then decreased gradually and in a stepwise manner to 10 28 M prostanoid (r), as shown in Fig. 4D (pink-violet line).
In terms of phosphorylation of ERKs, the simulated EP2 receptor-stimulated phosphorylation of ERKs was only weakly or not evoked by either PGE 2 or 15-keto-PGE 2 (Fig. 4E, pinkviolet line), whereas, as shown in Fig. 4F, the simulated EP4 receptor-stimulated phosphorylation of ERKs reached the E max level at 10 28 M PGE 2 (e) (pink-violet line). The E max level continued to the Schild area at around 10 27 M PGE 2 with ;10 25 M 15-keto-PGE 2 (j) (pink-violet line), gradually reduced during the last half of the Schild area, and then decreased in a stepwise manner to 10 29 M prostanoid (s), similar to the cAMP formation shown in Fig. 4D (Fig. 4F). With regard to b-catenin/TCFmediated activity, the simulated curves were more robustly reduced under EP4 receptor-stimulated conditions than EP2 receptor-stimulated conditions in the Schild area (pink-violet line). However, they were similarly activated by PGE 2 , reached almost identical E max levels at the same concentration, and then decreased gradually and in a stepwise manner to 10 29 M prostanoids (Fig. 4, G and H).
As described earlier, in terms of cAMP formation, although it was previously reported to have partial agonistic effects on EP2 and EP4 receptors (9), 15-keto-PGE 2 is widely considered as the inactive metabolite of PGE 2 (7,8,10). However, from the simulated results shown in Fig. 4, PGE 2 -evoked signaling may not be shut down immediately, but 15-keto-PGE 2 has the potential to take over the actions of PGE 2 to bring such signaling to an end. The phase-out decline of the signaling by 15keto-PGE 2 is probably attributable to the affinity, efficacy, and sensitivity of the system/surrounding environment of 15keto-PGE 2 for the receptors. Thus, regarding EP2 receptormediated signaling, the maximal effects/responses evoked by 15-keto-PGE 2 were similar or showed little decline when compared with the effects/responses evoked by PGE 2 . Therefore, the PGE 2 -evoked signaling may be prolonged with higher effect/response, and it may be terminated smoothly and gradually because 15-keto-PGE 2 could act as a full and/or potent partial agonist of EP2 receptors with higher binding affinity than that of EP4 receptors. On the other hand, regarding EP4 receptor-mediated signaling, the maximal effects/responses evoked by 15-keto-PGE 2 for EP4 receptors were lower than those of PGE 2 . Thus, the EP4 receptors were not fully activated by 15keto-PGE 2 so that each signaling was only marginally extended; the effects/responses from EP4 receptors evoked by 15-keto-PGE 2 were declined ;50% in cAMP formation, 40% in the case of activation of ERKs, and 75% in b-catenin/TCF-mediated activation when compared with the effects/responses evoked by a full agonist, PGE 2 (Figs. 4, D, F, and H). Nevertheless, 15keto-PGE 2 may not terminate the EP4 receptor-mediated PGE 2 -evoked signaling abruptly but does so in a stepwise manner.
Meanwhile, each apparent affinity of 15-keto-PGE 2 (K 15 ) calculated by the Black/Leff operational model was similar in EP2 receptors but different in EP4 receptors from the each signaling pathway as well as the IC 50 value obtained by binding assay as shown in Fig. 2B.
Therefore, it is possible to speculate that the partial agonistic activity of each signaling pathway evoked by 15-keto-PGE 2 on EP4 receptors could be differently regulated; the differences among the apparent affinities of the signaling pathways and IC 50 value of 15-keto-PGE 2 on EP4 receptors may reflect the different sensitivity of the system/surrounding environment among the signaling pathway-specific transitional states of EP4 receptors. Unlike the solely Ga s -protein-coupled EP2 receptors, EP4 receptors are shown to couple with additional Ga i -protein along with Ga s -protein (22). Thus, stepwise activation of each signaling pathway from EP4 receptors by 15-keto-PGE 2 , which may alter Ga s -and Ga i -protein coupling balances, depends on the signaling pathways, similarly to what we have discussed previously (3).
Incidentally, in terms of the maximal effects/responses of the signaling pathways of EP2 receptors, 15-keto-PGE 2 can fully activate the cAMP-mediated pathway (about 100% of PGE 2 ) but may act as partial agonist for the b-catenin/TCF-mediated signaling (;83% of PGE 2 ) and ERK pathways (around 70% of PGE 2 ). Thus, with respect to the E max values, 15-keto-PGE 2 showed biased activity for the cAMP-mediated pathway compared with other signaling pathways of EP2 receptors. In the case of EP4 receptors, 15-keto-PGE 2 could not fully activate all the EP4 receptor-mediated signaling pathways to the levels that were stimulated by PGE 2 , it acted as a partial agonist: about 53% in cAMP, around 68% in ERK activation, and ;49% in the b-catenin/TCF-mediated pathway.
Meanwhile, the transduction coefficient (TC) is known as a system/surrounding environment-independent parameter considering affinity and efficacy of the agonist (27)(28)(29). To use TC of the cAMP signaling as the standard, DTC can be calculated by simple subtraction; TC (cAMP) is subtracted from each signaling TC, which can be used as a signaling bias parameter. When compared with each DTC of PGE 2 , as shown in Fig.  4 (C-H), PGE 2 may prefer to activate the b-catenin/TCFmediated pathway (EP2, 0.66; EP4, 0.32) rather than cAMPmediated signaling (EP2, 0; EP4, 0) and the ERK-mediated pathway (EP2, 21.35; EP4, 20.81), in both EP2 receptors and EP4 receptors, whereas in the case of 15-keto-PGE 2 , it may prefer to activate the b-catenin/TCF-mediated pathway (EP2, 0.60; EP4, 1.11) rather than the ERK-mediated pathway (EP2, 0.14; EP4, 0.37) rather than cAMP-mediated signaling (EP2, 0; EP4, 0) in both EP2 receptors and EP4 receptors. Therefore, signaling biases of 15-keto-PGE 2 have changed from PGE 2 , and according to the values of DTC of 15-keto-PGE 2 , Ga s -proteinmediated cAMP signaling showed the smallest biased activity in both EP2 and EP4 receptors. Again, it could be a reason why 15-keto-PGE 2 has been regarded as an inactive metabolite of PGE 2 , which is also suggested by the Black/Leff operational model calculated DTC values.
Taken together, 15-keto-PGE 2 is able to act as a meaningful ligand to extend/sustain and/or terminate each signaling pathway evoked by PGE 2 , probably as a biased and/or partial agonist/ligand, in some cases as a full agonist/ligand, to pleiotropically fine-tune each signaling pathway.
In terms of effects/responses, from the results so far, EP2 receptors have biased activity for the cAMP-mediated pathway, whereas EP4 receptors have biased activity for activation of ERKs. Therefore, if the cells dominantly express EP2 receptors, PGE 2 /15-keto-PGE 2 would predominantly and continuously activate cAMP-mediated signaling, whereas if the cells mainly express EP4 receptors, ERK-mediated signaling would be predominant. On the other hand, in terms of apparent affinities, 15-keto-PGE 2 may preferentially affect (and hence inhibit) the ERK-mediated pathway (103 nM; Fig. 4E) via EP2 receptors and/or the b-catenin/TCF-mediated signaling (32.9 nM; Fig.  4H) via EP4 receptors as described above. Moreover, in terms of TC, 15-keto-PGE 2 may preferentially activate the b-catenin/ TCF-mediated signaling rather than the cAMP-mediated signaling through both receptor subtypes. In any case, depending on the dominantly expressing receptor subtype (i.e. EP2 or EP4), the predominantly activated signaling would be different. Thus, to clarify the extent to which the PGE 2 /15-keto-PGE 2evoked signaling differed, simple simulations were performed by altering the ratios of EP2 and EP4 receptors, such as EP2/ EP4 ratios of 4:0, 3:1, 2:2, 1:3, and 0:4, as depicted in Fig. 5A.
The formed maximal amounts and sustained ability to produce cAMP were highest if the cells expressed only the EP2 15-keto-PGE 2 acts as switched agonist of EP receptors receptor subtype, and those were decreased in a stepwise manner by increasing the expression of EP4 receptors (Figs. 4C and 5B), whereas if the expressed receptor subtype was only EP4, the maximal amounts formed were markedly decreased but were still a little less than half when compared with the EP2 receptor only. However, the ability to produce cAMP was markedly reduced at the point when PGE 2 was metabolized to 15keto-PGE 2 in the Schild area (Fig. 5B). On the other hand, in the case of activation of ERKs, if the cells expressed only the EP4 receptor subtype, PGE 2 -stimulated EP4 receptors evoked strong activation of ERKs, but this was reduced to some extent when PGE 2 was replaced with 15-keto-PGE 2 in the Schild area (Figs. 4F and 5C). Interestingly, although the maximal activation of ERKs was decreased in a stepwise manner, the decreased gap in activation caused by switching to 15-keto-PGE 2 became inconspicuous by increasing the expression of EP2 receptors (Fig. 5C), whereas if the expressed receptor was only the EP2 subtype, the activation of ERKs by PGE 2 as well as 15-keto-PGE 2 was very weak and may be negligible. It should be noted that the cAMP assay was performed under the effects of a phos-phodiesterase inhibitor, 3-isobutyl-1-methylxanthine, so that in the physiological conditions, the EP4 receptor-stimulated cAMP may be degraded much faster than the EP2 receptorstimulated cAMP, which may make it negligible like activation of ERKs evoked by EP2 receptors.
As shown in Fig. 5 (B and C), the cAMP-mediated signaling and ERK-mediated signaling could markedly change if the expression ratio of EP2 and EP4 receptors is altered. However, interestingly, the induction of b-catenin/TCF-mediated transcriptional activity was not changed by altering the ratio of receptor subtypes, although the ability to sustain the activation was gradually reduced by an increasing ratio of the EP4 receptor subtype after the point when PGE 2 was metabolized to 15keto-PGE 2 in the Schild area (Fig. 5D), plausibly because of potent apparent affinity (32.9 nM) and weak E max (about 49% of PGE 2 ) as shown in Figs. 4H and 2A. Thus, b-catenin/TCFmediated signaling would be stimulated to a similar extent regardless of the EP2 or EP4 receptor subtypes. Of particular interest, it has been reported that b-catenin/TCF-mediated signaling is involved in colorectal epithelial cell proliferation as well as differentiation for maintaining intestinal homeostasis (15). We previously showed that EP2 receptor-mediated b-catenin/TCF-mediated signaling is mainly involved in the cAMP/ PKA pathway, whereas EP4 receptor-mediated signaling is primarily involved in the PI3K/ERKs pathway (11). Thus, despite utilizing distinct pathways via EP2 receptors and/or EP4 receptors, PGE 2 may maintain b-catenin/TCF-mediated signaling to preserve homeostasis. This may be the reason why 15keto-PGE 2 preferentially activates b-catenin/TCF-mediated signaling in both EP2 receptors and EP4 receptors in terms of potencies for retaining homeostasis. Thus, regardless of the expression ratio of EP2 or EP4 receptors, PGE 2 -stimulated b-catenin/TCF-mediated signaling may not be so different among persons if the total amounts of EP2 and EP4 receptor subtypes are similar, so that colorectal tissue homeostasis may not be affected by the ratio of the expression levels of both EP receptor subtypes. Note that the simulations we have performed here were based on the situation in which sufficient PGE 2 was provided to each receptor subtypes around the environment. However, according to the results obtained in Fig. 4, if there is not much PGE 2 available, both receptor subtypes may be competing for the ligand so that simulation would be complicated because it must consider the apparent affinities and the effects/responses for each signaling pathway, which will need to be taken into account in the future. cAMP/PKAmediated signaling is widely accepted as regulating the inhibition of cellular growth (30), whereas PI3K/ERK-mediated signaling is often associated with cancer malignancy (20,31). Thus, to estimate the relationship between the ratio difference of EP2 and EP4 receptors and the possibility of cancer development, in silico analysis using the colon and rectal cancer data of the Cancer Genome Atlas (TCGA) database was performed. Fig. 6A shows the mRNA expression of 383 colorectal cancer samples, which were extracted from the COADREAD data set, and the mRNA expression ratio of EP4 and EP2 was calculated and plotted. The median ratio was 5.89. The EP4/EP2 ratio in the high group was the same as or higher than the median (192 samples); when the ratio was lower than the median, the sample 15-keto-PGE 2 acts as switched agonist of EP receptors was plotted as ratio-low (191 samples). Using Kaplan-Meier analysis, the probabilities of survival were calculated and plotted between EP4/EP2 ratio-high and ratio-low groups, shown in Fig. 6B. Thus, the survival probability of the EP4/EP2 ratiohigh group was significantly lower than that of the ratio-low group, indicating that relatively high EP4 receptor-expressing colorectal cancer may be related to a poor survival rate. Next, the expression levels of EP2 and EP4 receptors were shown side by side between high and low ratio groups. As shown in Fig. 6C, the expression levels of EP4 receptors were similar between high and low ratio groups (Fig. 6C). However, the EP2 receptor expression levels were significantly higher in the ratio-low group than in the ratio-high group (Fig. 6D). Thus, the poor survival probability of the EP4/EP2 ratio-high group may be due to the lower expression level of EP2 receptors, which is possibly caused by the relatively lower activation of the cAMP- Figure 6. The relationship between the ratio difference of EP2 and EP4 receptors and the possibility of cancer development using the TCGA database and schema of homeostasis or cancer development mechanisms regulated by expression levels of EP2 and EP4 receptors. A, the mRNA expression of 383 colorectal cancer samples, which were extracted from the COADREAD data set, and the mRNA expression ratio of EP4 and EP2 were calculated and plotted. B, with Kaplan-Meier analysis, the probabilities of survival were calculated and plotted between EP4/EP2 ratio-high and ratio-low groups. The expression levels of EP4 receptors (C) and EP2 receptors (D) are shown side by side between high-and low-ratio groups. E, homeostatic mechanism regulated by the expression levels of EP2 and EP4 receptors (left scheme). Relatively higher EP4 receptor-dominated signaling would turn to cancer malignancy signaling, probably by reducing the expression of EP2 receptors (right scheme).
15-keto-PGE 2 acts as switched agonist of EP receptors mediated signaling as well as the comparatively higher ERKmediated signaling.
When taken together, based on the results obtained in this study, the homeostatic mechanism may be tightly regulated by the expression levels of EP2 and EP4 receptors, as depicted in Fig. 6E (left scheme). As described in the Introduction, EP4 receptors are involved in the pathophysiology of colon cancer based on previous studies (12,(17)(18)(19). Thus, as depicted in Fig.  6E (right scheme), it is plausible that the relatively higher EP4 receptor-dominant signaling would derail this homeostatic mechanism and turn to cancer malignancy signaling by reducing the expression of the monitoring/guarding receptor subtype, the EP2 receptors.

Conclusions
It has long been considered that 15-keto-PGE 2 is the inactive metabolite of PGE 2 . However, this prostanoid may have an additional role as a biased/partial agonist to take over the actions of PGE 2 to gradually terminate reactions as soft-landing ways. Moreover, because of the marked differences in the binding affinity for the EP2 and EP4 receptors, 15-keto-PGE 2 also acts as a switch for cellular signaling to the EP2 receptor-mediated pathway from the EP4 receptor-mediated pathway, if both receptors are expressed closely on the same tissues and/or cells. In other words, PGE 2 -initiated EP4 receptor-mediated signaling would be terminated by the subsequent 15-keto-PGE 2 -adopted EP2 receptor-mediated restoring-signaling, which may have a role in maintaining homeostasis. Thus, once EP2 receptor-mediated signaling is relatively weaker and/or EP4 receptor-mediated signaling is persistently activated, the restoring signaling may not be started, so continuous PGE 2 -initiated signaling would evoke aberrant EP4 receptor-dominant signaling that would eventually lead to cancerous signaling. Note that those simulated 15-keto-PGE 2 -prolonged activations of each signaling pathway are likely to be altered in a physiological system/surrounding environment. Because it would be complicated and difficult to simulate, we did not consider the influences of 13,14-dyhydro-15-keto-PGE 2 , which is the further metabolized prostanoid of 15-keto-PGE 2 . The effects of 13,14-dyhydro-15-keto-PGE 2 should be examined in the near future; however, it is plausible that this prostanoid-mediated signaling may further take over 15-keto-PGE 2 signaling, which may terminate the PGE 2 -initiated signaling more smoothly, stepwisely, and pleiotropically. We are aware that these speculations are based on simulated calculations. As described earlier, it is very difficult to monitor metabolite conversion and their activities in actual experiments, which needs to be addressed in the future, but we believe that the present results shed new light on aspects of 15-keto-PGE 2 as not an inactive metabolite, but a biased and/or partial agonist to pleiotropically fine-tune each signaling pathway. Although there are some differences among the apparent affinities and the effects/ responses among the signaling pathways and between the receptor subtypes, according to the actual experimental IC 50 values obtained in Fig. 2B, 15-keto-PGE 2 may have important roles in translational activities from EP4 to EP2 receptors as a "switched agonist" for restoring/terminating the inflammatory reaction and/or maintaining homeostasis, such as in the colorectal tissues/cells functions.
cAMP assay HEK-EP2 or HEK-EP4 cells were cultured in 6-well plates and were switched from DMEM to Opti-MEM (Invitrogen, Carlsbad, CA, USA) containing 250 mg/ml Geneticin, 200 mg/ ml hygromycin B, and 100 mg/ml gentamicin 16 h prior to the experiments. Cells were treated with 0.1 mg/ml 3-isobutyl-1methylxanthine (Sigma) for 25 min followed by vehicle (0.1% DMSO) or the indicated concentrations of PGE 2 or 15-keto-PGE 2 for 60 min. Experiments were terminated by the removal of medium. The amount of cAMP that formed was calculated from a prepared standard curve using nonradiolabeled cAMP, as reported previously (3,4).

Western blotting
HEK-EP2 or HEK-EP4 cells were cultured in 6-well plates, and, prior to immunoblotting experiments, the culture medium was switched to Opti-MEM containing antibiotics at 37°C for 16 h, as stated above. Cells were then treated with either vehicle or the indicated concentrations of PGE 2 , 15-keto-PGE 2 , or 10 nM PMA for 5 min. Cells were scraped into lysis buffer consisting of 50 mM Tris-HCl (pH 8.0), 5 mM ethylene diamine (pH 8.0), 150 mM NaCl, 0.5% sodium deoxycholate, 1% Nonidet P-40, 10 mM sodium fluoride, 10 mM disodium pyrophosphate, 0.1% SDS, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, 10 mg/ml leupeptin, and 10 mg/ml aprotinin, as described previously (4). As reported previously (4), ;50 mg of protein samples was electrophoresed on 10% SDS-polyacrylamide gels and transferred to nitrocellulose membranes. After blocking for 1 h with 5% nonfat milk, the membranes were incubated with 1:1,000 dilutions of either an anti-phospho-ERK1/2 antibody (catalog no. 43705, Cell Signaling Technology, Danvers, MA, USA) in 5% BSA (Sigma) or a mixture of a 1:500 dilution of an anti-ERK1 antibody (sc-93, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) and a 1:20,000 dilution of an anti-ERK2 antibody (sc-154, Santa Cruz Biotechnology) in 5% nonfat milk at 4°C for 16 h. Membranes were washed twice after incubating with the primary antibody and then incubated with a 1:4,000 dilution of the appropriate secondary antibodies conjugated with horseradish peroxidase following the 15-keto-PGE 2 acts as switched agonist of EP receptors washouts, as described previously (4). To ensure the equal loading of proteins, membranes were stripped and reprobed with the anti-ERK1 and anti-ERK2 antibodies under the conditions described above.
TCF luciferase reporter gene assay HEK-EP2 or HEK-EP4 cells were cultured in 6-well plates and were switched from DMEM to Opti-MEM (Invitrogen) containing 250 mg/ml Geneticin, 200 mg/ml hygromycin B, and 100 mg/ml gentamicin prior to the experiments. Cells were transiently transfected with either TOP flash or FOP flash, as reported previously (3,4), along with the Renilla luciferase control plasmid, pRL-CMV (Promega, Madison, WI, USA), using HyliMax transfection reagent (Dojindo, Kumamoto, Japan). After ;4 h, transfection reagents were removed by a medium change, and cells were treated with the indicated concentrations of PGE 2 or 15-keto-PGE 2 for a further 16 h. Cells were then lysed and assayed using the Dual-Luciferase reporter assay system (Promega) according to the manufacturer's instructions with TECAN infinite M200 (TECAN, Mannedorf, Switzerland). Data were normalized by calculating the ratios of firefly luciferase scores to the corresponding Renilla luciferase values and corrected for background activity by the subtraction of FOP flash values from the corresponding TOP flash values, as described previously (3,4).
Prior to experiments, cell medium was switched to Opti-MEM containing antibiotics, as stated above, at 37°C for 16 h. Cells were trypsinized and resuspended at 10 5 cells/sample in 100 ml of ice-cold 10 mM MES (pH 6.0; Sigma) buffer containing 0.4 mM EDTA and 10 mM MnCl 2 (Sigma). As described previously (4), 2.5 nM [ 3 H]PGE 2 (GE Healthcare) was used for the binding assay with increased concentrations of PGE 2 or 15-keto-PGE 2 . Samples were incubated at 4°C for 2 h, and assays were terminated by filtration through a Whatman GF/C glass filter (Whatman, Maidstone, UK) followed by 3-5 washes with ice-cold MES buffer. Radioactivity was measured by liquid scintillation counting, as performed previously (4).

In silico analysis
The construction of the three-dimensional structure of human EP2 or human EP4 receptor and a docking simulation of PGE 2 or 15-keto-PGE 2 to either EP2 or EP4 receptor were performed with MOE (version 2016.08, CCG Inc., Montreal, Canada) based on the Protein Data Bank entry 4GRV.

In silico simulations
The simulated curves of cAMP formation, phosphorylation of ERKs, TCF-mediated transcriptional activity, and receptor occupancy were simulated using the results obtained in Figs. 1 and 2. The simulated cAMP formation curves were obtained using the average EC 50 value and E max shown in Fig. 1B

Black/Leff operational model calculation
The estimated affinity (K 15 ) and t value of 15-keto-PGE 2 were determined by GraphPad Prism software (version 8.0.1, La Jolla, CA, USA). The equation "operational model2partial agonist" was performed using the results obtained Fig. 1 (B-D) by using the formulas below. The basal level of the analysis for cAMP-mediated signaling was 0, for b-catenin/TCF-mediated signaling it was 100, and for phosphorylation of ERKs it was each bottom value. All Hill slopes used were specified as 1.
Each signaling E max value of PGE 2 obtained by each experiment as shown in Figs. 1 (B and C) and 2A was used as Effect max , an assumed maximal effect/response of each receptor (EP2 or EP4 receptor) activated by full agonist (PGE 2 ).
Operate ¼ 10 logK 15 110 ½B 10 logt 1 ½B (Eq. 1) where [B] represents 15-keto-PGE 2 concentration and Y is each effect/response. For partial agonist, TC (log(t/KA)) was obtained from K 15 and t calculated by the Black/Leff operational model for partial agonist. For full agonist, TC was directly calculated according to the formula below. The basal level of the analysis for cAMPmediated signaling was 0, for b-catenin/TCF-mediated signaling it was 100, and for phosphorylation of ERKs it was each bottom value. All Hill slopes used were specified as 1. Each signaling E max value of PGE 2 obtained by experiments was used as Effect max .

Y ¼ Basal1
Effectmax The apparent concentrations of PGE 2 and 15-keto-PGE 2 were estimated by Schild regression analysis. The apparent values of cAMP formed, phosphorylated ERKs, and activated b-catenin/TCF-mediated activity in the Schild area were obtained by the conversion from each apparent concentration of PGE 2 and 15-keto-PGE 2 with the simulated curves of each signaling pathway as described under "In silico simulations." The plotted total amounts of each signaling pathway in the Schild area were estimated using the sum of PGE 2 and 15-keto-PGE 2 , but if the total values were higher than E max values caused by PGE 2 , they were plotted as each E max value, respectively. The amounts of cAMP formed, phosphorylated ERKs, and activated TCF-mediated transcription evoked by PGE 2 at 0 M (a) to 10 25 M (h) and by 15keto-PGE 2 at 10 25 M (o) to 0 M (v) were plotted as curves, shown in Fig. 4B.

Bioinformatics analysis
The University of California Santa Cruz Xena browser (RRID:SCR_018938) was used to obtain the colon and rectal cancer data of TCGA (RRID:SCR_003193). The EP2 and EP4 gene expression values (log 2 of normalized count) of 383 colorectal cancer samples with no missing overall survival data were extracted from the COADREAD data set, which was converted to non-logarithmic value, and the expression ratio of the EP4 gene relative to the EP2 gene was calculated. The samples were divided into two groups, the EP4/EP2 ratio-high group (192 samples) and the EP4/EP2 ratio-low group (191 samples), by the median value of the EP4/EP2 expression ratio (5.89). The probabilities of survival of the groups were calculated using the Kaplan-Meier analysis method, and significance was analyzed using the log-rank test. Statistical analyses were performed with R (version 3.4.1, RRID:SCR_001905). Significance was assumed at p , 0.05.

Data availability
For bioinformatic analysis, the University of California Santa Cruz Xena browser (RRID:SCR_018938) was used to obtain the colon and rectal cancer data of TCGA (RRID:SCR_003193). The rest of the data are contained within the article.