Role of E-type prostaglandin receptor EP3 in the vasoconstrictor activity evoked by prostacyclin in thromboxane-prostanoid receptor deficient mice

Prostacyclin, also termed as prostaglandin I2 (PGI2), evokes contraction in vessels with limited expression of the prostacyclin receptor. Although the thromboxane-prostanoid receptor (TP) is proposed to mediate such a response of PGI2, other unknown receptor(s) might also be involved. TP knockout (TP−/−) mice were thus designed and used to test the hypothesis. Vessels, which normally show contraction to PGI2, were isolated for functional and biochemical analyses. Here, we showed that the contractile response evoked by PGI2 was indeed only partially abolished in the abdominal aorta of TP−/− mice. Interestingly, further antagonizing the E-type prostaglandin receptor EP3 removed the remaining contractile activity, resulting in relaxation evoked by PGI2 in such vessels of TP−/− mice. These results suggest that EP3 along with TP contributes to vasoconstrictor responses evoked by PGI2, and hence imply a novel mechanism for endothelial cyclooxygenase metabolites (which consist mainly of PGI2) in regulating vascular functions.

Scientific RepoRts | 7:42167 | DOI: 10.1038/srep42167 To resolve the above issues, in this study we generated a strain of TP −/− mice on a C57BL/6 background. Aortas, carotid and/or renal arteries, where PGI 2 evokes vasoconstrictor response under normal conditions 26,28,30,35 , were isolated for biochemical and/or functional analyses.

Results
Mutation in TP −/− mice and phenotype. As shown in Fig. 1a, sequencing of TP DNA PCR products revealed that as compared with that of wild-type (WT) mice, exon 3 of the TP locus in TP −/− mice has a 22 bp fragment deletion (CTG GGG GCC TGC TTT CGC CCG G) in the coding area, which was 18 bp after the start codon (NCBI Reference Sequence: NM_009325. 3). This resulted in a frame-shift in TP mRNA transcript and a premature termination of protein translation (only 7 amino acids were coded before the appearance of a stop codon (TGA) in TP −/− mice; Fig. 1a, bottom right). Indeed, RT-PCR revealed that un-mutated TP mRNAs, which were abundant in WT aortas, were not detected in the TP −/− counterparts (Fig. 1b). Also, compared to WT controls, TP −/− mice had an elongated bleeding time (Fig. 1c). However, these mice appear normal, and show no overt abnormality in mean arterial blood pressure (MAP; 92.3 ± 3.3 vs. 95.0 ± 2.8 mmHg in WT mice, n = 5 for each; P > 0.05) or in reproduction. Effect of TP −/− on contractions evoked by PGI 2 and other prostanoids. Abdominal aortas were then isolated for functional analyses. Vessels were treated with the NO synthase (NOS) inhibitor N ω -nitro-L-arginine methyl ester (L-NAME; 1 mM). In WT vessels, the TP agonist U46619 evoked potent contraction as noted previously 28 ; however, in TP −/− mice, U46619 did not evoke any response (Fig. 2a). Interestingly, not only the contraction evoked by PGI 2 (Fig. 2b), but that to PGF 2α (Fig. 2b), PGE 2 (Fig. 2c), or PGD 2 (Fig. 2d) was also diminished or largely removed in TP −/− vessels. At the same time, the contraction to low concentrations (0.1-0.3 μM) of PGE 2 remained intact in TP −/− vessels, and this contraction was abolished by the E-type prostaglandin receptor EP3 antagonist L798106 (1 μM), but not by the TP antagonist SQ29548 (10 μM; Fig. 2c). In addition, L798106 abolished the remaining contraction evoked by PGD 2 in TP −/− vessels (Fig. 2d). PGI 2 -induced response in TP −/− abdominal aortas precontracted with PE. Next, we determined whether PGI 2 -evoked contractile response in TP −/− vessels was masked by a dilator effect of the agonist. To this end, L-NAME-treated or endothelium-denuded abdominal aorta were precontracted with phenylephrine (PE; 2 μM), under which the vasoconstrictor response to an agonist is more readily detectable compared to baseline conditions 28 . Under either condition, PGI 2 (1 μM) evoked an increase of force on PE-induced contraction, which was however reversed by the EP3 antagonist L798106 (1 μM) into relaxation (Fig. 3a,b). In contrast, the EP1 antagonist SC19220 (10 μM) had no effect (Fig. 3a). Also, L798106, but not SC19220 inhibited a similar response evoked by PGE 2 (0.1 μM), although no relaxation was observed (Fig. 3c,d). Meanwhile, forces of PE-evoked contractions were found to be comparable among vessel groups that had been treated with L-NAME (Fig. 3a,c bottom panels). (a,b) comparison of contractions evoked by the TP agonist U46619 (a), and PGI 2 or PGF 2α (b) in WT and TP −/− vessels. (c,d) contraction to PGE 2 (c) or PGD 2 (d) in WT or TP −/− mice and that of TP −/− vessels treated with the TP antagonist SQ29548 (10 μM; +SQ) or the EP3 antagonist L798106 (1 μM; +L). Values are expressed as mean ± SEM; n = 5 for each. **P < 0.01 vs. the value of WT mice; ++ P < 0.01 vs. TP −/− mice.
As compared to that of WT controls, the contraction evoked by ACh (10 μM) in TP −/− vessels was indeed mostly abolished; however, a minor contraction could still be produced (Fig. 4a). Moreover, in such-treated TP −/− vessels precontracted with PE (2 μM), ACh evoked relaxation, which was blunted by a biphasic force sensitive to the non-selective COX inhibitor indomethacin. Interestingly, the EP3 antagonist L798106 also abolished the biphasic force, resulting in relaxation, which was to a greater extent than that obtained with indomethacin (Fig. 4b,c). In addition, such an enhancement of relaxation resulting from L798106 was removed by the IP antagonist CAY10441 (1 μM; Fig. 4c).
Likewise, in similar PE-precontracted WT vessels where the agonist usually evokes an increase of force 28 , treatment with the TP antagonist SQ29548 (10 μM) caused relaxation that was also blunted by a biphasic force in response to ACh. Again, the EP3 antagonist L798106 (1 μM) abolished the force, resulting in an enhanced relaxation that was reduced by indomethacin (Fig. 4d top). No significant difference was found among forces of PE-evoked contractions in TP −/− or WT vessels (Fig. 4c,d bottom panels). Effect of TP −/− on endothelial COX products and expressions of PG receptors. The production of PGI 2 and some other prostanoids evoked by ACh in TP −/− and WT aortas was then examined. As shown in Fig. 5a, in WT and TP −/− aortas ACh evoked an increase in the production of the PGI 2 metabolite 6-keto-PGF 1α , which was comparable between the two mouse strains. Also, an increase of PGE 2 was obtained with ACh, although levels were ~10-fold lower than those of 6-keto-PGF 1α (Fig. 5b). No significant difference was found in amounts of PGE 2 between TP −/− and WT vessels (Fig. 5b). In contrast, the TxA 2 metabolite TxB 2 was not increased by ACh in vessels from either mouse strain (Fig. 5c), similar to results reported previously 34,36 .
The expressions of IP, EP3 and the PGF 2α receptor (FP) mRNAs were also determined. As shown by real-time PCR, the level of IP, EP3 or FP mRNAs normalized by that of β-actin in TP −/− aortas was comparable with that of WT counterparts (Fig. 5d).

Effect of EP3 antagonism on varied vasoconstrictor responses in WT vessels. The effect of EP3
antagonism was further determined in L-NAME-treated WT vessels. As shown in Fig. 6a, in WT abdominal aortas the EP3 antagonist L798106 (1 μM) diminished the contraction evoked by PGI 2 . Also, in such vessels precontracted with PE (2 μM), PGI 2 (1 μM) or the COX substrate AA (3 μM), whose response is sensitive to TP antagonism under baseline conditions 35 , evoked an increase of force in the presence of the TP antagonist Scientific RepoRts | 7:42167 | DOI: 10.1038/srep42167 SQ29548 (10 μM) but a relaxation that was sensitive to the IP antagonist CAY10441 (1 μM) when both SQ29548 and L798106 were present (Fig. 6b). No significant difference was found among forces of PE-evoked contractions (Fig. 6b bottom).
Interestingly, in WT abdominal aortas, L798106, which drastically diminished the contraction evoked by ACh, only slightly reduced that evoked by a sub-maximal concentration of PGE 2 (10 μM; Fig. 6c). However, this contraction to PGE 2 was very sensitive to the TP antagonist SQ29548 (Fig. 6c). In addition, L798106 diminished the contraction evoked by 1 or 10 μM PGI 2 in carotid and renal arteries (Fig. 6d).

Discussion
In this study we show that in NOS-inhibited WT mouse abdominal aortas PGI 2 -or ACh (which activates endothelial COX to mainly produce PGI 2 ) evokes contraction that is diminished in TP −/− counterparts. More importantly, in TP −/− vessels or TP-inhibited vessels of WT mice, antagonizing EP3 abolishes the remaining vasoconstrictor responses to these agonists, resulting in relaxations sensitive to IP and/or COX blockade. EP3 antagonism also diminishes the contraction evoked by PGI 2 and/or ACh in NOS-inhibited WT abdominal aorta, carotid and renal arteries. These results not only demonstrate that TP contributes only partially to the contraction evoked by PGI 2 , but also suggest that EP3 has an important involvement in the response.
The deletion of TP in TP −/− mice was confirmed by DNA sequencing, mRNA analyses and an elongation of bleeding time 37 . Indeed, abdominal aortas from these mice (which posses a normal MAP, as reported previously 38 ) lost contraction in response to the TxA 2 analogue and TP agonist U46619 even under NOS-inhibited conditions. Notably, in such vessels, not only the contraction to PGI 2 , but also that to PGF 2α , PGE 2 , or PGD 2 was diminished. In contrast, levels of IP, FP and EP3 mRNAs were similar between TP −/− and WT vessels, arguing against that the above reduced PG responses resulted from altered expressions of receptors. Thus, TP, which appears able to be activated by all vasoactive prostanoids that are structurally similar 4 , mediates PGI 2 's contractile activity. Due to practical reasons, we were unable to clearly detect IP, EP3, and FP at the protein level; however, our above idea concurs with results in WT mice and some other species obtained here or previously with pharmacological blockade 8,28,34 . At the same time, responses evoked by low concentrations (0.1-0.3 μM) of PGE 2 in TP −/− vessels reveal a functional role of EP3 unaffected by the TP antagonist used.
Interestingly, we further noted that antagonizing EP3 abolished the remaining contraction, resulting in relaxation in response to PGI 2 in either NOS-inhibited or endothelium-denuded TP −/− abdominal aortas. This suggests that EP3, which exists in medial smooth muscles in a manner similar to that of IP and TP, mediates PGI 2 's contractile response, although its effect is largely offset by IP when TP is absent. In support of the idea, in NOS-inhibited, TP-antagonized WT vessels a relaxation sensitive to IP antagonism was also evoked by PGI 2 following EP3 blockade. Moreover, after EP3 antagonism the contraction to PGI 2 was minimal. This suggests that the part of EP3-mediated activity could be only slightly smaller than that of TP, which alone could also be largely masked by IP-mediated effect. Therefore, the robust contractile response to PGI 2 in WT vessels reflects activities from both TP and EP3 overcoming the effect of concomitantly activated IP. In contrast, EP1 (another vasoconstrictor PGE 2 receptor), EP2 and EP4 (dilator PGE 2 receptors) do not appear to have a role in the vessels studied, as suggested by the lack of effect caused by antagonism or the absence of relaxation to PGE 2 in TP −/− vessels even after EP3 is antagonized 39,40 .
Also, the above effects of TP −/− , TP and/or EP3 antagonism under NOS-inhibited conditions were similarly obtained in responses evoked by ACh or AA, which stimulates endothelial COX to mainly produce PGI 2 28,35 . As seen from EIA measurements, the profile of COX-derived products in aortas was unaltered by TP −/− . Thus, the mechanism for the contraction evoked by endothelial COX metabolites produced in situ is similar to that of PGI 2 . Due to an endothelium-derived hyperpolarizing factor (EDHF)-mediated relaxation concomitantly activated 28,41 , the effect of EP3 antagonism on the response evoked by ACh in NOS-inhibited, TP −/− or TP-inhibited vessels was reflected by abolition of the contractile activity blunting EDHF-mediated relaxation and a relaxation that is sensitive to IP or COX blockade, but adds to that of EDHF. One must note that the EDHF activity in the vessel does not originate from non-COX AA metabolites, as we put forward previously 35 . Indeed, this point is also supported here by the lack of IP-independent relaxation to AA after TP and EP3 were both antagonized. Previously, the contractile role of EP3 in PGE 2 -evoked vasoresponse was established in vessels of mice as well as those of humans 32,42 . In the present study, our results further suggest an intimate link between EP3 and the contractile activity evoked by PGI 2 . Notably, PGD 2 might also act on EP3 to mediate a minor contraction, as revealed by functional studies of its response in TP −/− vessels. These results could again be possibly due to a structural similarity among PGs. In support of this, iloprost, a PGI 2 analog, also activates EP3 43 . Moreover, EP3 antagonism exerts a greater inhibitory effect on PGI 2 -evoked contraction than on that of PGE 2 (whose response via EP3 peaks at 0.3 μM, as seen from its response in TP −/− vessels). This implies not only that the EP3 antagonist used has limited if any, unintended effects on TP, but also that PGI 2 , although it might have a lower potency, is more effective on EP3 than PGE 2 , underscoring the importance of PGI 2 in EP3-mediated vasoconstrictor activities. Besides, the effects of its antagonism among WT vessels studied further indicate that the involvement of EP3 in PGI 2 's vasoconstrictor activity is not limited to any specific vascular bed. Therefore, our above results make important amendments to previous proposals on the mechanisms of PGI 2 or endothelial COX metabolites (which consist mainly of PGI 2 ) in mediating vasoconstrictor responses 20,44 . It should be noted that the contraction to PGI 2 only exists in vessels with limited expression of IP 44 . This is also true in the abdominal aorta examined here where we previously showed that IP expression is lower than in mesenteric arteries (where 0.3 μM PGI 2 almost completely relaxes 2 μM PE-evoked contraction) 28 . A reason for this could be that PGI 2 (the prototype IP agonist) is more potent on IP than TP and/or EP3, leading to PGI 2 being more likely to evoke relaxation than to cause contraction. Indeed, this idea explains why PGI 2 has been recognized as a potent vasodilator in many vascular beds and used clinically as an effective therapy for pulmonary hypertension or peripheral arterial diseases 2,6,45,46 .
On the other hand, it must also be emphasized that the minimal concentration of PGI 2 required to initiate vasoconstrictor activity could be 0.003-0.03 μM (under precontracted conditions), far below the amount (1 ng/ mg 6-keto-PGF 1α can be translated into 2.7 μmol PGI 2 per kg of vessel) released by agonists, such as ACh, or similar to that of it (PGI 2 ) to evoke relaxation in vessels, such as mouse mesenteric arteries even after TP is antagonized 28,30,44 . Also, PGI 2 -mediated contraction or endothelial COX-derived vasoconstrictor activity has been found in many vessels across species (including those of humans), of which some are small or resistance arteries 23,24,29,34,47 . Moreover, PGI 2 's contractile activity exists in vessels that show a dilator response to the agonist 27,32 . As a result, although PGI 2 may cause a hypotensive effect in general, concurrent activities via TP and/or EP3 can negate some of its beneficial effects via IP, especially on local vascular pathology under disease conditions 22,26,36 . For this reason, antagonizing TP and/or EP3 might be needed for an optimal therapeutic effect obtained with PGI 2 or its analogues under clinical conditions.
In contrast to our findings, EP1 antagonism has also been suggested to diminish PGI 2 -evoked contraction 48 . However, the EP1 antagonist used was also a partial antagonist of TP, which was deleted in the vessels we studied 49 , not to mention the variation that might exist among species or vascular beds. Also, the COX inhibitor indomethacin may cause off-target effects 50,51 ; however, this agent has been shown not to alter ACh responses in similar vessels of COX-1 −/− mice 28 . Indeed, IP blockade inhibited the relaxation in a manner similar to that of indomethacin. Thus, the effect of indomethacin noted here can be considered to result mainly from COX inhibition. However, the precise structural properties responsible for different PGs to activate the same receptor or for one PG to act on different receptors still require further investigation. Also, reasons for one PG to evoke contraction mainly through receptors other than its own, e.g. FP of PGF 2α need to be resolved, given that contractions evoked by endothelial COX metabolites can result from non-PGI 2 products, including PGF 2α 19,20,34,52,53 .
In summary, our results demonstrate that TP, which appears able to be activated by all vasoactive prostanoids, only partially mediates PGI 2 's vasoconstrictor activity. Interestingly, our data further suggest that PGI 2 also effectively activates EP3, whose activity along with that of TP can overcome the dilator effect of concomitantly activated IP to produce a robust vasoconstrictor response, and hence imply a novel mechanism for endothelial COX metabolites (which consist mainly of PGI 2 ) in regulating vascular functions.
L-NAME, PE, AA, and ACh were dissolved in distilled water (purged with N 2 for dissolving AA), while PGI 2 was dissolved in carbonate buffer (50 mM, pH 10.0). PGF 2α , PGE 2 , PGD 2 , CAY10441, SQ29548, L798106, and indomethacin were dissolved in dimethyl sulfoxide (DMSO). The final ratio of a solvent (distilled water, carbonate buffer, or DMSO) to working PSS was 0.5/1,000, which doesn't alter the final pH value of the working buffer (pH 7.4). The concentration of an inhibitor or antagonist used was based on previous reports, which would selectively inhibit the effect of its intended target 27,54,55 . Animals and tissue preparation. All procedures were in conformance with the Guide for the Care and for experiments. Mice were killed by CO 2 inhalation. For in vitro functional and biochemical analyses, aortas, carotid and/or renal arteries were isolated and dissected free of adherent tissues with the help of a binocular microscope.
DNA sequencing. The gene mutation in TP −/− mice was verified by sequencing PCR products (the sense and anti-sense PCR primers were 5′-GAA AGG GTA TTT TGT TCC TGA GGC-3′ and 5′-GCT ACC CCC ATG AAG TAG CAC AGG-3′, respectively) of DNA isolated from tail biopsies and performed by Sangon Biotech (Shanghai, China).
RT-PCR and real-time PCR. The preparation of total RNA from whole sections of mouse aortas and RT reactions were performed as described elsewhere previously 28 . First-strand cDNA was synthesized using total RNA (250 ng) and oligo(dT)15 primers (TaKaRa; Dalian, China).
TP mRNA transcripts were detected with RT-PCR. Primers for TP were 5′-CTG GGG GCC TGC TTT CGC CCG G-3′ (sense; using the deleted fragment) and 5′-GTC AGG AAG CAC CAA GAG CC-3′ (antisense), while those for β-actin (internal control) were as described previously 28 . The expected sizes of the RT-PCR products were 530 bp for TP and 300 bp for β-actin.

Blood pressure measurement.
In some experiments, blood pressure in mice (body weight of 26-30 g) was measured using a computerized noninvasive blood pressure system (Kent Scientific Corporation, Torrington, CT, USA). Mice were accustomed to tail-cuff blood pressure measurements for 3 consecutive days, and then blood pressure was measured on the 4 th day. MAP taken from the averaged value of three measurements was used for analysis.
Tail bleeding time assay. To evaluate in vivo bleeding time, WT and TP −/− tails (age, tail size and length matched) were cut 2 mm from the tips, and wounds were then gently wiped with sterilized filter paper every 30 s, until no more blood was visible. The bleeding time was calculated from the ending of cutting to the time when no more blood would be seen on paper.
Analysis of vascular function. Abdominal sections of aortas and main stems of carotid or renal arteries were cut into 1 mm rings. Analysis of vascular function was performed with isometric force measurement as described elsewhere previously 28,30 . For some experiments, the endothelium was denuded by rotating vessel rings around two wires with passive tension kept at 100 mg (endothelial removal was confirmed by absence of relaxation to 10 μM ACh at the end of experiment).
To remove the influence of endothelial NO, vessels were treated with the NOS inhibitor L-NAME (1 mM), under which the response of arteries appears similar to that of eNOS −/− mice 14 . Inhibitors or solvents were added 30 min before the vessel was contracted with an agonist, and was kept in the solution throughout the experiment. The response elicited by an agonist under baseline conditions was expressed relative to the contraction evoked by 60 mM K + , while that during the contraction evoked by PE (2 μM) was expressed as a change of force relative to the value before the application of the agent.
Assay of COX-derived metabolites. Measurement of the PGI 2 metabolite 6-keto-PGF 1α , the TxA 2 metabolite TxB 2 , or PGE 2 was performed by EIA 28,36 . Briefly, after being rinsed of blood components, whole sections of aortas were incubated with PSS at 37 °C for 30 min, followed by exposure to PSS (300 μl) and ACh (10 μM) in 300 μl PSS (37 °C) for 15 min each. Thereafter, vessels were taken out, and 1, 10, or 100 μl of reaction solutions was used for 6-keto-PGF 1α , PGE 2 , or TxB 2 measurements, respectively (2 replicates for each single measurement), using protocols according to instructions of the manufacturer. The amount of 6-keto-PGF 1α , TxB 2 , or PGE 2 was expressed in ng per mg of wet tissue.

Data analysis.
Values were expressed as means ± SEM from n numbers or pools of vessels from different animals. The normality of data sets with n of 5 or more was confirmed using the Kolmogorov-Smirnov test. Thereafter, statistical analyses were performed with a Student's t-test (unpaired) or ANOVA (1-way or 2-way), followed by Bonferroni's or Dunnett's post-hoc test. For some data sets with undeterminable normality (n = 3), the Mann-Whitney U test was used. P < 0.05 was considered to be statistically significant.