Mouse mpkCCD_c14 cells were maintained essentially as described (Hasler et al., 2002). Cells were seeded at a density of 1.5 × 10⁵ cells/cm² on semipermeable filters (Transwell®, 0.4 μm pore size, Corning Costar, Cambridge, MA) and cultured for 8 days. Unless stated otherwise, the cells were exposed to 1 nM of the V2R agonist desmopressin (dDAVP) at the basolateral side during the last 96 h, to maximally induce the AQP2 expression (Li et al., 2006). Cells were incubated with 10 μM indomethacin, 1 μM PGE₂ (both Sigma, St. Louis, MO, USA), 1 μM PGF_2α (Calbiochem, San Diego, CA), 300 nM of EP1/EP3 agonists sulprostone (Sigma, St. Louis, MO, USA), 1 μM of EP4 agonists CAY10580, 0.5 μM of EP4 antagonist Gw627368, 2.5 nM of the EP4 antagonist L161982, 20 μM of EP1 antagonist Sc-51089, or 100 nM of EP1 antagonist Ono-8711 (all Cayman Chemical, Ann Arbor, Michigan, USA) during the last 48 h. The medium was replaced after 24 h, or in experiments using the EP agonists or antagonists, the medium was replaced every 12 h.

MpkCCD_c14 cells grown on 1.13 cm² filters were lysed using 200 μl Laemmli. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis, blotting, and blocking of the polyvinylidene fluoride membranes were carried out as described previously (Kamsteeg et al., 1999). Membranes were incubated for 16 h with 1:3,000-diluted affinity-purified rabbit anti-AQP2 antibodies [R7 (Deen et al., 1994) or Novus Biologicals, Littleton, CO] in Tris-buffered saline Tween-20 (TBS-T) supplemented with 1% w/v nonfat dried milk. Blots were incubated for 1 h with 1:5,000-diluted goat anti-rabbit IgGs (Sigma, St. Louis, MO) as secondary antibodies coupled to horseradish peroxidase. Proteins were visualized using enhanced chemiluminescence (ECL, Pierce, Rockford, IL).

MpkCCD_c14 cells were grown as described above, and total RNA was isolated using the TriZol extraction reagent (Gibco, Life Technologies, Rockville, MD), according to the instructions of the manufacturer. To remove genomic DNA, total RNA was treated with DNase (Promega, Madison, WI) for 1 h at 37°C, extracted with phenol/chloroform, and precipitated. RNA was reverse-transcribed into cDNA using Moloney Murine Leukemia Virus reverse-transcriptase and random primers (Promega, Madison, WI). During cDNA production, a control reaction without the reverse-transcriptase enzyme was conducted to exclude genomic DNA amplification. Exon overlapping primers were designed for prostaglandin receptors (see Table 1). Amplification was performed using the cDNA equivalent of 5 ng RNA for 40 cycles (i.e., 95°C 45 s, 50°C 1 min, and 72°C 1.5 min). β-actin was used as a positive control for cDNA amplification. cDNA from the tissue reported to express the particular receptor was taken along as a positive control. The proper identity of products was confirmed using the restriction analysis.

SYBR Green real-time quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) was performed on an iQ5 Real-Time PCR Detection System from Bio-Rad by utilizing the SYBR Green PCR Master Mix (Applied Biosystems Foster City, CA). Signals for the ribosomal 18S were used to normalize for differences in the amount of starting cDNA.

Samples were prepared as described previously (Schweer et al., 1994) with minor modifications. In brief, cell culture supernatants were spiked with ~1 ng of deuterated internal standards, and the methoximes were obtained through the reaction with an O-methylhydroxylamine hydrochloride-acetate buffer. After acidification to pH 3.5, prostanoid derivatives were extracted, and the pentafluorobenzylesters were formed. Samples were purified by thin layer chromatography, and a broad zone with R_F 0.03–0.4 was eluted. After withdrawal of the organic layer, trimethylsilyl ethers were prepared by the reaction with bis(trimethylsilyl)-trifluoroacetamide and thereafter, subjected to the gas chromatography-tandem mass spectrometry (GC/MS/MS) analysis on a Finnigan MAT TSQ700 GC/MS/MS (Thermo Electron Corp., Dreieich, Germany) equipped with a Varian 3400 gas chromatograph (Palo Alto, CA) and a CTC A200S autosampler (CTC Analytics, Zwingen, Switzerland).

Student's t-test was applied to compare two groups with Gaussian distribution. Comparisons of more than two groups were performed using a one-way ANOVA followed by a Dunnett multiple comparison test. Levene's test was used to compare variances. P-values < 0.05 were considered significant. Immunoblotting signals were analyzed using the Bio-Rad software. Data are presented as mean ± standard error of the mean (SEM).

To analyze the effect of PGE₂ on the AQP2 expression, mpkCCD_c14 cells were grown to confluence for 8 days, either with or without 1 nM of the V2R agonist dDAVP for the last 4 days and with or without 1 μM PGE₂ during the last 48 h. PGE₂ increased the AQP2 abundance in the absence of dDAVP but decreased it in the presence of dDAVP (Figure 2). In the presence of dDAVP, 1 μM PGF_2α also decreased the AQP2 abundance.

To test whether COX inhibition affects the dDAVP-induced AQP2 expression, cells were grown as described above, i.e., the last 4 days in the presence of dDAVP and the last 48 h in the presence of 10 μM indomethacin. Subsequent immunoblotting showed an increased AQP2 abundance with indomethacin (Figure 2), suggesting that dDAVP-treated mpkCCD_c14 cells produce prostanoids, which decrease the AQP2 abundance.

To determine whether mpkCCD_c14 cells produce PGE₂ or other prostanoids, and whether the presence of dDAVP affects the release of these prostanoids, cells were grown as above, i.e., with or without dDAVP for the last 4 days, after which the medium was collected and analyzed for the presence of prostanoids. Prostaglandin concentrations from the fresh medium were subtracted. The major prostanoids released from control cells were PGE₂ and PGF_2α, while levels of PGD₂, 6-keto-PGF_1α (i.e., a stable metabolite of PGI₂), and TxB₂ (i.e., a stable metabolite of TxA₂) were lower and bordering on their detection limit (Figure 3). The dDAVP treatment significantly increased the production of PGD₂ and PGE₂, while PGF₂α levels were decreased. No effect of dDAVP was observed on the release of 6-keto-PGF_1α or TxB₂.

The effects of prostaglandins on the AQP2 expression are conferred by effects on their respective receptors. Immunoblotting was unsuitable to examine the expression of the individual prostaglandin receptors (not shown). Therefore, we determined the mRNA expression of the prostaglandin receptors in mpkCCD cells using the RT-PCR.

From unstimulated cells, cDNA products of the expected size were obtained for EP1, EP4, FP, and TP receptors (Figure 4A). While PCR products for EP2, EP3, or DP receptors were found in control tissues, no products were obtained in mpkCCD_c14 cells, indicating that these receptors are not expressed. A detectable expression of the IP receptor was inconsistent. The same receptors were expressed in mpkCCD cells treated with dDAVP (not shown).

To test if the levels of the expressed prostanoid receptors were influenced by dDAVP, we determined their relative expression by using the qRT-PCR. dDAVP increased the expression of the EP1 and FP receptor, while the expression of the EP4 receptor was significantly decreased (Figure 4B). No difference was detected in the expression of the TP receptor.

As the EP1/FP receptors and EP4 receptors are coupled to Gi/Gq and Gs (Figure 1), respectively, an altered activation of these receptors due to their changes in the expression with dDAVP could explain the differential effect of prostanoids on the AQP2 abundance. To further explore the roles of the different PGE₂ receptor subtypes in mediating the effects of PGE₂ on AQP2 levels, we used EP receptor-specific agonists and antagonists.

MpkCCD cells were grown as described above, i.e., stimulated with or without dDAVP, and incubated with the EP4 agonist CAY10580 or the EP1/EP3 agonist sulprostone (Kiriyama et al., 1997; Billot et al., 2003) during the last 48 h. As the EP3 receptor is not expressed in mpkCCD cells (Figure 4A), sulprostone will act as a specific EP1 agonist in these cells. Consistent with a contribution of EP4 to the prostanoid-stimulated AQP2 abundance in unstimulated cells, CAY10580 and PGE₂ increased the AQP2 abundance as compared with unstimulated cells or cells incubated with sulprostone (Figure 5A). In cells stimulated with dDAVP, however, CAY10580 did not affect the AQP2 abundance, while both sulprostone and PGE₂ decreased the AQP2 abundance, therewith, illustrating an important contribution of the EP1 receptor in reducing the AQP2 abundance in dDAVP-stimulated mpkCCD, cells (Figure 5B).

To further investigate the role of the EP4 receptor in the prostanoid-induced AQP2 abundance, mpkCCD cells were treated with PGE₂ with or without the EP4 antagonists L161982 and Gw627368. While PGE₂ alone again increased the AQP2 abundance significantly, Gw627368 completely blocked the PGE₂-mediated AQP2 increase, whereas L161982 had a tendency to decrease the AQP2 expression relative to cells treated with PGE₂ alone (Figure 5C).

To investigate the role of EP1 in the PGE₂-mediated AQP2 decrease in dDAVP-treated cells, mpkCCD cells were stimulated with dDAVP and incubated with or without PGE₂ and the specific EP1 antagonists Sc-51089 or Ono-8711. Both antagonists fully prevented the PGE₂-mediated downregulation of AQP2 (Figure 5D), illustrating an important contribution of the EP1 receptor in the regulation of AQP2.

Prostaglandin E₂ reduce the AVP-stimulated water reabsorption in the collecting duct (Hebert et al., 1990; Nadler et al., 1992), while in the absence of AVP, ex vivo water permeability is increased by PGE₂ (Sakairi et al., 1995). A short-term action of PGE₂ is to alter the localization of AQP2 at the plasma membrane (Zelenina et al., 2000; Nejsum et al., 2005; Olesen et al., 2011). Here, we showed that long-term PGE₂ affects the abundance of the AQP2 protein. PGE₂ attenuated the dDAVP-induced AQP2 expression, while PGE₂ stimulated the AQP2 abundance in the absence of dDAVP. In addition, dDAVP-stimulated AQP2 levels were decreased after the application of PGF_2α, explaining the inhibition of water reabsorption in the collecting duct observed after the PGF_2α treatment (Zook and Strandhoy, 1981; Hebert et al., 2005). Furthermore, blocking the prostaglandin production by indomethacin increased the AQP2 abundance, showing that the dDAVP-stimulated AQP2 abundance is likely reduced due to the effects of endogenously produced prostaglandins. The major prostaglandins produced in mpkCCD cells are PGE₂ and PGF_2α. The dDAVP stimulation significantly increased both the production of PGE₂ and PGD₂, while levels of PGF_2α were decreased. In agreement with these findings, it has been shown that AVP stimulates the PGE₂ synthesis in isolated collecting ducts (Schlondorff et al., 1985; Bonvalet et al., 1987).

Consistent with previous studies, the PGE₂ receptors expressed in mpkCCD_c14 cells are EP1 and EP4 (Olesen et al., 2016). Our experiments using receptor antagonists and agonists show that it is the EP4 receptor that is involved in the stimulatory effect of PGE₂ on the AQP2 expression in mpkCCD cells. The EP4 receptor can couple to Gs-stimulated cAMP generation, thereby activating the same pathway as AVP. Incubation with dDAVP increased the expression of the EP1 receptor in mpkCCD cells but decreased the expression of the EP4 receptor. Additionally, our experiments showed that the activation of EP1 is the pathway by which PGE₂ inhibits the dDAVP-induced AQP2 expression in mpkCCD cells (Figure 6). The EP1 receptors can couple to Gq and increase cytosolic Ca²⁺ and activate protein kinase C (PKC; Funk et al., 1993; Watabe et al., 1993). In microperfused collecting ducts, the inhibitory effect of PGE₂ on AVP-stimulated water permeability was dependent on the activity of PKC (Hebert et al., 1990; Nadler et al., 1992). PKC activation also promotes AQP2 endocytosis, similar to PGE₂ (Zelenina et al., 2000; Van Balkom et al., 2002; Nejsum et al., 2005), and increases AQP2 ubiquitination, leading to lysosomal degradation (Kamsteeg et al., 2006). This suggests that the EP1 activation will decrease the AQP2 abundance by lysosomal degradation (Figure 6).

The expression of the FP receptor was increased by dDAVP incubation in mpkCCD cells. As the activation of the FP receptor inhibits water reabsorption, the increase in the FP expression might be a compensatory mechanism to counteract AVP stimulation, similar to the increase in the EP1 expression.

As no DP receptor was detected in mpkCCD_c14 cells, the role of the dDAVP-stimulated increase in the PGD₂ production after dDAVP incubation is unclear. However, PGD₂ has been shown to bind to the FP receptor with an affinity close to that for the DP receptor, indicating that PGD₂ may act on the FP receptor (Kiriyama et al., 1997). The increase in PGD₂ might counteract the dDAVP-induced increase in the AQP2 expression, although levels are low compared with the PGE₂ and PGF_2α production.

While the TP receptor is expressed in mpkCCD_c14 cells, the expression of the IP receptor is inconclusive. Both thromboxane and PGI₂ were produced in very low amounts in mpkCCD_c14 cells, and the production was not affected by dDAVP. Whether these prostanoids have any role in water reabsorption remains unclear.

A limitation of our study is that all experiments are performed in mpkCCD cells. However, a problem with in vivo studies investigating the effect of prostaglandins on the collecting duct is that these studies are complicated by the effect of prostaglandins on AVP release and on medullary osmolality, both of which will influence the AQP2 expression (Yamamoto et al., 1976; Stoff et al., 1981; Hasler et al., 2005). To study the effect of prostaglandins directly on principal cells, experiments were performed in mpkCCD cells, shown to display the essential functionalities characteristic of principal cells like the AVP-regulated AQP2 expression and aldosterone-mediated sodium transport via the epithelial sodium channel (Bens et al., 1999; Hasler et al., 2002).

The major prostaglandins produced in our cell system were PGE₂ and PGF_2α, which is in agreement with in vivo findings, showing that PGE₂ is the most abundant prostanoid in both the renal cortex and medulla, followed by PGI₂ and PGF_2α (Qi et al., 2006). The synthases involved in the production of PGD₂, PGE₂, and PGF_2α are detected in the nephron (Vitzthum et al., 2002; Sakurai et al., 2005), where the production of PGE₂ and PGF_2α has been shown to occur mainly in the collecting ducts (Farman et al., 1987). Neither PGI synthase nor thromboxane synthase mRNA is detected in any tubular structure (Vitzthum et al., 2002).

The effects of prostaglandins on the AQP2 expression are conferred by PG receptors. In mpkCCD_c14 cells, EP1, EP4, and FP receptors are found, in agreement with expression in the collecting duct (Breyer et al., 1998; Saito et al., 2003).

In line with our data showing the role of EP4 in the stimulatory effect of PGE₂ on AQP2, a study by Gao et al. demonstrates that disruption of EP4 in the collecting duct impaired the urinary concentration by decreasing the AQP2 abundance and apical membrane targeting, providing evidence that EP4 can regulate the urine concentration in vivo (Gao et al., 2015). In addition, a selective EP4 agonist has been shown to increase the urine osmolality, decrease the urine volume, and increase the AQP2 expression in a mouse model for congenital NDI (Li et al., 2009).

In agreement with our findings that the activation of the EP1 receptor decreases the AVP-induced AQP2 expression, the stimulation of the EP1 receptor has been shown to decrease the vasotocin-induced osmotic water permeability of the frog urinary bladder, a model system of the collecting duct (Bachteeva et al., 2007). In addition, EP1-knockout mice have a urinary concentrating defect (Kennedy et al., 2007), and recent studies show that PGE₂ does not decrease AVP-mediated water transport in isolated collecting ducts of these mice (Nasrallah et al., 2018). Taken together with the present data, this suggests that EP1 conveys both acute and long-term modulation of the V2R activity.

Furthermore, TP and IP receptors are mainly localized in the glomerulus and vasculature, respectively, but have also been located in the collecting duct (Takahashi et al., 1996; Komhoff et al., 1998), in agreement with the expression seen in mpkCCD_c14 cells. Based on our mpkCCD data, however, we anticipated that the IP receptor does not have a major impact on the principal cell AQP2 expression in the presence or absence of AVP.

None of the receptors DP, EP2, and EP3 seems to be expressed in mpkCCD_c14 cells. While DP is also not expressed in the kidney, the presence of EP2 along the nephron is a matter of considerable debate (Breyer and Breyer, 2000; Olesen and Fenton, 2013). However, a previous study has shown that functionally, the collecting duct can respond to the stimulation of the EP2 receptor (Olesen et al., 2011).

The inhibitory effects of PGE₂ on AVP-induced water reabsorption have, besides via the activation of EP1, also been suggested to occur through the activation of EP3 (Hebert et al., 1993; Fleming et al., 1998). Our cell model does not express the EP3 receptor, which was found in vivo by the in situ hybridization and RT-PCR on microdissected tubules to be expressed in the collecting duct (Breyer et al., 1998). However, a study using single-cell RNA-Seq of intercalated and principal cells from the mouse kidney demonstrated that EP3 was selectively expressed in collecting duct-intercalated cells, while EP1 and EP4 were expressed in the principal cells (Chen et al., 2017). In addition, EP3-knockout mice exhibit a similar urine-concentrating ability during basal conditions as well as in response to AVP compared with wild-type mice, arguing against a role of EP3 in the AQP2 regulation (Fleming et al., 1998). The exact role of the EP3 receptor in the AQP2 regulation needs further investigation.

It is interesting to note that, while dDAVP increases the PGE₂ production and release, the mRNA expression of the EP4 receptor is reduced, whereas that of the EP1 receptor is increased. As both receptors are bound and activated by PGE₂, these data suggest that it is not the agonist per se that determines the expression level of the receptors. Instead, our data indicate that the signaling cascade that is mainly activated exerts a negative feedback regulation on receptors stimulating the same pathway and a positive feedback on receptors activating an opposite pathway: dDAVP increases the cAMP-AQP2 pathway, which can be stimulated by EP4, whereas EPl activates a pathway that leads to a decreased AQP2 expression and water permeability.

The same antagonizing mechanism can be seen in response to endothelin, which counteracts the AVP-mediated water permeability (Edwards et al., 1993), and at the same time, leads to an increased expression of the vasopressin V2 receptor in the inner medullary collecting duct of the rat (Sonntag et al., 2004). Similar to this antagonizing mechanism, dDAVP increases the mRNA levels of the purinergic receptor subunit P2Y₂ in mpkCCD cells and targets the subunits P2Y₂ and P2X₂ to the plasma membrane, where the activation of these receptors leads to the AQP2 internalization and a decrease in the water permeability (Wildman et al., 2009). A similar mechanism can be seen with the hormone angiotensin II, which increases renal proximal sodium reabsorption but at the same time increases expression of the D4 dopamine receptor in renal proximal tubule cells, which activation will decrease sodium reabsorption, thereby counteracting the direct effect of angiotensin II (Tang et al., 2017).

In conclusion, our study shows that in mpkCCD_c14 cells, both PGE₂ and PGF_2α decrease the dDAVP-stimulated AQP2 abundance, while in the absence of dDAVP, PGE₂ increases AQP2 levels. Furthermore, our study suggests that EP4 mediates the PGE₂-induced increase in the AQP2 abundance in the absence of dDAVP, while the PGE₂-mediated decrease in the AQP2 abundance in the presence of dDAVP is likely mediated via EP1. This paradoxical difference in response to PGE₂ is likely explained by the different receptor subtype expression induced by the dDAVP treatment, leading to an increase in EP1 and a decrease in EP4.

Based on our data above that a negative feedback is mediated by the signaling pathways activated instead of the agonist, we hypothesized that in vivo AVP increases, besides AQP2, the expression of EP1 and decreases the expression of EP4 receptors. Consequently, in conditions with the increased PGE₂ release, such as with lithium-NDI or bilateral uteral obstruction, the AVP-induced AQP2 expression would be reduced via the activation of these EP1 receptors.