Small-Molecule Screening Identifies Modulators of Aquaporin-2 Trafficking

Abstract

In the principal cells of the renal collecting duct, arginine vasopressin (AVP) stimulates the synthesis of cAMP, leading to signaling events that culminate in the phosphorylation of aquaporin-2 water channels and their redistribution from intracellular domains to the plasma membrane via vesicular trafficking. The molecular mechanisms that control aquaporin-2 trafficking and the consequent water reabsorption, however, are not completely understood. Here, we used a cell-based assay and automated immunofluorescence microscopy to screen 17,700 small molecules for inhibitors of the cAMP-dependent redistribution of aquaporin-2. This approach identified 17 inhibitors, including 4-acetyldiphyllin, a selective blocker of vacuolar H⁺-ATPase that increases the pH of intracellular vesicles and causes accumulation of aquaporin-2 in the Golgi compartment. Although 4-acetyldiphyllin did not inhibit forskolin-induced increases in cAMP formation and downstream activation of protein kinase A (PKA), it did prevent cAMP/PKA-dependent phosphorylation at serine 256 of aquaporin-2, which triggers the redistribution to the plasma membrane. It did not, however, prevent cAMP-induced changes to the phosphorylation status at serines 261 or 269. Last, we identified the fungicide fluconazole as an inhibitor of cAMP-mediated redistribution of aquaporin-2, but its target in this pathway remains unknown. In conclusion, our screening approach provides a method to begin dissecting molecular mechanisms underlying AVP-mediated water reabsorption, evidenced by our identification of 4-acetyldiphyllin as a modulator of aquaporin-2 trafficking.

Arginine-vasopressin (AVP) stimulates vasopressin V2 receptors on the surface of renal collecting duct principal cells and thereby production of cAMP, which activates protein kinase A (PKA). Initiation of this signaling results in the redistribution of the water channel aquaporin-2 (AQP2) from intracellular vesicles into the plasma membrane by an exocytosis-like process. The membrane insertion of AQP2 facilitates water reabsorption from primary urine and fine-tunes blood osmolality.^1,2 Loss of AVP secretion causes central diabetes insipidus, loss of function mutations in vasopressin V2 receptors, or AQP2 lead to nephrogenic diabetes insipidus,^3–5 whereas pathologically elevated levels of AVP with excessive water retention are associated with chronic heart failure or the syndrome of inappropriate antidiuretic hormone secretion.⁶

AVP induces the PKA-catalyzed phosphorylation of AQP2 at serine 256 (S256). This phosphorylation is the key trigger for the redistribution of AQP2 from intracellular vesicles into the plasma membrane.^7–11 AVP also induces phosphorylations of S264 and S269, which are associated with a predominant plasma membrane localization of AQP2.^12–17 Under resting conditions, AQP2 is phosphorylated at S261.¹² AVP mediates dephosphorylation of S261.^15,18 This is associated with decreased polyubiquitination and proteasomal degradation and an enhanced AQP2 abundance, which contributes to the increase in water reabsorption of the collecting duct in response to AVP.¹⁹

Although several proteins controlling AQP2 trafficking were identified and the paths of AQP2 to and from the plasma membrane are defined in general terms,^1,17 the molecular details underlying AQP2 trafficking are unclear. We present a novel, unbiased, high-throughput cell-based assay that identifies small-molecule inhibitors of the cAMP-dependent redistribution of AQP2. Identification of the targets of candidate molecules reveals new proteins and mechanisms controlling AQP2.

Results

High-Throughput Screening Identifies Small-Molecule Inhibitors of the cAMP-Dependent AQP2 Redistribution

Mouse collecting-duct cells stably expressing human AQP2 (MCD4 cells²⁰) were used to establish a high-throughput assay to identify small-molecule inhibitors of the cAMP-dependent redistribution of AQP2 from intracellular vesicles into the plasma membrane (Supplemental Figure 1). MCD4 cells were incubated with each of the 17,700 small molecules from the ChemBioNet library (40 µM).²¹ Forskolin, a direct activator of adenylyl cyclases, was added to induce the redistribution of AQP2. The localization of AQP2 and cortical F-actin as plasma membrane marker were evaluated by automated immunofluorescence microscopy (Figure 1A). The localization of AQP2 was expressed as the ratio of fluorescence signal intensity at the plasma membrane to intracellular fluorescence signal intensity (Figure 1B).^11,22 Forskolin induced the redistribution of AQP2 from a perinuclear localization to the plasma membrane (ratio, 1.40±0.1). As expected, blocking PKA with the kinase inhibitor H89^23,24 prevented the AQP2 redistribution^7,8,11,19 (ratio, 0.91±0.1) (Figure 1, A and B). In previous studies cells were incubated with H89 for 30 minutes,^11,22 whereas we incubated cells for 2 hours, as long as with the library compounds. This most likely leads to the dramatic change in the localization of AQP2 compared with earlier studies. On the basis of the ratios determined in the presence of forskolin (1.4) and the combination of forskolin and H89 (0.9), ratios ≤1.2 were considered to indicate low plasma membrane abundance of AQP2 (Supplemental Figure 2). Treatment with forskolin in the presence of 83 of the library compounds resulted in ratios ≤ 1.2 (Supplemental Figure 2 and Supplemental Table 1), defining them as inhibitors of the forskolin-induced AQP2 redistribution; 17 of these compounds (Table 1) inhibited in a concentration dependent manner (40 µM, 4 µM, and 0.4 µM). Fourteen of the 17 were commercially available and could thus be tested in secondary screens using rat primary inner medullary collecting duct (IMCD) cells expressing AQP2 endogenously.^19,22 In this model, five of the compounds inhibited the forskolin-induced redistribution of AQP2 (Table 1), two of which were further characterized: 4-acetyldiphyllin (4AD; ratio, 0.88; Table 1) and triazolpropenon (ratio, 1.06; Table 1).

Figure 1.

Identification of 4AD as an inhibitor of the forskolin (FSK)-induced AQP2 redistribution in MCD4 cells by cell-based high-throughput screening. (A) MCD4 cells were seeded into 384-well plates, left untreated, treated with forskolin (10 µM, 20 minutes), or incubated with each of the 17,700 small molecules from the ChemBioNet small-molecule library (40 µM) for 100 minutes before forskolin (10 µM) was added for an additional 20 minutes. As a control, the forskolin-induced redistribution of AQP2 was inhibited with H89 (30 µM, 2 hours), a known blocker of the AQP2 translocation. The cells were fixed with paraformaldehyde, and AQP2 was detected by immunofluorescence microscopy using specific antibodies (H27) and Cy5-coupled antirabbit secondary antibodies (red).⁵¹ F-actin was visualized using TRITC-conjugated phalloidin (green). Fluorescence signals were detected using an automated screening microscope (10× magnification). The magnified views are from the indicated squares in the merge. Shown are representative images from one of three independent experiments. (B) The chemical structure represents 4AD. The intensities of intracellular and plasma membrane immunofluorescence signals arising from AQP2 were determined and related to nuclear signal intensities (mean ± SEM; three independent experiments). The ratios of plasma membrane to intracellular fluorescence signal intensities were calculated. According to the ratios determined for cells treated with forskolin alone (1.4) and forskolin in combination with H89 (0.9), ratios <1.2 indicate a predominant intracellular localization. Values significantly different from forskolin-treated cells are indicated (***P<0.001).

Table 1.

Small molecule inhibitors of the cAMP-induced redistribution of AQP2.

No.	ID	Structure	Formula	Ratio	Plate	Well	Molecular Weight	Partition Coefficient	H-Acc	H-Don	IUPAC Nomenclature
1	4-acetydiphillin (4AD) 216407		C23H18O8	0.88	115	N9	422.4	3.08	6	0	9-benzo[d][1,3] dioxol- 5-yl)-6,7-dimethoxy-1-oxo-1,3-dihydronaphtho[2,3-c]furan-4-yl acetate
2	212678		C23H24FN3O2S	0.84	105	K1	425.5	4.99	4	0	(8-ethyl-3-(4-fluorophenyl)-2-thioxo-1,4,8-triazaspiro[4.5]dec-3-en-1-yl)(4-methoxyphenyl)methanone
3	212634		C31H36N4O4S	0.88	104	D14	560.7	5.2	6	3	N-{2[benzyl (isopropyl)amino] ethyl}-6-[(4isopropylphenyl)sulfamoyl]-4-oxo-1H-quinoline-3-carboxamide
4	202336		C18H22N2O2	1	75	P11	298.4	1.69	3	3	3-((3-(9H-carbazol-9-yl) 2 hydroxypropyl)amino)propan-1-ol
5	209614 triazolpropenon		C17H10Cl3N3O2	1.06	96	K6	394.6	4.58	4	1	(Z)-3-(3-chloro-6-hydroxycyclohexa-2,4-dienyl)-1-(2,4-dichlorophenyl)-2-(11H,2,4-triazol-1-yl)prop-2-en-1-one
6	214590		C23H20N4O2	0.69	110	K18	384.4	3.24	2	2	N-(naphthalen-1-yl)-4-(2-oxo-3H-1,3-benzodiazol-1-yl)-3,6-dihydro-2H-pyridine-1-carboxamide
7	205653		C24H18N2O7S2	0.78	85	I5	510.5	3.52	6	4	N,N′-bis(4-hydroxyphenyl)dibenzo[b,d]furan-2,8-disulfonamide
8	216086		C25H32N4O2	0.79	114	L17	420.5	3.47	4	1	N-{3-[ethyl(3-methylphenyl)amino]propyl}-2-[3-(4-methylphenyl)-6-oxo-4,5-dihydropyridazin-1-yl]acetamide
9	214645		C19H23N3	0.8	110	J9	293.4	3.62	2	1	9-cyclohexyl-2,3,5,6-tetrahydro-1H-3,4,6a-triazafluoranthene
10	203240		C13H18NS2	0.89	78	O17	252.4	1.6	0	0	2-cyclohexyl-5-ethyl-2,3-dihydrothieno[3,2-d]isothiazole
11	215223		C22H20BrNO4S	0.9	112	M21	474.4	4.23	3	1	5-[(4-bromobenzenesulfonyl)methyl]-N-(1,2,3,4-tetrahydronaphthalen-1-yl)furan-2-carboxamide
12	203313		C17H21N3O3	1	78	A16	315.4	0.79	3	1	N-(3,5-dimethylphenyl)-2-(1,4 dioxohexahydropyrrolo[1,2-a] pyrazin-2(1H)-yl) acetamide
13	201213		C22H20N2O3	1.02	72	I18	360.4	4.47	5	1	N-(3,4,5-trimethoxyphenyl)acridin-9-amine
14	208677		C21H15N4O2S	1.07	93	J13	387.4	1.13	5	0	2-(1,3-benzothiazol-2-yl)-5-benzyl-4-methyl-3-oxo-3,5-dihydro-2H-pyrazolo[4,3-c]pyridin-6-olate
15	202329		C22H21ClN2O3S	1.12	75	B11	428.9	3.69	3	1	N-(3-(9H-carbazol-9-yl)-2-hydroxypropyl)-N-(4-chlorophe nyl)methane-sulfonamide
16	204332		C18H19NO4	1.16	81	G6	313.3	3.84	4	1	3-[(3,4dimethyl-phenyl)amino]-6,7-dimethoxy-3H-2-benzofuran-1-one
17	209044		C20H15NO5S2	1.18	94	H17	413.5	4.34	5	0	N-(2-oxonaphtho[2,1-d][1,3]oxathiol-5-yl)-N-tosylacetamide

Primary screening of 17,700 small molecules identified 83 inhibitors of the cAMP-induced redistribution of AQP2 in MCD4 cells. Secondary screening revealed that 17 chemically highly diverse hits inhibited the AQP2 redistribution in a concentration-dependent manner in MCD4 cells (compounds 1–17). Compounds 1–5 also inhibited the redistribution in IMCD cells. Compounds are ranked according to increasing ratios of plasma membrane/intracellular AQP2 fluorescence signal intensities. Compounds with ratios ≤ 1.2 were defined as inhibitory (a predominant intracellular localization of AQP2). Compound ID numbers (ID), compound library positions (plate, well), molecular weight, numbers of proton donor (H-Don.) and acceptor (H-Acc.) moieties are given for each substance. Compound names are according to the International Union of Pure and Applied Chemistry nomenclature. IUPAC, International Union of Pure and Applied Chemistry.

4AD Prevents the cAMP-Induced Redistribution of AQP2

In resting IMCD cells, AQP2 is located intracellularly; in the presence of forskolin, it predominantly resides in the basolateral plasma membrane, while a minor fraction of AQP2 redistributes to the apical plasma membrane (Figure 2A).^11,19,22 This resembles the situation in the collecting duct where in response to AVP the AQP2 amount in the basolateral plasma membrane can be equal to or greater than that in the apical plasma membrane.² 4AD concentration-dependently inhibited the forskolin-induced AQP2 redistribution (Figure 2A). Cell-surface biotinylation assays were used to quantify the effect. IMCD cells were treated as indicated above; cell-surface proteins were biotinylated and precipitated, and AQP2 and cadherin (loading control) were detected by Western blotting. Densitometric analysis confirmed that 4AD significantly reduced the forskolin-induced increase in plasma membrane abundance of AQP2 (Figure 2B); 4AD alone did not alter the membrane abundance of AQP2 compared with untreated controls. In line with these observations, 4AD prevented the forskolin-induced increase of AQP2 phosphorylation at S256 (AQP2-pS256) and S269 (AQP2-pS269) at the plasma membrane (Figure 3). The reason for the nuclear staining with the anti-pS269 antibody in the absence of forskolin is unclear; it also occurred with a custom-made antibody²⁵ (data not shown), suggesting that the antibodies cross-react with a nuclear protein containing a similar epitope. The effective concentrations of 4AD did not induce any visible signs of toxicity in MCD4 or IMCD cells (24-hour toxicity assay) (Supplemental Figure 3B).

Figure 2.

4AD concentration-dependently inhibits the forskolin (FSK)-induced redistribution of AQP2 in primary IMCD cells. (A) Primary IMCD cells were left untreated and were stimulated with forskolin (10 µM, 20 minutes) or with a combination of 4AD (40, 4, or 0.4 µM, 30 minutes) and forskolin. AQP2 was detected by immunofluorescence microscopy using specific primary antibodies (H27) and Cy3-coupled antirabbit secondary antibodies (red).⁵¹ Nuclei were stained with DAPI (blue). Shown are representative images from one of three independent experiments. (B) Upper panel. Cell-surface biotinylation assays were carried out to quantify the effect of 4AD on the localization of AQP2. To detect AQP2 on the surface of IMCD cells, surface proteins were biotinylated with EZ-Link Sulfo-NHS-SS-Biotin, precipitated with streptavidin-agarose, separated by SDS-PAGE, and AQP2, and as a loading control cadherin were detected by Western blotting (shown is a representative blot of three independent experiments). Lower panel. The amounts of complex (compl) glycosylated AQP2 and the high mannose (hm) form of AQP2 and nonglycosylated (ng) AQP2 on the blots were quantitatively analyzed by densitometry; each bar of the histogram represents the total amounts of the three AQP2 forms (mean ± SEM). *P<0.05, **P<0.01, and ***P<0.001 versus forskolin-treated cells.

Figure 3.

4AD directs the forskolin (FSK) induced phosphorylation of S269 (pS269) to intracellular domains and prevents S256 phosphorylation of AQP2 in primary IMCD cells. (A) The cells were left untreated, incubated with forskolin (10 µM, 20 minutes), 4AD (4 µM, 30 minutes), or a combination of 4AD and FSK. AQP2 was detected using specific primary antibodies (C-17, Santa Cruz) and Cy2-coupled antigoat secondary antibodies (green).¹⁹ AQP2-pS256 was detected using specific antiphospho primary and Cy3-coupled secondary antibodies.²⁵ AQP2-pS269 was detected using specific antiphospho primary (p112–269, Phosphosolutions) and Cy3-coupled secondary antibodies. Nuclei were stained with DAPI (blue). Shown are representative images from one of three independent experiments. (B) The intensities of intracellular and plasma membrane immunofluorescence signals arising from AQP2, AQP2-pS256, and AQP2-S269 were determined and related to nuclear signal intensities (n≥30 cells per condition; mean ±SEM; three independent experiments). The ratios of plasma membrane to intracellular fluorescence signal intensities were calculated. Ratios >1 indicate a predominant localization at the plasma membrane. Values significantly different from control cells and cells treated with 4AD or forskolin and 4AD are indicated (***P<0.001).

4AD Increases Intravesicular pH and Causes Accumulation of AQP2 in the Golgi Compartment

4AD is a selective blocker of vacuolar H⁺-ATPases(V-ATPases);²⁶ to compare its effects with another inhibitor of V-ATPases, we used bafilomycin A1 (baf A1).^27,28 Baf A1 also blocked the forskolin-induced AQP2 translocation in IMCD cells (Figure 4A). To investigate whether baf A1 and 4AD inhibited intravesicular acidification, MCD4 cells were incubated with LysoTracker RED DND-99, which stains acidic intracellular compartments (red vesicular structures in Figure 4B). Both 4AD and baf A1 shifted intravesicular acidity to a higher pH, as indicated by the strong decrease of the mean fluorescence signal intensity per µm² of individual cells (Figure 4B).

Figure 4.

Inhibition of V-ATPase with 4AD or baf A1 increases acidity of intracellular vesicles in IMCD cells and is associated with inhibition of the FSK-induced redistribution of AQP2 from intracellular vesicles to the plasma membrane. (A) IMCD cells were left untreated or treated with baf A1 in the indicated concentrations (30 minutes) in the absence or presence of forskolin (10 µM, 20 minutes). AQP2 was detected by immunofluorescence microscopy using specific primary antibody H27 and Cy3-coupled antirabbit secondary antibodies (red).⁵¹ Nuclei were stained with DAPI (blue). Representative images from one of three independent experiments are shown. (B) Upper panel. To visualize acidic compartments in MCD4 cells in the absence or presence of the V-ATPase inhibitors 4AD or baf A1, the cells were incubated with the acidotropic dye LysoTracker Red DND-99 (75 nM, 30 minutes), which is retained in acidified compartments (red vesicular structures). Nuclei were stained with Hoechst 33258 (blue). Shown are representative images from one of three independent experiments. Lower panel. The intensities of fluorescence signals arising from Lysotracker Red DND-99 were determined in individual cells (>50 cells per condition; three independent experiments; mean ± SEM). Statistically significant differences versus untreated cells are indicated (***P<0.001).

Baf A1 has previously been shown to arrest AQP2 in the trans-Golgi network (TGN) in LLC-PK₁ cells, a porcine cell line stably expressing AQP2.^29,30 In MCD4 cells, 4AD caused an accumulation of AQP2 in perinuclear patches (Figure 1, A and B) resembling those in LLC-PK₁ cells that represent the TGN.^17,30 In IMCD cells, 4AD did not induce the formation of patches clearly classifiable as TGN (Figure 2A). To confirm, however, that AQP2 is located in the Golgi in the presence of 4AD, we performed co-localization analyses of AQP2 and two Golgi proteins: (1) Golgi 58K (formiminotransferase cyclodeaminase), distributed throughout the Golgi, and (2) a marker for the endoplasmic reticulum Golgi intermediate compartment, ERGIC-53/p58. Figure 5 shows co-localization of AQP2 with both proteins in resting IMCD cells and a decrease in the co-localization upon treatment with forskolin when AQP2 exits intracellular compartments, including the Golgi, en route to the plasma membrane. Combined with forskolin 4AD decreased the co-localization with ERGIC-53/p58 and increased it with Golgi 58K, indicating that 4AD caused AQP2 accumulation in the Golgi. Whether this effect involves inhibition of Golgi V-ATPase is unclear; it is likely because inhibition of V-ATPase with baf A1 inhibits acidification of Golgi vesicles.³¹

Figure 5.

4AD causes an accumulation of AQP2 in the Golgi compartment. (A) IMCD cells were seeded on 12-mm cover slides, left untreated or treated with 4AD (4 µM or 0.4 µM, 30 minutes). AQP2 was detected with specific primary (C17, Santa Cruz) and secondary Cy2-coupled antibodies (red). ERGIC-53/p58 and Golgi apparatus 58K proteins were detected with specific primary and Cy5-conjugated secondary antibodies (both green). Fluorescence signals were visualized by laser-scanning microscopy. Shown are representative images from one of three independent experiments. (B) The co-localization of AQP2 with ERGIC-53 and Golgi apparatus 58K was determined from the images in part A using ZEN2010 software (Carl Zeiss; >25 individual cells per condition; three independent experiments; mean ±SEM). Statistically significant differences are indicated (*P<0.05; ***P<0.001). FSK, forskolin.

4AD Inhibits Forskolin-Induced Phosphorylation of AQP2 at S256 without Affecting cAMP Levels, Global PKA Activity, or Phosphorylations at S261 and S269

To further elucidate the molecular mechanisms underlying the inhibitory effect of 4AD, we investigated whether 4AD influences the cAMP/PKA signaling pathway. 4AD did not alter AVP- or forskolin-induced cAMP formation in IMCD cells (Figure 6A). cAMP is hydrolyzed by phosphodiesterases (PDEs). We had previously shown that nonselective inhibition of PDEs with 3-isobutyl-1-methylxanthine (IBMX) or specific inhibition of the PDE4 family with rolipram enhances AVP- and forskolin-induced increases in cAMP and the redistribution of AQP2 to the plasma membrane in IMCD cells.²² 4AD did not visibly modify the effect of IBMX on the localization of AQP2 (Figure 6B), suggesting that it does not influence cAMP hydrolysis by PDEs.

Figure 6.

4AD does not affect formation of cAMP or global PKA activity. (A) Primary IMCD cells were left untreated, treated with 4AD (30 minutes), forskolin (FSK; 20 minutes), or AVP (20 minutes) in the indicated concentrations. Where indicated, cells were treated with 4AD for 30 minutes before addition of AVP or forskolin for an additional 20 minutes. Cell lysates were prepared and cAMP concentrations were determined by radioimmunoassay (three independent experiments, triplicates per condition; mean ± SEM; ***P< 0.001). (B) IMCD cells were seeded on 12-mm cover slides, left untreated, or treated with forskolin (10 µM, 20 minutes), IBMX (250 µM, 30 minutes), or the indicated combinations of forskolin and 4AD. AQP2 was detected by immunofluorescence microscopy using specific primary antibody H27 and Cy3-coupled antirabbit secondary antibodies (red).⁵¹ Nuclei were stained with DAPI (blue). Shown are representative images from one of three independent experiments. (C) MCD4 cells were left untreated, treated with forskolin (10 µM, 20 minutes), or treated with 4AD in the absence or presence of forskolin as indicated. Upper panels. Cell lysates were prepared and PKA activity was determined by measuring its ability to phosphorylate the substrate peptide PepTag A1. Where indicated, cAMP (1 µM) was added to induce maximal PKA activity. Agarose gels from one of three representative experiments show PKA-phosphorylated and nonphosphorylated PepTag A1 peptide. Lower panels. The amounts of phosphorylated and nonphosphorylated PepTag A1 peptides were densitometrically evaluated. PKA activity is expressed as the ratio of phosphorylated to nonphosphorylated PepTag A1 peptides. Values are mean ± SEM; ***P<0.001.

A potential influence of 4AD on total cellular PKA activity was assessed by measuring the kinase’s ability to phosphorylate a PKA-specific substrate peptide added to lysates derived from MCD4 cells (Figure 6C). In the absence or presence of forskolin, 4AD did not influence the phosphorylation of the peptide, indicating that 4AD does not globally affect PKA activity.

In IMCD cells, elevation of cAMP in response to forskolin or AVP increases the abundance of endogenous AQP2 approximately two-fold within 30 minutes.¹⁹ We had previously shown that this increase is detectable with antibodies directed against both the N and the C terminus of AQP2 and is thus independent from phosphorylations in the C terminus.¹⁹ The forskolin-induced increase in AQP2 abundance occurred similarly in the presence of 4AD (Figure 7A).

Figure 7.

4AD inhibits the forskolin (FSK)-induced phosphorylation of AQP2 at S256 but does not affect cAMP-regulated phosphorylations at S261 and S269. IMCD cells were left untreated or were treated with forskolin (10 µM, 20 minutes), 4AD (0.4 or 4 µM, 30 minutes), or the indicated combinations of forskolin and 4AD. Lysates were prepared, proteins were separated by SDS-PAGE, and AQP2 was detected using specific antibodies (C17, Santa Cruz) (A). AQP2 phosphorylated at S256 (pS256-AQP2) (B) was detected with specific custom-made antibodies.²⁵ pS261-AQP2 (C) and pS269 (D), and, as a loading control, α-tubulin were detected using commercially available antibodies. Representative blots from one of five or more independent experiments are shown. Upper panels. The signals emerging from the complex glycosylated form of AQP2 were quantified densitometrically. Statistically significant differences are indicated (mean ± SEM; *P<0.05; **P<0.01, ***P<0.001). compl, complex glycosylated APQ2; hm, high mannose form of AQP2; ng, nonglycosylated AQP2.

Surprisingly, although 4AD did not inhibit forskolin-induced increases in cAMP formation and cellular PKA activity, 4AD prevented the forskolin-induced cAMP/PKA-dependent AQP2 phosphorylation at S256, the key trigger for its redistribution to the plasma membrane (Figures 3 and 7B). In contrast, 4AD did not affect the forskolin-induced dephosphorylation at S261 or the forskolin-induced phosphorylation of S269 (Figure 7, C and D). Thus, 4AD uncouples the cAMP-induced S256 phosphorylation from cAMP-induced phosphorylation of S269 and dephosphorylation of S261 of AQP2.

Fluconazole Inhibits the cAMP-Induced Redistribution of AQP2

Another promising inhibitory compound from our primary screen is triazolpropenon (Table 1). However, it inhibited the AQP2 redistribution in concentrations close to its 50% lethal concentration (34 µM), and the less toxic derivatives (50% lethal concentration, 235 µM) that we synthesized lost their inhibitory potency (data not shown). Triazolpropenon possesses structural similarity to the fungistatic agent fluconazole. Fluconazole blocks fungal lanosterol 14α-demethylase³² and thereby inhibits ergosterol biosynthesis, causing membrane defects. In IMCD cells, fluconazole inhibits the forskolin-induced redistribution of AQP2 (Figure 8). This inhibitory effect is apparently not due to obvious membrane defects similar to those fluconazole causes in fungi. In humans, fluconazole is a substrate for elimination and an inhibitor of several cytochrome P450 enzymes (CYP2C9, CYP2C19, CYP3A4).³³ Thus, such P450 systems may be involved in the control of AQP2 or fluconazole modulates another yet unknown target controlling AQP2.

Figure 8.

Fluconazole inhibits the forskolin (FSK)-induced redistribution of AQP2 in primary IMCD cells. (A) Primary IMCD cells were left untreated, stimulated with forskolin (10 µM, 20 minutes), or treated with fluconazole (10 µM, 30 minutes) alone or in combination with forskolin. AQP2 was detected by immunofluorescence microscopy using specific primary antibodies (H27) and Cy3-coupled antirabbit secondary antibodies (red).⁵¹ Nuclei were stained with DAPI (blue). Shown are representative images from one of three independent experiments. (B) The intensities of intracellular and plasma membrane immunofluorescence signals arising from AQP2 were determined and related to nuclear signal intensities (n≥30 cells per condition; mean ±SEM; three independent experiments). The ratios of plasma membrane to intracellular fluorescence signal intensities were calculated. Ratios >1 indicate a predominant localization at the plasma membrane. ***P<0.001.

Discussion

We report here a novel, cell-based, high-throughput small-molecule screening approach as a feasible starting point for dissecting molecular mechanisms underlying AVP-mediated water reabsorption. The system permits monitoring effects of small molecules on AQP2 localization, and cell-damaging hits can easily be recognized and excluded from follow-up studies.

One of the small-molecule inhibitors of the AQP2 redistribution, which we identified, is 4AD, a blocker of V-ATPase.²⁶ In renal principal cells, 4AD interferes with acidification of intracellular vesicles (Figure 4) and leads to inhibition of the PKA-dependent phosphorylation of AQP2 at S256, the key trigger for its redistribution to the plasma membrane (Figure 7B). This effect cannot be explained by global inhibition of PKA because 4AD does not affect total cellular PKA activity (Figure 6C). However, 4AD may inhibit PKA locally. PKA is tethered to AQP2-bearing vesicles by A-kinase anchoring proteins (AKAPs), such as AKAP220 and AKAP18δ, and AKAP-PKA interactions are crucial for the cAMP-induced phosphorylation of S256 and the consequent redistribution of AQP2.^11,34–36 AKAP18δ is recruited to AQP2-bearing vesicles through ionic interaction with negatively charged membrane lipids and, in addition, through protein-protein interactions.³⁷ Ablation of the proton gradient across vesicular membranes by 4AD could interfere with the electrostatic interaction of AKAP18δ with AQP2-bearing vesicles and thereby with the local AKAP18δ-dependent PKA activation. Because AKAPs generally interact with membranes through protein-protein or electrostatic interactions or lipid modifications,³⁸ an altered proton gradient may also interfere with other AKAPs on AQP2-bearing vesicles. Notably, V-ATPase binds PKA.³⁹ It acts as a pH sensor and binds the small GTPase ARNO at low vesicular pH. Baf A1 prevents this interaction.⁴⁰ Similarly, the tethering of PKA by V-ATPase to AQP2-bearing vesicles may be pH-dependent and could be disrupted by 4AD. An altered proton gradient could also directly inhibit local PKA by affecting PKA’s “catalytically competent protonation state”⁴¹ between a pH of 6.5 and 8.⁴² Whether changes in protonation affect AQP2, in particular its accessibility for phosphorylation by PKA, is unknown.

AQP2-pS256 is slightly elevated (1.5-fold; Figure 7B) in the presence of forskolin and 4AD. 4AD (Figure 4) and baf A1³⁰ arrest AQP2 in the Golgi compartment. However, we have not detected PKA and Golgi markers in the same preparations of immunoisolated AQP2-bearing vesicles.^22,43 Thus, if PKA plays a role in phosphorylating S256 in the presence of 4AD and forskolin, it occurs outside the Golgi. In line, in IMCD cells AVP initiates activation of a perinuclear pool of PKA co-localizing with AQP2, apparently outside the Golgi, prior to mediating activation of PKA in other compartments.²² Alternatively, the S256 phosphorylation may be catalyzed by kinases other than PKA. One candidate is Golgi casein kinase. It is located within the Golgi and has been suggested to phosphorylate AQP2 in this compartment (Figure 5).⁴⁴ Other candidates are casein kinase II and protein kinase Cα, which phosphorylates S256 in vitro.^15,45

In IMCD cells, 4AD does not affect the phosphorylations at S261 and S269 (Figures 3 and 7), suggesting that the responsible kinases (e.g., p38 mitogen-activated protein kinases or Jun kinase¹⁹) are insensitive to 4AD or that they catalyze the phosphorylations independently from V-ATPase activity. Because 4AD inhibits only the S256 phosphorylation, our data corroborate the many earlier studies in providing further evidence that the S256 phosphorylation is the crucial trigger for the membrane insertion of AQP2.^7–9,15,25 Our data also show that the cAMP-dependent dephosphorylation at S261 and the phosphorylation at S269 are independent from the S256 phosphorylation by PKA. Tamma et al. found that the cAMP/PKA-dependent S261 phosphorylation also occurs in the AQP2-S256A mutant, which cannot be phosphorylated at S256, and that forskolin induces dephosphorylation of S261 in this mutant.¹⁸ However, our data seem in contrast to studies by Hoffert and coworkers, who had observed that cAMP-stimulated and PKA-dependent phosphorylation of S256 precedes and primes for phosphorylation of S269.¹⁵ The reason for the discrepancy is unclear. The residual phosphorylation of S256 in the presence of 4AD and forskolin (Figure 7B) may be sufficient to facilitate phosphorylation of S269. Alternatively, S269 phosphorylation may occur independently from S256 phosphorylation, and location may play a role: Because the combination of 4AD and forskolin maintains AQP2 mainly inside IMCD cells (Figure 5), S269 may be accessible for phosphorylation at a different location compared with the normal situation, in which forskolin stimulates its phosphorylation at the plasma membrane. This is underlined by immunofluorescence microscopic analysis (Figure 3). Under resting conditions, AQP2-pS269 is hardly detectable, whereas in the presence of forskolin alone, AQP2-pS269 is almost exclusively located at the plasma membrane, supporting the proposed inhibitory role of pS269 for endocytosis of AQP2.¹⁵ In the presence of both 4AD and forskolin, AQP2-pS269 is detectable mainly intracellularly; the observed arrest of AQP2 in Golgi compartments in the presence of 4AD (Figure 5) suggests that S269 phosphorylation could be catalyzed there.

So far, baf A1 is widely used to block V-ATPases in studies of protein and vesicle trafficking. However, baf A1 not only inhibits V-ATPase but also functions as an ionophore that causes influx of potassium into mitochondria. This can be lethal for various cells, including cancer and osteoclast-like cells,^46–48 but most importantly this function blurs interpretation of experimental data because intracellular ion homeostasis is disturbed. We are not aware of such effects of 4AD, suggesting that 4AD and other novel V-ATPase inhibitors, such as new derivatives of baf A1 that selectively inhibit V-ATPases,⁴⁹ represent alternatives to baf A1 as molecular tools for studying cellular trafficking processes.

4AD and structurally related compounds were suggested for treatment of bone resorption disorders (e.g., osteopetrosis) and various cancers.^26,49,50 However, because of the vital role of V-ATPases in normal physiology, adverse effects on blocking of V-ATPase may be expected. Hence, it is unclear whether such candidates reach the clinic. With regard to the work presented here, our identification of 4AD confirms the feasibility of our approach to discover new candidate molecules for further development into drugs for the treatment of diseases associated with dysregulated AVP-mediated water reabsorption (e.g., heart failure or syndrome of inappropriate antidiuretic hormone secretion).

With fluconazole, we discovered a small molecule targeting apparently a yet unknown player in the control of AQP2 (Figure 8). In addition, we identified three further inhibitors of the AQP2 redistribution, which we have not yet characterized but which will potentially identify additional novel proteins controlling AQP2 (compound IDs 212678, 212634 and 202336) (Table 1). Moreover, through screening further small-molecule libraries covering the chemical space more extensively, our approach has the potential for discovering further molecules and their targets controlling AQP2.

In summary, our screening lays the foundation for uncovering molecular mechanisms directing AQP2. In addition, identified molecules may constitute novel tools to study cellular trafficking processes and may pave the way to innovative treatments of diseases associated with disturbances of AVP-mediated water reabsorption.

Concise Methods

Materials and Reagents

AVP was synthesized by M. Beyermann, Leibniz-Institut für Molekulare Pharmakologie (Berlin, Germany), and forskolin was from BIOLOG Life Sciences Institute (Bremen, Germany). All other chemicals, if not stated otherwise, were purchased from Sigma-Aldrich (Munich, Germany).

Cell Culture

Primary rat IMCD cells were cultured as described previously.^19,22 Mouse collecting-duct cells stably expressing human AQP2 (MCD4 cell line) were grown in Dulbecco’s modified Eagle medium: Nutrient Mixture F-12 (DMEM/F-12) 1:1 with GlutaMAX (catalog number 31331–028; Gibco) supplemented with 5% FBS, 5 µM dexamethasone, 100 IU/ml penicillin, and 100 µg/ml streptomycin.²⁰

High-Throughput Screening

The ChemBioNet compound collection at the Leibniz-Institut für Molekulare Pharmakologie was purchased from ChemDiv (San Diego, CA). All library compounds are accessible at www.fmp-berlin.de, and further information on every compound is publically available at http://www.ncbi.nlm.nih.gov/pccompound. All dispensing and washing steps were performed using the EL406 washer/dispenser combination (BioTek Instruments, Bad Friedrichshall, Germany). The screen was performed in a 384-well plate format, with each plate containing 32 control wells (16 with forskolin alone and 16 with a mixture of forskolin and the PKA inhibitor, H89) (Supplemental Figure 2). MCD4 cells were seeded and after 24 hours incubated with each of the 17,700 library compounds (40 µM, 2 hours) or H89 (30 µM, 2 hours) and stimulated with forskolin (10 µM within the last 20 minutes). The cells were fixed with paraformaldehyde (2%), and AQP2 was detected using rabbit anti-AQP2 antibody H27⁵¹ and Cy5-coupled secondary antirabbit antibody (Dianova, Hamburg, Germany). Nuclei were stained with DAPI, and F-actin was labeled with TRITC-phalloidin.⁵² Fluorescence signals were visualized with an automated ArrayScan V^TI HCS reader microscope (Cellomics, ThermoFisher, Pittsburgh, PA) fitted with a 10× objective, in which the microscope automatically focused on the DAPI-stained nuclei. The localization of AQP2 was assessed by CytoCellMemTrans V3 application software (Thermo Fisher Scientific Inc., Waltham, MA). Automated object recognition of nucleus and plasma membrane was optimized on the basis of F-actin and DAPI staining. From these objects a cytosolic object mask with a variable distance from the nucleus was defined. AQP2 localization was semi-quantitatively expressed as the ratio of plasma membrane to intracellular fluorescence signal intensities, as described previously.^11,34,52 On the basis of the controls (forskolin alone or in combination with 4AD; see also Supplemental Figure 2) ratios <1.2 were defined as a predominant intracellular localization of AQP2. Acquired fluorescence signal data were processed with Pipeline Pilot software (Accelrys, Inc.; San Diego, CA), integrating the obtained information from this screen with compound details from the library, such as false-positive activity in previous screening assays.

Immunofluorescence Microscopy and Visualization of Acidic Intracellular Vesicles

MCD4 and IMCD cells were seeded on 12-mm cover slides. Immunofluorescence microscopy was carried out as described (LSM 710; Carl Zeiss MicroImaging, Jena, Germany).^19,22 In brief, AQP2 was detected employing primary anti-AQP2 H27 and Cy3-coupled secondary donkey antirabbit antibodies or with goat polyclonal anti-AQP2 (1:100; C-17, Santa Cruz Biotechnology, Inc.) and Cy2-coupled secondary donkey antigoat antibodies.^19,22 ERGIC-53/p58 and Golgi 58K protein were detected with polyclonal antirabbit (1:500; Sigma-Aldrich, Munich, Germany) and monoclonal mouse (1:100; Abcam; Cambridge, United Kingdom) antibodies, respectively, and Cy3-coupled polyclonal donkey antirabbit antibodies. AQP2-pS256 was detected with phospho-specific custom-made polyclonal rabbit antibody (Eurogentec Deutschland GmbH, Germany; 1:200 and 1:25, respectively) and antirabbit Cy3-coupled secondary antibodies.^19,25 AQP2-pS269 was detected with phospho-specific primary antibody (p112–269, Phosphosolutions, Aurora, CO) and antirabbit Cy3-coupled secondary antibodies. All secondary antibodies were purchased from Dianova (Hamburg, Germany) and used in a dilution of 1:600.

For co-localization studies of AQP2 with ERGIC-53/p58 and Golgi 58K, ZEN2010 software (Carl Zeiss, Jena, Germany) was used. Co-localizations were analyzed in defined regions of interest (i.e., the intracellular region between the nucleus and the plasma membrane).

LysoTracker RED DND-99 (Invitrogen) selectively accumulates in acidic intracellular vesicles. When pH levels shift to >6.5–7, LysoTracker RED DND-99 protonation and retention are impaired, indicated by decreased fluorescence signal intensity (Figure 4B). MCD4 cells were seeded on 30-mm cover slides and incubated with LysoTracker RED DND-99 (75 nM, 30 minutes, 37°C) and washed twice with Dulbecco PBS. Incubation with 4AD or baf A1 with Hoechst 33258 (5 µM, nuclei staining) was carried out simultaneously. Cover slides were transferred to a custom-made cuvette, and fluorescence signals were evaluated by confocal laser-scanning microscopy (LSM 510 META microscope, Carl Zeiss, Jena, Germany). Mean fluorescence signal intensities of LysoTracker RED DND-99 were measured in 10 fields per condition using ImageJ software.

Western Blotting, cAMP Measurements by Radioimmunoassay, Biotinylation of Cell Surface Proteins, and PKA Activity Assay

Western blotting was carried out as described elsewhere.^19,22 AQP2 was detected with goat polyclonal anti-AQP2 antibody C-17 (see above). AQP2-pS261 was detected with phospho-specific polyclonal phospho-S261 rabbit antibody (Abcam; Cambridge, UK). AQP2-pS256 was detected with phospho-specific custom-made polyclonal rabbit antibody (see above),^19,25 and AQP2-pS269 with a commercially available antibody (Phosphosolutions, see above). We detected α-tubulin with a CalBiochem antibody (CP06). Secondary antibodies, all coupled to horseradish peroxidase, were purchased from Dianova. Signals were visualized by utilizing chemiluminescence, i.e., incubation of the Western blot membranes with Lumi-Light Western blotting substrate solution (Roche Applied Science) and imaging at the Lumi Imager F1 (Roche Applied Science). Protein amounts were semi-quantified by densitometry (LumiAnalyst 3.0 software; Roche Diagnostics GmbH, Mannheim, Germany). Determination of cAMP levels by RIA, cell-surface biotinylation assays^19,22,53 and analysis of PKA activity using Promega’s PepTag A1 peptide assay kit⁵⁴ were carried out as described.

Statistical Analyses

Statistical analyses of data were carried out with GraphPad Prism 5.0 software using a one-way ANOVA test combined with a Bonferroni post hoc comparison test; P<0.05 was considered to represent statistically significant differences. All values are expressed as mean ± SEM.

Disclosures

None.