Phosphorylation and ubiquitylation are opposing processes that regulate endocytosis of the water channel aquaporin-2

Phosphorylation and ubiquitylation are highly dynamic and regulated protein post-translational modifications (PTMs). Both modifications have the potential to mediate inducible protein–protein interactions and thereby regulate protein function. For membrane proteins, induction of phosphorylation, including poly-phosphorylation, is a mechanism for regulating the activity and intracellular trafficking of channels and transporters (Moeller et al., 2010; Rosenbaek et al., 2012; Tong et al., 2009; Yang et al., 2006). Ubiquitylation of membrane proteins through specific linkage is a common way of mediating internalization and promotion of protein degradation by the lysosomal pathway (Kazazic et al., 2009; Piper and Luzio, 2007; Vina-Vilaseca et al., 2011). ‘Cross-talking’ of phosphorylation and ubiquitylation has been proposed (Hunter, 2007), and phosphorylation often occurs as a priming event for ubiquitylation (Hunter, 2007; Murakami et al., 2009; Orford et al., 1997; Sehat et al., 2007). This occurs through a variety of mechanisms, such as phosphorylation-mediated regulation of the activity of E3-ligases, by generation of a phospho-E3-recognition (phosphodegron) site or by promoting localization of the protein to intracellular compartments where ubiquitylation can occur (Hunter, 2007). Although less frequent, and to our knowledge not reported for plasma membrane proteins, phosphorylation can mediate the opposite effect, for example, the phosphorylation of the kinase c-MOS inhibits ubiquitylation (Sheng et al., 2002). Alternatively, ubiquitylation can regulate protein phosphorylation by, for example, regulation of kinase activity (Hunter, 2007). Multiple PTMs in a protein are likely to act in concert rather than mediate regulatory effects alone. Such mechanisms could greatly increase fine tuning of protein function but also provide a great challenge in deciphering the complexity of protein regulation.

In this study, we examined the potential interplay between polyubiquitylation and poly-phosphorylation of the water channel aquaporin-2 (AQP2). AQP2 is expressed in the kidney collecting duct. Vasopressin-mediated accumulation of AQP2 in the apical plasma membrane of kidney principal cells is an essential event for urine concentration. The mechanism underlying membrane accumulation of AQP2 involves regulated insertion and removal of AQP2 from the plasma membrane (Brown, 2003; Knepper and Nielsen, 1993), a process that involves phosphorylation and ubiquitylation (Moeller et al., 2011). The C-terminal 15 amino acids of AQP2 contain a vasopressin-regulated poly-phosphorylated region with four phosphorylated serine residues at S256, S261, S264 and S269 (T in human) (Hoffert et al., 2006). These sites have differing roles for regulation of channel function and localization (see, for example, Fushimi et al., 1997; Hoffert et al., 2008; Kamsteeg et al., 2000; Moeller et al., 2011; Moeller et al., 2010; Nedvetsky et al., 2010; Tamma et al., 2011). It has been proposed that phosphorylation of AQP2 results in increased accumulation of AQP2 on the plasma membrane owing to reduced internalization (Lu et al., 2007; Moeller et al., 2010). The C-terminus of AQP2 is also subjected to polyubiquitylation at K270, with two or three ubiquitin molecules added through K63-linkage (Kamsteeg et al., 2006). Polyubiquitylation of AQP2 enhances its removal from the plasma membrane (Kamsteeg et al., 2006), similar to what has been reported for other proteins polyubiquitylated through K63-linkage (Haglund and Dikic, 2005).

Owing to the close proximity of AQP2 phosphorylation and ubiquitylation and the opposing roles of these PTMs for determining AQP2 localization, we utilized AQP2 as a model protein to examine the interplay between phosphorylation and ubiquitylation and whether one of these PTMs is always dominant to the other with respect to protein localization. We determined that although AQP2 S269 phosphorylation and K270 ubiquitylation can occur in parallel, the effects of phosphorylation can override the effect of K63-linked ubiquitylation on AQP2 endocytosis. Similar mechanisms might exist for other plasma membrane proteins.

When the AQP2 mutants AQP2-S256D and AQP2-S269D are expressed in MDCK cells, they localize predominantly to the plasma membrane in basal conditions (Hoffert et al., 2008; Moeller et al., 2010; van Balkom et al., 2002). We and others (Rice et al., 2012) have also shown a decreased rate of AQP2 internalization in cells expressing single mutations at AQP2-S256D and AQP2-S269D compared to those expressing wild-type AQP2 (AQP2-WT); indicating that phosphorylation at both positions might be involved in retaining AQP2 in the plasma membrane. To dissect out the roles of individual sites we generated various polarized cell lines expressing double mutant AQP2 to examine whether AQP2-pS256 or AQP2-pS269 dominates in respect to AQP2 cellular localization. In cells expressing AQP2-S256A-S269A or AQP2-S256A-S269D, AQP2 was predominantly localized within the cell in basal conditions (Fig. 1), with greater colocalization with markers of the trans-Golgi network (TGN) (Vti1b and adaptin-G) and endosomes (EEA1) compared to WT-AQP2 (Fig. 2A and data not shown). In contrast, AQP2-S256D-S269A and AQP2-S256D-S269D showed predominantly membrane localization (Fig. 1), emphasized by very little colocalization with Vti1b or EEA1 (Fig. 2A). AQP2-S256D-S269D had significantly reduced colocalization with Vti1b and EEA1 compared to AQP2-S256D-S269A (Fig. 1; Fig. 2A). These results demonstrate that phosphorylation of AQP2 S256 is essential for AQP2 membrane localization, but the S256D-S269D form can accumulate on the plasma membrane more than the S256D-S269A form of AQP2. To confirm these results, we used a biotin-based internalization assay to examine AQP2 internalization rates in AQP2-S256D-S269A and AQP2-S256D-S269D double mutant cell lines. The AQP2-S256D mutation resulted in a decreased rate of AQP2 internalization compared to AQP2-WT (Fig. 2B). An additional decrease in internalization was observed when the AQP2-S269 residue was mutated to aspartic acid, suggesting that AQP2-S269-phosphorylation serves to further decrease AQP2 internalization when AQP2-S256 is phosphorylated.

Despite multiple attempts by our group (data not shown) and others (for example, Hoffert et al., 2008), a specific kinase that can modulate AQP2-S269 phosphorylation in vivo has not been identified. Thus, we made use of phosphatase inhibitors in an attempt to increase AQP2 phosphorylation specifically at individual sites and examine the effects of this inhibition on AQP2 plasma membrane accumulation. In the presence of the cell permeable phosphatase inhibitors okadaic acid (200 nM) and calyculin A (100 nM), following forskolin (an activator of adenylate cyclase resulting in increased AQP2 membrane localization) washout, AQP2 had a significantly decreased rate of internalization compared to controls (Fig. 3A). Concurrently with these effects on internalization, under the same conditions AQP2-S269 phosphorylation levels were significantly prolonged and enhanced (Fig. 3B), which supports a role for pS269 for decreasing AQP2 internalization. In contrast, there were no significant differences between control and treated groups when analyzing AQP2-pS256 levels, with levels of pS256-AQP2 remaining relatively constant throughout the internalization time period examined (Fig. 3C; supplementary material Fig. S3A). Although not significant from untreated controls, at the earliest time point examined during the washout period, there was a trend for increased AQP2-pS256 levels with okadaic acid and calyculin A, suggesting that AQP2-pS256 might be important for modulating AQP2 internalization at the early stages of endocytosis. Although there was a trend towards greater membrane abundance of AQP2 following forskolin treatment in the presence of inhibitors, the difference was not significant (Fig. 3D). Okadaic acid and calyculin A had no significant effect on AQP2-pS256 levels (Fig. 3E). AQP2-S269 phosphorylation was absent under control or upon okadaic acid or calyculin A treatment alone, significantly increased after forskolin treatment, and increased to a greater degree in the presence of phosphatase inhibitors (Fig. 3F). Taken together, these observations indicate that the effects of the phosphatase inhibitors on AQP2 internalization occur predominantly through modulation of AQP2-S269 phosphorylation. Additionally, AQP2-S269 phosphorylation was only detected during forskolin treatment and AQP2-pS269 was not observed in the internalized fraction (supplementary material Fig. S1A), suggesting rapid dephosphorylation of AQP2 at S269 during the internalization process.

The observed effects of okadaic acid and calyculin A on the phosphorylation status of AQP2 at S269 could be direct or indirect. At the concentrations utilized, the most likely phosphatases inhibited were PP1 or PP2A. AQP2-pS269 peptides were dephosphorylated by both PP1 and PP2A in vitro (Fig. 3G), indicating that a direct effect in our cell studies was plausible. However, in three independent experiments fostriecin, endothall or cantharidic acid (inhibitors of PP2A) did not significantly alter AQP2-S269 phosphorylation (supplementary material Fig. S1). Deltametrin, an inhibitor of PP2B, also did not affect AQP2-S269 phosphorylation (supplementary material Fig. S1), but mildly increased pS256 levels. These data indicate that PP1 is the likely phosphatase responsible for dephosphorylating AQP2-pS269 in vivo.

Short-chain ubiquitylation of AQP2-K270 can occur in the membrane and enhance AQP2 internalization (Kamsteeg et al., 2006). To examine potential cross-talk between K270 ubiquitylation and AQP2-S269 phosphorylation, we generated cell lines expressing N-terminal FLAG-tagged AQP2-WT, and AQP2-K270R and AQP2-S269A and -S269D mutants, and assessed the AQP2 ubiquitylation levels under control, forskolin, forskolin washout and forskolin plus TPA conditions. TPA is an activator of PKC that enhances apical endocytosis in MDCK cells (Holm et al., 1995) and enhances AQP2 ubiquitylation in the presence of forskolin (Kamsteeg et al., 2006). In MDCK cells, AQP2 was ubiquitylated at K270 with one to three moieties (Fig. 4A; supplementary material Fig. S2). Ubiquitylation of AQP2-WT increased upon forskolin washout and forskolin plus TPA treatment (Fig. 4A,B). Examination of the AQP2-S269A and AQP2-S269D mutants revealed similar ubiquitylation patterns (Fig. 4B). Quantitative assessment from at least three independent experiments is shown in Fig. 4C, demonstrating no significant differences in AQP2 ubiquitylation between AQP2-WT, AQP2-S269A and AQP2-S269D. These data demonstrate that the levels of K270 ubiquitylation are not directly affected by mimicking phosphorylation at S269. The use of linkage-specific ubiquitin antibodies on immunoprecipitated AQP2 samples demonstrated that AQP2 polyubiquitylation in MDCK cells occurs through K63 linkages (supplementary material Fig. S2A), confirming previous mutational studies (Kamsteeg et al., 2006). Furthermore, in mpkCCD_c14 cells, a mouse collecting duct cell line with endogenous expression of the type II vasopressin receptor (V2R) (Yu et al., 2009), a similar pattern of AQP2 ubiquitylation was observed, with the greatest ubiquitylation levels occurring following vasopressin removal (supplementary material Fig. S2B). This data indicates that AQP2 ubiquitylation and regulation is not specific to a cell line type.

Analysis of the ubiquitylation and phosphorylation datasets at PhosphoSite demonstrated that there are 95 other reported peptides (rat, mouse, human) that contain an ubiquitylated lysine residue at the penultimate position of the protein, similar to K270 of AQP2 (http://interpretdb.au.dk/database/Penultimate_Ubi.htm). Alignment of these sequences highlighted a preference for S, T or Y residues at the −1 position (relative to the K), and S and T residues at the −2 and −3 positions (supplementary material Fig. S3). These data indicate that phosphorylation and ubiquitylation often occur in close association with each other at the terminal tail of proteins and point towards a general PTM regulatory crosstalk.

In AQP2-WT cells, phosphorylation at AQP2-S269 significantly increased following forskolin treatment and decreased after wash-out. A similar pattern of phosphorylation was observed for cells expressing AQP2-K270R (Fig. 5A). Quantification from three independent experiments demonstrated that forskolin-induced AQP2-S269 phosphorylation in AQP2-K270R mutant cells was significantly higher than for AQP2-WT, but returned to similar baseline levels after washout (Fig. 5C). Similarly, although not as pronounced, AQP2-S256 phosphorylation levels were also significantly higher in AQP2-K270R mutant cells after forskolin stimulation (Fig. 5B).

WT-AQP2 cells were subjected to forskolin or forskolin washout in the presence or absence of the phosphatase inhibitors okadaic acid and calyculin A (Fig. 6A). AQP2 ubiquitylation was greater under all conditions in the presence of phosphatase inhibitors, although this was only significant under control or forskolin conditions (Fig. 6B). AQP2-pS256 levels were not significantly higher under conditions where AQP2 ubiquitylation was increased (Fig. 6C). AQP2-S269 phosphorylation was only detected after forskolin treatment and was significantly higher in the presence of phosphatase inhibitors (Fig. 6D). Thus, phosphatase inhibitor treatment reduces AQP2 internalization (Fig. 3), and induces both ubiquitylation and phosphorylation of AQP2 at S269, two PTMs with opposing effects on membrane localization of AQP2.

Assessment of AQP2 ubiquitylation versus membrane abundance and internalization rate was performed in a single study using AQP2-S256D-S269D cells, which had the lowest rate of AQP2 internalization (Fig. 1; Fig. 2A). Supporting our confocal imaging studies, AQP2-S256D-S269D was predominantly detected in the plasma membrane with no significant increase following forskolin stimulation (Fig. 7A). Ubiquitylation of AQP2-WT significantly increased following 15 min forskolin washout plus TPA treatment (Fig. 7B). In contrast, despite its greater plasma membrane abundance, AQP2-S256D-S269D had significantly greater ubiquitylation than AQP2-WT under control conditions and the levels of ubiquitylation were not increased by forskolin washout plus TPA treatment (Fig. 7B). In addition, despite the higher ubiquitylation levels, AQP2-S256D-S269D had a significantly lower degree of internalization from the apical plasma membrane (Fig. 7C). These data support a model whereby site-specific AQP2 phosphorylation can override the effect of K63-linked ubiquitylation on AQP2 endocytosis.

To examine whether AQP2 ubiquitylation could be regulated by vasopressin in vivo, AQP2 ubiquitylation levels were examined in kidney homogenates isolated from rats at various time points following treatment with the selective vasopressin type 2 receptor (AVPR2) agonist dDAVP (Fig. 8). A similar pattern of AQP2 ubiquitylation was observed in vivo as in our cell models. Supporting our cell model mechanisms of AQP2 retention mediated by phosphorylation at S269, maximum AQP2 ubiquitylation occurred 30 min after vasopressin stimulation, a time point when maximal levels of pS269-AQP2 were observed and AQP2 levels on the apical plasma membrane of collecting duct principal cells are the greatest (Moeller et al., 2009a).

Phosphorylation and ubiquitylation have been well studied for their roles in regulating protein function, yet crosstalk between the PTMs for regulation of plasma membrane proteins is not well defined. AQP2 undergoes both PTMs within a short span of C-terminal amino acids and thus is an ideal model to assess the interplay between phosphorylation and ubiquitylation. Previous studies have highlighted the role of AQP2-S256 phosphorylation for plasma membrane targeting of AQP2 (Fushimi et al., 1997; Hoffert et al., 2008; Katsura et al., 1997; McDill et al., 2006) and have suggested a special role for vasopressin-induced phosphorylation of AQP2 at S269 in plasma membrane accumulation (Hoffert et al., 2008; Moeller et al., 2009a; Moeller et al., 2009b; Moeller et al., 2010). Owing to the close association and interdependence of these two phosphorylation sites, it has proven to be difficult to understand the role of each individual site in the regulation of AQP2. In this study, we aimed to address this problem and, furthermore, suggest the possibility of a general cell biological phenomenon where site-specific phosphorylation of other plasma membrane proteins might override the endocytic effect of K63-linked ubiquitylation.

Both S256 and at S269 play roles in AQP2 membrane localization. When AQP2-S256 was mutated to aspartic acid, AQP2 was predominantly in the plasma membrane irrespectively of the status at the AQP2-S269 position. In contrast, preventing phosphorylation at residue AQP2-S256 resulted in intracellular localization of the protein regardless of the AQP2-S269 status. Examination of AQP2 internalization rates following agonist removal demonstrated that AQP2-S256D-S269A or AQP2-S256D-S269D mutants had slower internalization compared to AQP2-WT, with the AQP2-S256D-S269D form having the most prolonged period at the apical membrane. Combined with the reduced colocalization with markers of the endocytic system, these results strengthen the previously suggested role of the S269 site in modulating and/or prolonging AQP2 localization in the plasma membrane.

To investigate the isolated effect of AQP2-S269 phosphorylation in AQP2 membrane retention, we attempted to specifically inhibit de-phosphorylation of AQP2 at S256 or S269. In vitro de-phosphorylation assays demonstrated that PP1 and PP2A could dephosphorylate AQP2 at S256 or S269 in vitro. Calyculin A and okadaic acid in cell culture had no significant effect on pS256 levels following forskolin stimulation, but did greatly increase pS269 levels; this suggests that both PP1 and PP2A could play a role in AQP2 de-phosphorylation at S269. Selectively inhibiting PP2A or PP2B using endothall, fostriecin, cantharidic acid or deltamethrin did not result in significant changes in pS269 levels, suggesting that PP1 is the most likely protein phosphatase targeting pS269-AQP2 in vivo. Our results are in line with other studies that suggest a role for PP1 in regulation of AQP2 phosphorylation (Moeller et al., 2010; Zwang et al., 2009). A functional role for protein phosphatases in regulation of AQP2 has previously been suggested, although the effects of okadaic acid observed were attributed to a depolymerizing effect (disruption) on the actin cytoskeleton (Valenti et al., 2000).

Our studies reported here strengthen the hypothesis that vasopressin-induced AQP2-S269 phosphorylation mediates membrane retention of AQP2. In contrast, vasopressin removal (mimicked by forskolin washout) results in increased AQP2-K270 K63-linked ubiquitylation in the plasma membrane, which mediates enhanced AQP2 internalization (Kamsteeg et al., 2006). Thus, these neighboring sites have opposite effects on AQP2 subcellular localization. Owing to their close proximity we hypothesized that the two PTMs were linked, with enhanced AQP2-S269 phosphorylation reducing the levels of K270 ubiquitylation. However, relative to WT-AQP2, mutating AQP2-S269 to either alanine or aspartic acid had no significant effect on AQP2 ubiquitylation levels. Thus, it does not appear that AQP2-S269 phosphorylation directly regulates ubiquitylation. An alternative cross talk mechanism is that ubiquitylation regulates phosphorylation of AQP2. Mutation of AQP2-K270 to an arginine (preventing ubiquitylation) did not prevent forskolin-induced AQP2 phosphorylation but, in fact, facilitated AQP2-S269 phosphorylation. A possible explanation for this is that the presence of ubiquitin affects the binding or activity of the kinases involved in AQP2 phosphorylation (Hunter, 2007). Bioinformatics demonstrated that when a ubiquitylated lysine residue lies within the terminal three amino acids of a protein, there was a preference for S, T or Y residues at the −1 position (relative to the K). These data indicate that in other proteins, phosphorylation and ubiquitylation can occur in close association, suggesting the potential for a general PTM regulatory crosstalk mechanism.

As PTM of AQP2 at S269 and K270 can occur in parallel, we attempted to determine whether one of the modifications was dominant in terms of AQP2 membrane localization. Our data support a model whereby phosphorylation of AQP2 at S269 is able to override the internalization signal of neighboring K63-linked polyubiquitylation. This is supported by the following major findings. First, following forskolin stimulation and subsequent washout, in the presence of okadaic acid and calyculin A, there was a general increase in AQP2 ubiquitylation. However, this occurred with a paradoxical decrease in AQP2 internalization combined with prolonged and increased phosphorylation of S269. Second, AQP2-S269D cells had reduced AQP2 internalization rates despite the levels of AQP2 ubiquitylation being higher. Third, in AQP2-S256D-S269D cells AQP2 was predominantly on the plasma membrane under control conditions, despite 2-fold–3-fold greater levels of AQP2 ubiquitylation. Fourth, in AQP2-S256D-S269D cells AQP2 had significantly reduced internalization rates despite constitutively high levels of AQP2 ubiquitylation. Finally, in vivo, at a time point when AQP2 plasma membrane levels are greatest, AQP2 ubiquitylation and pS269 phosphorylation levels are also greatest. To our knowledge, this is the first report that phosphorylation can sequester a protein in the plasma membrane despite the protein being marked for internalization by K63-linked ubiquitylation. What is the basis for such a mechanism (summarized in supplementary material Fig. S4)? The first possibility is centered on previous observations (Lu et al., 2007; Moeller et al., 2009b) that phosphorylation of AQP2 decreases its interaction with proteins of the ‘endocytic machinery’. It is feasible that, even when marked for internalization by K63-linked ubiquitylation, phosphorylation of AQP2 sequesters the protein in a membrane compartment that is not susceptible to clathrin-mediated endocytosis, the major pathway for AQP2 internalization (Brown and Orci, 1983; Brown et al., 1988; Sun et al., 2002). Such AQP2 ‘endocytic-resistant domains’ have been proposed earlier by Brown and colleagues (Bouley et al., 2006). Alternatively, phosphorylation of AQP2 might alter the plasma membrane dynamics of AQP2 between raft and non-raft compartments, and indirectly affect the rate at which ubiquitylated AQP2 can be processed for clathrin-mediated endocytosis. The final mechanism that we propose is centered on the molecular recognition of the ubiquitylated AQP2 by ubiquitin-binding domain (UBD)-containing proteins, e.g. epsin (Horvath et al., 2007; Hurley and Wendland, 2002). It is plausible that phosphorylation of AQP2 alters the affinity of UBD proteins for AQP2, thus affecting clathrin pit formation and endocytosis. Furthermore, epsins have low affinity to single ubiquitin moieties and multiple ubiquitin moieties are required to form an efficient internalization signal (Piper and Luzio, 2007). As we cannot determine how many ubiquitin moieties are present within the AQP2 tetramer formed on the surface, it is possible that the stoichiometry of phosphorylation versus ubiquitylation within each tetramer in the plasma membrane might be a factor that determines the internalization rate of AQP2.

In conclusion, our data suggest that site-specific phosphorylation of AQP2 can override the effect of K63-linked ubiquitylation on AQP2 endocytosis. We propose that the effects of closely associated phosphorylation and ubiquitylation that we observe for AQP2 could be a general mechanism for other plasma membrane proteins and act as a mechanism to fine-tune protein localization and function.

Mutant forms of AQP2 and N-terminal FLAG-tagged AQP2 were generated by site-directed mutagenesis. Generation of stable MDCK cell lines and cell culture conditions were as previously described (Hoffert et al., 2008). Multiple individual cell lines were individually characterized by examination of cell morphology, and AQP2 expression by western blotting, immunocytochemistry and RT-PCR. For all experiments, cells were cultured on semi-permeable supports (0.4 µM pore size, Corning) until a confluent monolayer formed. mpkCCDc14 cells were cultured on filters as described previously (Yu et al., 2009) until the trans-epithelial resistance (TER) was above 5 kOhm/cm². Subsequently dDAVP (Sigma, 10⁻⁹ M) was added in serum-free medium to the basolateral compartment for 4 days to induce AQP2 expression.

Affinity-purified rabbit phospho-specific antibodies against pS256-AQP2 and pS269-AQP2 or against total AQP2 upstream of phosphorylation sites have previously been characterized (Hoffert et al., 2008; Moeller et al., 2009a; Nishimoto et al., 1999). Mouse anti-ubiquitin (P4D1) was from Cell Signaling and rabbit anti-FLAG antibody (F7425) was from Sigma-Aldrich. The phosphatase inhibitors cantharidic acid, okadaic acid, endothall, deltamethrin, fostreicin and calyculin A were from Calbiochem. Forskolin (Sigma) was used at a final concentration of 25 µM. Intracellular marker antibodies were purchased from BD Transduction Laboratories and used for immunocytochemistry at: mouse-anti EEA1, 1∶100 (early endosome marker), Vti1b, 1∶500 (marker of post-Golgi vesicle trafficking), and adaptin G, 1∶500 (marker of late Golgi and the trans-Golgi network and/or endosomes).

Standard procedures were utilized and blots were developed using ECL detection. Quantification was performed on non-saturated films by determining signal intensity in each band using Quantity One 4.2.3 software densitometry analysis.

This was performed as described in detail previously (Moeller et al., 2010).

A Leica TCS SL confocal microscope with an HCX PL APO 63× oil objective lens (numerical aperture 1.40) was used for obtaining image stacks with a z-distance of 0.1 µm between images. Images were obtained at room temperature. Microscope settings (PMT offset and gain, sampling period, and averaging) were identical. To quantify the degree of colocalization, a minimum of four independent stacks from two different experiments were obtained sequentially by using a 488-nm laser line and emission between 505 and 540 nm for Alexa Fluor 488 and a 546-nm laser line and emission over 585 nm for Alexa Fluor 546 and 555. Background correction and quantification of colocalization was performed using the colocalization function of Imaris image analysis software.

Phosphopeptides corresponding to the C-terminus of AQP2 phosphorylated at either S256 or S269 were subject to in vitro dephosphorylation (peptide sequence: Biotin-LC-Val-Arg-Arg-Arg-Gln-Ser-Val-Glu-Leu-His-Ser-Pro-Gln-Ser-Leu-Pro-Arg-Gly-Ser-Lys-Ala-OH, underlined residues correspond to S256 and S269). In a 30 µl reaction, 1.5 nmol phosphopeptide was incubated at 30 min for 30°C with 50 mM HEPES, 100 mM NaCl, 2 mM DTT, 0.01% Brij 35, pH 7.5, 1 mM MnCl in the presence or absence of 2.5 or 7.5 units PP1 (P0754, NEB) or 0.1 or 0.5 units PP2A (14-111, Millipore). Reactions were stopped using Laemmli sample buffer containing DTT and heating for 10 min at 65°C.

Cells were washed twice in pure medium and pre-incubated for 30 mins in the presence of inhibitors or vehicle. In the presence of inhibitors or vehicle, cells were stimulated with forskolin (25 µM) for 20 min at 37°C. After three washes on ice in ice-cold PBS-CM, pH 8.0 (PBS containing 1 mM CaCl and 0.1 mM MgCl₂), cells were incubated for various time points in medium containing vehicle or inhibitors. For some studies, gel samples were prepared directly after two washes in ice-cold PBS-CM by addition of sample buffer containing DTT. For internalization studies, the cells were surface biotinylated after the forskolin treatment followed by re-incubation at 37°C in medium containing inhibitors or vehicle for various timepoints. MesNa-based stripping of surface biotin and internailization assays were performed as previously described (Moeller et al., 2010). All experiments were performed on two wells of cells at least three times.

For immunoprecipitations using phosphatase inhibitors (okadaic acid and calyculin A) cells were incubated in control or forskolin-containing medium for 30 min in the presence or absence of inhibitors. For studies indicating ‘washout’, forskolin was removed, and cells were washed and re-incubated for 20 min in presence or absence of inhibitors. For other studies, cells were pretreated as annotated for 30 min followed by ‘washout’ in either control medium, or medium containing forskolin (25 µM) or forskolin and 0.1 µM 12-O-tetradecanoylphorbol 13-acetate (TPA) for 15 mins. Samples were lysed in 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.25% Na-Deoxycholate, 1% Triton X-100, 1 mM EDTA, 20 mM N-ethylmelamide, containing protease inhibitors Leupeptin and Phefa-block (Boehringer Mannheim) and phosphatase inhibitor cocktail tablets (PhosSTOP, Roche Diagnostics A/S). Following sonication, samples were centrifuged at 10,000 g for 10 min at 4°C. Samples were assayed for protein concentration and immunoprecipitation was performed at 4°C for 1 h using ∼150 µg of lysate and 1 µg of FLAG antibody in a total volume of 500 µl. Lysates were subsequently incubated with 20 µls of Protein-A–agarose (Santa Cruz Biotechnology) followed by washing three times with lysis buffer and elution in sample buffer. Specificity of AQP2 ubiquitylation was confirmed by performing immunoprecipitations following full denaturation of cell lysates in SDS-buffer (0.5% SDS, 150 mM NaCl, 10 mM Tris-HCl, pH 8.0) before heating to 65°C for 10 min (data not shown).

Basolateral dDAVP was washed out using pure DMEM/F12 medium for 2 h before the experiment. Where required, cells were stimulated using dDAVP (1 nM) for 20 min or stimulated for 20 min followed by an additional 20-min washout. Cells were lysed in a buffer containing 50 mM Tris-HCl pH 7.2, 1 mM MgCl, 0.2 uM CaCl, 100 mM KCl, 0.5% Triton, 0.25% NaDeoxycholate, 0.01% SDS and 0.5 mg/ml DTT before immunoprecipitation as above.

All animal protocols comply with the European Community guidelines for the use of experimental animals and were performed in accordance to licenses for the use of experimental animals issued by the Danish Ministry of Justice. Before each experiment, rats had free access to standard rat chow and water. 16 male wistar rats (200 g) were randomly separated into experimental or control groups. Rats were injected intramuscularly with 10 ng dDAVP in saline (or saline alone) and killed after 30 or 120 min by cervical dislocation. Kidneys were homogenized in 5% sorbitol, 5 mM histidine-imidazole, and 0.5 mM Na₂EDTA containing phosphatase and protease inhibitors and 20 mM N-ethylmelamide. Homogenates were mixed 1∶1 with 2× lysis buffer 100 mM Tris-HCl pH 7.4, 300 mM NaCl, 0.5% Na-Deoxycholate, 2% Triton X-100, 2 mM EDTA, 20 mM N-ethylmelamide, containing protease inhibitors Leupeptin and Phefa-block (Boehringer Mannheim) and phosphatase inhibitor cocktail tablets (PhosSTOP, Roche Diagnostics A/S). IP was performed as for MDCK cells.

Data was tested for normal distribution using the D'Agostino-Pearson omnibus test and Graphpad Prism Software. Data fitting a normal distribution were analyzed using one-way ANOVA followed by Bonferroni's Multiple Comparison Test. Multiple comparisons tests were applied only when a significant difference (P<0.05) was determined in the ANOVA. Data not fitting a normal distribution were assessed using a nonparametric Kruskal–Wallis test. All data are presented as mean±s.e.m.

We thank Mark Knepper (Epithelial Systems Biology Laboratory, NHLBI, NIH, USA) for the K5006 AQP2 antibody and critical reading of the manuscript, and Christian Westberg, Tina Drejer and Helle Høyer (Dept. Biomedicine, Aarhus University) for technical assistance.