Regulation of activity and localization of the WNK1 protein kinase by hyperosmotic stress

WNK1 (with-no-K[Lys] kinase-1) was originally identified as an unusual kinase that lacked an invariant catalytic Lys residue in subdomain II of the catalytic domain that is crucial for binding of ATP (Xu et al., 2000). WNK1 is a catalytically active kinase, and modeling (Xu et al., 2000), as well as structural analysis, of the WNK1 catalytic domain (Min et al., 2004) revealed that a Lys residue in subdomain I substitutes for the missing Lys residue in subdomain II. WNK1 is a widely expressed protein kinase comprising 2,382 residues. It possesses a kinase catalytic domain at its N terminus (residues 221–479), and apart from three putative coiled-coil domains, the remainder of the WNK1 polypeptides possess no obvious structural features (Verissimo and Jordan, 2001; Xu et al., 2005). Great interest in WNK1 was aroused after the finding that intronic deletions that increased WNK1 expression were observed in humans with an inherited hypertension and hyperkalemia (elevated plasma K⁺) disorder termed Gordon's syndrome or pseudohypoaldosteronism type II (OMIM 145260; Wilson et al., 2001). These findings indicated that overexpression of WNK1 may result in hypertension and, consistent with this, heterozygous WNK1^−/+ mice possess reduced blood pressure (Zambrowicz et al., 2003). WNK1-knockout embryos fail to develop, indicating that WNK1 is also required for normal development. There are four isoforms of WNK (WNK1, -2, -3, and -4) in humans encoded by distinct genes (Verissimo and Jordan, 2001). Mutations in WNK4 have also been found in patients with Gordon's syndrome, but in contrast to WNK1, these comprise point mutations lying within noncatalytic regions of this enzyme (Wilson et al., 2001). It is not yet clear how mutations in WNK4 lead to Gordon's syndrome, but overexpression of a Gordon's syndrome mutant of WNK4, but not the wild-type enzyme, increased blood pressure in mice (Lalioti et al., 2006).

Most functional studies on WNK isoforms have focused on the overexpression of these enzymes in Xenopus laevis oocytes or epithelial cells and monitoring the effects that this has on the activity and membrane localization of coexpressed ion cotransporters or ion channels. These have thus far revealed that WNK isoforms have effects on the activity and/or membrane expression of the thiazide-sensitive Na⁺:Cl⁻ cotransporter (NCC), the bumetanide-sensitive Na⁺:K⁺:2Cl⁻ cotransporter-1/2 (NKCC1/2), the K⁺:Cl⁻ cotransporter-2, the Cl⁻:HCO3⁻ exchanger, the inwardly rectifying K⁺ channel, the epithelial Na⁺ channel, the tight junction claudin proteins, and the transient receptor potential vanilloid-4 Ca²⁺ channel (for reviews see Delaloy et al., 2005; Kahle et al., 2005; Gamba, 2006).

Recent findings indicate that the protein kinases WNK1 and -4 interact with high affinity with the protein kinases STE20/SPS1-related proline alanine–rich kinase (SPAK) and the oxidative stress response kinase-1 (OSR1; Piechotta et al., 2003; Vitari et al., 2005; Gagnon et al., 2006). These observations were followed by the finding that WNK1 and -4 could phosphorylate and activate SPAK and OSR1 in vitro (Moriguchi et al., 2005; Vitari et al., 2005; Anselmo et al., 2006). SPAK and OSR1 are phosphorylated by WNK1/WNK4 at a Thr residue located within the T-loop (Thr233-SPAK and Thr185-OSR1) as well as at a conserved noncatalytic Ser residue (Ser373-SPAK and Ser325-OSR1) lying within a region termed the S-motif (Vitari et al., 2005). Mutational analysis indicated that phosphorylation of the T-loop rather than the S-motif was required for the activation of SPAK and OSR1 by WNK1 (Vitari et al., 2005). SPAK and OSR1 were originally identified through their ability to interact, phosphorylate, and activate NKCC1 (Piechotta et al., 2002; Dowd and Forbush, 2003) and may also regulate NCC (Pacheco-Alvarez et al., 2006). SPAK and OSR1 are 68% identical in sequence and possess a highly similar kinase catalytic domain as well as a conserved C-terminal (CCT) domain, which interacts with RFXV/I motifs present in both WNK isoforms as well as NKCC1 (Piechotta et al., 2002; Moriguchi et al., 2005; Gagnon et al., 2006; Vitari et al., 2006). The activity and phosphorylation of NKCC family cotransporters is stimulated by hyperosmotic stress (Lytle and Forbush, 1992; Kurihara et al., 1999; Darman and Forbush, 2002), conditions that have also been reported to enhance WNK1 activity (Xu et al., 2002; Lenertz et al., 2005) and induce phosphorylation of NKCC1 at the sites targeted by SPAK/OSR1 in vitro (Vitari et al., 2006). In this study, we investigate the mechanism by which WNK1 as well as its substrates SPAK/OSR1 are regulated by hyperosmotic stress.

It has been reported that hyperosmotic stress stimulates WNK1 activity (Xu et al., 2002; Lenertz et al., 2005). As a prelude to investigating the mechanism by which WNK1 is regulated, we further assessed the activity of endogenous WNK1 in 293 cells. We found that WNK1 immunoprecipitated from sorbitol-treated cells phosphorylated OSR1 at an approximately threefold higher rate than WNK1 derived from nontreated cells (Fig. 1 A). Consistent with WNK1 mediating this phosphorylation, no phosphorylation of OSR1 was detected when a mutated form of OSR1 in which the T-loop (Thr185) and S-motif (Ser325) sites phosphorylated by WNK1 were changed to Ala or when preimmune antibody was used instead of the specific anti-WNK1 antibody in the immunoprecipitation step (Fig. 1 A). A dose-dependent increase in WNK1 activity was observed by treatment with 0.1–1.0 M sorbitol, resulting in up to a fivefold increase of WNK1 activity (Fig. 1 B). Activation of WNK1 by 0.5 M sorbitol was detected within 0.5 min, with maximal activation being reached within 1–2 min, which was sustained for 80 min (Fig. 1 C). Consistent with hyperosmotic stress activating WNK1, treatment of cells with increasing concentrations of NaCl also induced activation of WNK1, which was detectable at 50 mM and maximal at 0.15 M (Fig. 1 D). Similarly, treatment of cells with 0.5 M KCl led to a threefold increase of WNK1 activity (Fig. 1 E). We next explored whether WNK1 was activated by other stimuli, namely, serum, IGF-1, phorbol ester (TPA), oxidative stress (H₂O₂), hypotonic stress (medium diluted twofold in H₂O), an NKCC1/2 inhibitor (bumetanide), TNFα, a tyrosine phosphatase inhibitor (pervanadate), or a serine/threonine phosphatase inhibitor (calyculin A). These treatments failed to induce activation of WNK1 under conditions in which they triggered activation of other signaling pathways (Fig. 1 E). As 0.5 M sorbitol stimulated the extracellular signal–regulated kinase (ERK) 1/2, p38, and JNK protein kinases (Fig. 1 E), we investigated whether these enzymes might be involved in the activation of WNK1. Treatment of cells with the MEK inhibitor PD184352 (Fig. 1 F) or U0126 (not depicted) abolished ERK activation by sorbitol, without affecting WNK1 activation. The p38 inhibitor SB203580 and high concentrations of BIRB0796 that inhibit all p38 and JNK isoforms (Kuma et al., 2005) also failed to prevent activation of WNK1 by sorbitol (Fig. 1 F). Consistent with p38 and JNK not regulating WNK1, anisomycin that potently stimulates p38 and JNK failed to activate WNK1 (Fig. 1 F).

Despite the large size of WNK1 (∼250 kD), we consistently observed a reduction in the electrophoretic mobility of endogenous WNK1 isolated from sorbitol-treated cells on a polyacrylamide gel (Fig. 2 A). To assess whether this was due to enhanced phosphorylation, 293 cells were labeled with ³²P-orthophosphate and endogenous WNK1 was immunoprecipitated from control and sorbitol-treated cells. Electrophoresis on a polyacrylamide gel revealed a Coomassie-stained band at ∼250 kD that was identified as WNK1 by mass spectrometry (Fig. 2 B). Autoradiographic analysis of the gel revealed that phosphorylation of WNK1 was stimulated by sorbitol (Fig. 2 B). The ³²P-labeled WNK1 from control and sorbitol-treated cells was digested with trypsin, and the resulting peptides were chromatographed on a C₁₈ column. Sorbitol substantially increased the abundance of several ³²P-labeled peptides (Fig. 2 C). All phosphopeptides were subjected to mass spectroscopy and some of them to solid phase Edman sequencing analysis, which enabled us to identify five sites of phosphorylation (Fig. 2 C). Phosphorylation of two sites, Ser1261 and Ser 2372, was stimulated by treatment of cells with sorbitol, whereas Ser2012, Ser2029, and Ser2032 were phosphorylated at similar levels in WNK1 isolated from control and sorbitol-stimulated cells.

As several of the phosphopeptides observed in the ³²P-cell–labeling experiments presented in Fig. 2 B were not present at sufficient levels to enable phosphorylation site assignment, we undertook a mass spectrometry phosphopeptide mapping analysis of a larger amount of endogenous WNK1 immunoprecipitated from unlabeled control and sorbitol-treated 293 cells (Fig. 2 D). The immunoprecipitated WNK1 was digested with trypsin, and the resulting peptides were subjected to liquid chromatography/mass spectrometry with precursor ion scanning on a Q-TRAP mass spectrometer, which enabled us to assign five residues (Ser15, Ser167, Ser382, Ser1261, and Thr1848) whose phosphorylation was markedly enhanced by treatment of cells with sorbitol (Fig. 2 D). One of these sites, Ser382, which lies within the T-loop of the WNK1 catalytic domain, has been reported to be an autophosphorylation site on WNK1 catalytic domain expressed in Escherichia coli (Xu et al., 2002). The location of all the phosphorylation sites identified within WNK1 and the amino acid sequences surrounding these sites are illustrated in Fig. 2 (E and F).

To verify whether the activity of WNK1 was influenced by its phosphorylation state, we incubated WNK1 isolated from control and sorbitol-treated 293 cells with the serine/threonine protein phosphatase-1γ (PP1γ) and assessed its activity. We observed that incubation with PP1γ substantially decreased both the basal and sorbitol-stimulated activity of WNK1 and also increased WNK1 electrophoretic mobility (Fig. 2 G). Microcystin-LR, a toxin that specifically inhibits serine/threonine phosphatases, prevented PP1γ-induced inactivation as well as the increase in WNK1 electrophoretic mobility (Fig. 2 G).

We observed that an N-terminal fragment of WNK1 encompassing residues 1–667 retained the property of becoming activated after stimulation of 293 cells with sorbitol (Fig. 3). Mutation of Ser382 that lies within the T-loop of WNK1 to Ala prevented activation of WNK1 by sorbitol treatment (Fig. 3 A), consistent with the previous report that this mutation inhibited WNK1 activity expressed in E. coli (Xu et al., 2002). Mutation of Ser382 to Glu enhanced the basal activity of WNK1[1–667] to a level that was over eightfold higher than that observed for wild-type WNK1[1–667] isolated from untreated cells (Fig. 3 A). The activity of the WNK1[S382E, 1–667] mutant was not enhanced by sorbitol stimulation.

Sorbitol markedly increased recognition of WNK1[1–667] by a phospho-Ser382 antibody that we generated (Fig. 3 A). Consistent with this antibody being specific, mutation of Ser382 to Ala abolished its recognition of WNK1 (Fig. 3 A). We also observed that sorbitol treatment induced Ser382 phosphorylation of a catalytically inactive WNK1[D368A, 1–667] mutant to the same extent as the catalytically active WNK1[1–667] fragment (Fig. 3 A), indicating that phosphorylation of WNK1 at Ser382 was not mediated by intramolecular autophosphorylation.

To determine whether WNK1 possesses the intrinsic ability to autophosphorylate on Ser382, we expressed wild-type WNK1[1–661] or kinase-inactive WNK1[D368A, 1–661] in E. coli. Consistent with previous work (Xu et al., 2002), we observed that the wild-type WNK1[1–661] was heavily phosphorylated at Ser382 and also possessed a high specific activity of 5.2 U/mg (Fig. 3 B). The activity of E. coli expressed WNK1[1–661] was much higher than that of the WNK1[1–667] fragment isolated from sorbitol-stimulated 293 cells, which had a specific activity of only ∼0.13 U/mg (Fig. 3). We observed that the kinase-inactive WNK1[D368A, 1–661] mutant expressed in E. coli was not phosphorylated at Ser382 (Fig. 3 B). Moreover, the WNK1[S382A, 1–661], WNK1[S382E, 1–661], and WNK1[S382D, 1–661] mutants expressed in E. coli possessed low activities of 0.15, 0.66, and 0.26 U/mg, respectively (Fig. 3 B). Incubation of wild-type WNK1[1–661] with MgATP in vitro did not further increase phosphorylation of Ser382 (Fig. 3 C).

As outlined in the introduction, WNK1 interacts with its substrates SPAK and OSR1 through RFXV/I motifs. Interestingly, the site of phosphorylation, Ser1261, is located adjacent to such a motif (Fig. 2 E), suggesting that phosphorylation of Ser1261 might influence the ability of WNK1 to bind SPAK and OSR1. To investigate this, we generated biotinylated peptides that encompass residues surrounding Ser1261 in its phosphorylated or nonphosphorylated form. These were conjugated to streptavidin–Sepharose and tested for ability to affinity purify endogenously expressed SPAK and OSR1 from 293 cell extracts. The nonphosphorylated Ser1261 peptide interacted with SPAK and OSR1 to a markedly greater extent than the phosphorylated peptide (Fig. 4 A). Using a surface plasmon resonance binding assay, we also observed that the nonphosphorylated Ser1261 peptide interacted with a dissociation constant of ∼40 nM with the isolated CCT domain of OSR1, whereas the phosphorylated Ser1261 peptide bound to OSR1 with a markedly lower affinity (Fig. 4 B).

We next studied whether the downstream substrates of WNK1, SPAK and OSR1, were activated by hyperosmotic stress in HeLa and 293 cells. Endogenous SPAK and OSR1 were immunoprecipitated from control and sorbitol-treated cells with an antibody that immunoprecipitates both proteins and assayed using a peptide substrate termed CATCHtide (Vitari et al., 2006). The activity of SPAK/OSR1 was increased approximately fivefold by sorbitol stimulation of HeLa cells (Fig. 5 A, top) and approximately fourfold in 293 cells (Fig. S1, available). This is consistent with previous reports that these enzymes are activated by sorbitol (Chen et al., 2004; Anselmo et al., 2006). We raised phosphospecific antibodies that recognize the T-loop and S-motif residues on SPAK and OSR1 phosphorylated by WNK1. The sequences surrounding these sites are almost identical in SPAK and OSR1 (Vitari et al., 2005), suggesting that the T-loop and S-motif phosphospecific antibodies should recognize the phosphorylated forms of both SPAK and OSR1. The specificity of the antibodies was confirmed by overexpressing OSR1 (Fig. 5 B) or SPAK (Fig. 5 C) in 293 cells and finding that mutation of the T-loop or S-motif residues to Ala abolished antibody recognition. Using these antibodies, we demonstrated that sorbitol treatment markedly stimulated the phosphorylation of endogenously expressed SPAK and OSR1 at both the T-loop and S-motif in HeLa cells (Fig. 5 A) as well as in 293 cells (Fig. S1). Sorbitol also induced a decrease in the electrophoretic mobility of endogenous SPAK. Studying overexpressed OSR1 (Fig. 5 B) and SPAK (Fig. 5 C) in 293 cells, we observed that mutation of the T-loop, but not the S-motif, to Ala, inhibited sorbitol-induced activation of these enzymes. We also found that mutation of the T-loop residue to Glu markedly increased the basal activity of OSR1 and SPAK and the activity of these mutants was not further increased by sorbitol treatment of cells. Mutation of the S-motif to Glu moderately enhanced basal and sorbitol-stimulated OSR1 (Fig. 5 B) and SPAK (Fig. 5 C) activity.

We next investigated the effect of depleting the levels of endogenous WNK1 using a siRNA duplex targeting WNK1 that reduces expression >90% (Fig. 5 A). Under these conditions, sorbitol-induced phosphorylation of the T-loop and S-motif of SPAK and OSR1 were markedly reduced. Depletion of WNK1 also inhibited the basal as well as sorbitol-induced SPAK/OSR1 activity by >50% without affecting sorbitol-induced phosphorylation of JNK isoforms at their T-loop residues (Fig. 5 A).

We also studied the localization of GFP-WNK1 (Fig. 6) that was stably expressed at a similar level to that of endogenous WNK1 (Vitari et al., 2004) as well as the localization of endogenous WNK1 in 293 cells (Fig. 7). In unstimulated cells, WNK1 was diffusely localized throughout the cytosol (Figs. 6 and 7, panel 1), but after treatment with 0.2 M sorbitol for only 1 min, WNK1 was strikingly observed on discrete intracellular structures. Colocalization experiments revealed that both GFP-WNK1 and endogenous WNK1 colocalized with clathrin (Figs. 6 and 7, panel 2) as well as AP-1 (Figs. 6 and 7, panel 3), which is an adaptor for clathrin and is recruited to budding vesicles at the TGN and endosomes (Hirst and Robinson, 1998). Partial colocalization between GFP-WNK1 and the TGN46 integral membrane protein that is predominantly localized to the TGN was also observed (Fig. 6, panel 4). In contrast, WNK1 did not colocalize with the early endosomal markers EEA1 (Mu et al., 1995) and Hrs (Raiborg et al., 2001) or the late endosome and lysosome marker LAMP1 (Rohrer et al., 1996; Figs. 6 and 7). Nor did it colocalize with the AP-2 adaptor (Figs. 6 and 7), which plays a central role in clathrin-mediated endocytosis by linking transmembrane receptors to be internalized to the clathrin lattice (Hirst and Robinson, 1998). We next investigated the redistribution of stably expressed GFP-WNK1 in living 293 cells using time-lapse microscopy (Fig. 8 and Videos 1 and 2, available). The redistribution of GFP-WNK1 to intracellular vesicles after sorbitol (Video 1) or NaCl (Video 2) treatment was rapid and observed within ∼0.5 min, the earliest time point that we could monitor (Fig. 8 A). Frames of the video were taken every 2 s, and substantial movement of some of the GFP-WNK1 localized was observed within a 10–30-s time frame (Videos 1 and 2 and Figs. S2 and S3). This effect was reversible, as removal of sorbitol or NaCl resulted in WNK1 becoming diffusely localized in the cytosol within 2 min (Fig. 8 B). We also undertook an experiment of FRAP to monitor the dynamics of WNK1 movement within the cells. In sorbitol- or NaCl-treated cells (Fig. 8 C), WNK1 became redistributed to the photobleached area within 2 min. This was slower than in untreated cells, where the recovery occurred within 0.5 min (unpublished data).

We observed that the C-terminal noncatalytic domain of WNK1 comprising residues 1504–2382 fused to GFP redistributed to intracellular vesicles after sorbitol stimulation, similar to full-length GFP-WNK1 (Fig. 9). In contrast, the N-terminal fragment (residues 1–667) comprising the catalytic domain remained diffusely localized throughout the cytoplasm after sorbitol stimulation.

Our findings are consistent with previous reports (Xu et al., 2002; Lenertz et al., 2005) suggesting that endogenously expressed WNK1 is specifically activated by hyperosmotic conditions (sorbitol, NaCl, and KCl) and not by other stresses (H₂O₂, anisomycin, and phosphatase inhibitors), growth factors, cytokines, phorbol esters, or the diuretic hypotensive agent bumetanide (Fig. 1). The activation of WNK1 by sorbitol is rapid and observed within 0.5 min, the earliest time point that we can practically investigate. Our data indicate that hyperosmotic stress activates WNK1 by stimulating its phosphorylation as incubation of WNK1 with PP1γ lead to a decrease in its activity. Phosphopeptide mapping resulted in the identification of six residues on endogenous WNK1 whose phosphorylation is stimulated with sorbitol (Fig. 2). Two of these sites are located N-terminal to the kinase domain (Ser15 and Ser167), one within the T-loop of the kinase domain (Ser382), and the other three sites are within the C-terminal noncatalytic region (Ser1261, Thr1848, and Ser2372). In addition, we have identified three other phosphorylation sites within the C-terminal region of WNK1 (Ser2012, Ser2029, and Ser2032) that are constitutively phosphorylated. We cannot rule out the possibility that there are additional phosphorylation sites on WNK1 that we have not been able to identify. Apart from Ser15, all the phosphorylation residues identified are conserved in mouse and rat WNK1. In Drosophila melanogaster WNK1, Ser382 and Ser2372 are conserved, whereas in C. elegans WNK1, only Ser382 is conserved. The only phosphorylation site that is present in all human WNK isoforms is Ser382, and the residues surrounding this site are also identical in all WNK isoforms. Ser2372 is found in WNK2 and -3, but not WNK4. Alignment of the sequences surrounding the WNK1 phosphorylation sites (Fig. 2 F) indicates that they are quite distinct, suggesting that different upstream kinases may be phosphorylating these residues in vivo. The only similarity between three of the phosphorylation sites (Ser1261, Ser2029, and Ser2032) is that they possess a proline residue following the site of phosphorylation. It would be interesting to investigate whether these residues were phosphorylated by a proline-directed kinase, such as an isoform of p38 or JNK. If this was the case, the phosphorylation of these sites may not be regulating WNK1 activity, as inhibitors of p38–JNK–ERK pathways did not prevent WNK1 activation (Fig. 1 F).

Our results suggest that phosphorylation of Ser382 is required for sorbitol-induced activation of WNK1, as mutation of Ser382 to Ala prevented WNK1 activation by sorbitol, whereas mutation of Ser382 to Glu to mimic phosphorylation increased activity of WNK1 and prevented further activation by sorbitol (Fig. 3). The conservation of Ser382 in all species of WNK isoforms is consistent with the notion that phosphorylation of this residue plays a crucial role in controlling the activity of WNK isoforms. It was previously reported that the isolated catalytic domain of WNK1 encompassing residues 198–491, when expressed in E. coli, was phosphorylated at Ser378 and Ser382 (Xu et al., 2002). This study showed that mutation of Ser382 to Ala inactivated the WNK1 enzyme, whereas mutation of Ser378 to Ala only moderately reduced activity (Xu et al., 2002). In our peptide-mapping studies with endogenously expressed full-length WNK1 we have not been able to detect phosphorylation of Ser378.

Our findings suggest that, in mammalian cells, phosphorylation of Ser382 may be mediated by a transautophosphorylation reaction. This conclusion is based on the finding that wild-type WNK1, but not kinase-inactive WNK1, when expressed in E. coli, is phosphorylated at Ser382 (Fig. 3, B and C). In contrast, we observe that a catalytically inactive WNK1 mutant is normally phosphorylated at Ser382 in response to sorbitol when expressed in 293 cells (Fig. 3). These observations could be explained by the ability of endogenous WNK1 to transphosphorylate the kinase-inactive WNK1 mutant at Ser382 (Fig. 3). However, our data do not rule out the possibility that there is another upstream kinase capable of phosphorylating Ser382 in response to hyperosmotic stress. Nor do our data rule out the possibility that hyperosmotic stress stimulates phosphorylation of Ser382 by inhibiting a protein phosphatase.

Further work is required to establish the roles of the novel sites of phosphorylation on WNK1 that we have identified. Apart from Ser382 (Fig. 3), we have also analyzed the effects that mutating other sorbitol-stimulated phosphorylation sites (Ser167, Ser1261, Thr1848, and Ser2372) had on regulating the activity of WNK1 expressed in 293 cells. We observed that individual mutation of these residues to either Ala or Glu did not markedly affect basal or sorbitol-stimulated WNK1 activity (unpublished data). We suggest that phosphorylation of Ser1261 inhibits the interaction with the CCT domain of SPAK and OSR1 (Fig. 4). There are four RFXV/I motifs in the C-terminal region of WNK1, and a Ser or Thr residue follows all of these potential CCT binding sites. Interestingly, a Ser/Thr residue also follows several other RFXV/I motifs in other proposed SPAK/OSR1 binding proteins, including WNK4, NKCC2, and NCC. It is possible that phosphorylation of residues following RFXV/I motifs comprises a mechanism for dissociating WNK1 from the CCT domains of SPAK and OSR1. We have attempted to detect dissociation of SPAK/OSR1 from WNK1 after sorbitol stimulation of cells but have not observed a marked decrease in the association between SPAK/OSR1 and WNK1. It is possible that nonstoichiometric phosphorylation of Ser1261 and/or the presence of other RFQV/I SPAK/OSR1 binding sites in WNK1 masks detection of the dissociation of SPAK/OSR1 from WNK1 in such an experiment.

We establish that hyperosmotic stress induces activation and phosphorylation of endogenous SPAK and OSR1 at their T-loop and S-motif, the sites phosphorylated by WNK1 in vitro (Fig. 5). Moreover, depletion of WNK1 by ∼90% using siRNA methodology markedly inhibited T-loop and S-motif phosphorylation of SPAK/OSR1 and repressed basal and sorbitol-induced activation of these enzymes. Other isoforms of WNK or the remaining low level of WNK1 in these cells may mediate the residual SPAK/OSR1 phosphorylation and activity. These data provide further evidence that SPAK/OSR1 are indeed regulated by WNK1 in vivo. While this study was under review, it was reported that siRNA knock down of WNK1 reduced OSR1 activity as well as its total phosphorylation in sorbitol-treated cells, but the sites of phosphorylation affected by WNK1 knockdown were not investigated in this study (Anselmo et al., 2006). Our mutational analysis also confirms that phosphorylation of the T-loop of SPAK/OSR1 is required for sorbitol-induced activation of these enzymes (Fig. 5, B and C). Moreover, we find that individual mutation of the T-loop or S-motif does not affect phosphorylation of the other residue in sorbitol-stimulated cells.

We observed that hyperosmotic stress (sorbitol and NaCl) induced a striking relocalization of WNK1 to intracellular vesicles that are highly mobile (Videos 1 and 2) and colocalize with clathrin, AP-1, and partially with TGN46, but not with the endosomal markers EEA1, Hrs, LAMP1, or AP-2, which colocalizes with plasma membrane–coated pits (Figs. 6 and 7). These results are consistent with the notion that after hyperosmotic stress, a considerable pool of WNK1 is localized to TGN/recycling endosomes. We have not been able to demonstrate that endogenous clathrin and WNK1 coimmunoprecipitate with each other from sorbitol-stimulated cells (unpublished data), indicating that these proteins may not interact directly. The trafficking of several ion channels and cotransporters between intracellular vesicles and plasma membrane (see Introduction) is strongly influenced by the overexpression of WNK isoforms. It is therefore possible that relocalization of WNK1 to TGN/recycling endosomes may enable it to regulate trafficking and/or activity of certain ion channels/cotransporters. Our findings based on the overexpression of fragments of WNK1 in 293 cells indicate that a C-terminal noncatalytic region of WNK1 mediates this relocalization (Fig. 9). Overexpression of catalytically inactive WNK1 or -4 decreased the membrane expression of the renal outer medullary potassium channel (ROMK) through a clathrin-dependent endocytosis mechanism (Kahle et al., 2003; Cope et al., 2005). Overexpression of the C-terminal noncatalytic region of WNK3 decreased membrane expression of the potassium channel (Leng et al., 2006), and the C-terminal portion of WNK4 coimmunoprecipitated with ROMK in 293 cells (Kahle et al., 2003). Overall, these data suggest that the C-terminal portion of the WNK isoforms plays an important role in influencing WNK cellular localization and function.

In conclusion, our study defines the striking effects that exposure of cells to hyperosmotic conditions has on WNK1 phosphorylation, cellular localization, and catalytic activity as well as on its ability to interact with and activate its substrates SPAK and OSR1. In future studies, it would be important to determine whether the capacity of hyperosmotic stress to stimulate WNK1 and SPAK/OSR1 plays a role in controlling cell volume and blood pressure. It will also be interesting to establish whether the reported effects that WNK isoforms have on ion channels and other cotransporters are mediated through activation of SPAK/OSR1 and/or translocation of WNK1 to TGN/recycling endosomes. It would also be essential to address whether WNK2, -3, and -4 isoforms are regulated in a manner similar to WNK1.

The following antibodies were raised in sheep and affinity purified on the appropriate antigen: the SPAK/OSR1 (total) antibody was raised against the human full-length SPAK and OSR1 proteins, and the SPAK/OSR1 (T-loop) phospho-T233/T185 antibody was raised against a phosphopeptide encompassing residues 226–238 of human SPAK or residues 178–190 of human OSR1 (TRNKVRKTpFVGTP). The SPAK/OSR1(S-motif) phospho-S373/S325 was raised against a phosphopeptide encompassing residues 367–379 of human SPAK (RRVPGSSpGHLHKT), which is highly similar to residues 319–331 of human OSR1 in which the sequence is RRVPGSSpGRLHKT. The antibody used for immunoblotting WNK1, termed WNK1(total), was raised in sheep against a fragment encompassing residues 61–661 of human WNK1. The antibody used for immunoprecipitation of WNK1, termed WNK1(CT), was raised against a peptide encompassing residues 2360–2382 of human WNK1 (QNFNISNLQKSISNPPGSNLRTT).

The PKB (total) antibody was raised in sheep against a peptide encompassing residues 466–480 of rat PKBα (RPMFPQFSYSASGTA). The following antibodies were purchased from Cell Signaling: ERK1/2 (total), ERK1/2 phospho-Thr202/Tyr204, p38α (total), p38α phospho-Thr180/Tyr182, and PKB phospho-Thr308. The JNK (total) and JNK phospho-Thr183/Tyr185 antibodies were obtained from Biosource International. The mouse monoclonal antibody recognizing the GST tag was purchased from Roche. Secondary antibodies coupled to horseradish peroxidase used for immunoblotting were obtained from Pierce Chemical Co. For Li-COR analysis, the IRDye 800-conjugated anti-sheep antibody was purchased from Rockland. Preimmune IgG used in control immunoprecipitation experiments were affinity purified from preimmune serum using protein G–Sepharose. The clathrin antibody was purchased from Abcam, the EEA1 and LAMP1 antibodies (raised in mouse) were purchased from BD Biosciences, and the Hrs antibody was a gift from H. Stenmark (The Norwegian Radium Hospital, Oslo, Norway) and has been described previously (Raiborg et al., 2001). The AP-1 antibody (monoclonal anti–γ-adaptin AP6) was obtained from Sigma-Aldrich. The AP-2 antibody (monoclonal anti–α-adaptin) was obtained from Affinity BioReagents, Inc. The TGN46 antibody (polyclonal produced in sheep) was purchased from Serotec. Alexa Fluor 488 donkey anti–sheep, Alexa Fluor 595 donkey anti–rabbit, and Alexa Fluor 595 donkey anti–mouse secondary antibodies were obtained from Invitrogen.

Anti-WNK1(CT) and preimmune IgG antibodies were covalently coupled to protein G–Sepharose (1 μg of antibody per 1 μl of beads) using a dimethyl pimelimidate cross-linking procedure. 0.5 mg of clarified cell lysate was incubated with 5 μg of anti-WNK1(CT) or preimmune IgG antibody conjugated to 5 μl of protein G–Sepharose and incubated for 1 h at 4°C with gentle agitation. The immunoprecipitates were washed twice with 1 ml of lysis buffer containing 0.5 M NaCl and twice with 1 ml of buffer A. The in vitro phosphorylation reaction mix contained a final volume of 25 μl in buffer A containing 5 μM OSR1[D164A], 0.1 mM [γ³²P]ATP, and 10 mM magnesium acetate. The assays were performed for 20 min at 30°C, and the reactions were terminated by adding SDS sample buffer. The samples were electrophoresed on a polyacrylamide gel, which was stained with Coomassie blue, dried, and autoradiographed. The OSR1 Coomassie bands were excised, and incorporation of ³²P-radioactivity was quantified by Cerenkov counting. 1 U of activity was defined as the amount of WNK1 that incorporated 1 nmol of ³²P into OSR1[D164A]. For experiments in Fig. 3 A using overexpressed forms of GST-WNK1, WNK1 was affinity purified from 0.1 mg of cell lysate using 5 μl of glutathione–Sepharose. The beads were washed, and assays were undertaken as described above.

3 mg of clarified cell lysate was incubated with 5 μg of the SPAK/OSR1 (total) antibody conjugated to 5 μl of protein G–Sepharose and incubated for 1 h at 4°C with gentle agitation. The immunoprecipitates were washed twice with 1 ml of lysis buffer containing 0.5 M NaCl and twice with 1 ml of buffer A. The SPAK/OSR1 immunoprecipitates were either assayed with the CATCHtide peptide substrate (RRHYYYDTHTNTYYLRTFGHNTRR) that encompasses the SPAK/OSR1 phosphorylation sites on NKCC1 (Vitari et al., 2006) or using the N-terminal fragment of NKCC1 encompassing residues 1–260 (NKCC1[1–260]; Vitari et al., 2005). Assays were set up in a total volume of either 50 μl (CATCHtide assay) or 25 μl (NKCC1 assay) in buffer A containing 10 mM MgCl₂, 0.1 mM [γ³²P]ATP (∼300 cpm/pmol) and 300 μM CATCHtide (RRHYYYDTHTNTYYLRTFGHNTRR), or 5 μM NKCC1[1–260]. After incubation for 10–60 min at 30°C, the reaction mixture was applied onto P81 phosphocellulose paper (for the CATCHtide assay), the papers were washed in phosphoric acid, and incorporation of ³²P-radioactivity in CATCHtide was quantified. For the NKCC1 assay, the reaction was stopped by the addition of SDS sample buffer. The samples were electrophoresed on a polyacrylamide gel, which was stained with Coomassie blue, dried, and autoradiographed. The NKCC1[1–260] Coomassie bands were excised, and incorporation of ³²P-radioactivity was quantified by Cerenkov counting. For experiments in Fig. 5 (B and C) using overexpressed forms of GST-OSR1 and -SPAK, the GST fusion proteins were affinity purified from 0.5 mg of cell lysate using 5 μl of glutathione–Sepharose. The beads were washed, and assays were undertaken as described above, using NKCC1[1–260] as a substrate.

293 cells stably expressing GFP-WNK1 at levels similar to that of endogenous WNK1 were described previously (Vitari et al., 2004). In experiments in which the cells were fixed before analysis, the cells were grown on coverslips (no. 1 1/2) and, after stimulation, were fixed for 5 min in freshly prepared 3% vol/vol formaldehyde in PBS. The cells were washed twice in PBS (5 min each wash) and once in PBS containing 0.2% wt/vol Triton X-100, incubated 10 min in PBS-TG (PBS containing 0.2% wt/vol Tween +3% wt/vol of fish skin gelatine), and incubated for 1 h with anti-clathrin/Hrs/AP-2/TGN46 antibodies (diluted 1:1,000 in PBS-TG), anti-EEA1 antibody (diluted 1:100 in PBS-TG), anti-LAMP1 antibody (diluted 1:50 in PBS-TG), or anti–AP-1 (diluted 1:2,500 in PBS-TG). Cells were washed three times in PBS-T (PBS containing 0.2% wt/vol Tween), incubated for 30 min with the secondary antibody diluted 1:500 in PBS-TG, washed three times in PBS-T and once in water, and mounted onto slides using hydromount (National Diagnostics). Images were collected using the α Plan 100× 1.45 NA Plan-Fluor objective on a confocal microscope (LSM 510 META; Carl Zeiss MicroImaging, Inc.). Fixed cells were imaged at room temperature (∼20°C) and live cells at 37°C. Images were acquired using LSM 510 acquisition software (Carl Zeiss MicroImaging, Inc.), and no further processing of images was performed apart from assembling montages in Photoshop/Illustrator (Adobe). Videos were edited in QuickTime Pro. For imaging of endogenous WNK1, untransfected 293 cells were treated as described above except that 5 μg/ml of anti-WNK1(Total) antibody was used. For live-cell imaging, 293 cells stably expressing GFP-WNK1 were grown on 35-mm-diameter glass-bottomed dishes (Willco). Cells were maintained at 37°C and 5% CO₂ by the use of a microscopy incubator chamber (Carl Zeiss MicroImaging, Inc.). For each cell, optical sections of 0.5 μm were recorded at 2-s (for Videos 1 and 2 and Figs. S2 and S3) or 10-s (Fig. 8) intervals. FRAP was performed using a bleached region of interest defined by the LSM 510 META software. Recovery of fluorescence was monitored by collecting images every second until recovery was complete. See the supplemental text (available) for further methodological details.

The supplemental text contains additional methodological details on materials, buffers, and DNA constructs used, as well as protocols used for in vivo ³²P-labeling, identification of phosphorylation sites by mass spectrometry, immunoprecipitation, immunoblotting, and siRNA knockdown protocol. Fig. S1 shows that hyperosmotic stress leads to activation and phosphorylation of SPAK and OSR1 in 293 cells. Figs. S2 and S3 show selected time frames of videos of cells treated with sorbitol (Fig. S2 and Video 1) or NaCl (Fig. S3 and Video 2) to illustrate movement of vesicles to which GFP-WNK1 is recruited.

We thank Harold Stenmark for providing the Hrs antibody, Natalia Shpiro and Rodolfo Marquez for synthesis of PD184352 and BIRB0796, Colin Bell and Agnieszka Kieloch for tissue culture, the Sequencing Service (School of Life Sciences, University of Dundee, Scotland) for DNA sequencing, and the Post Genomics and Molecular Interactions Centre for Mass Spectrometry Facilities (School of Life Sciences, University of Dundee, Scotland) and the protein production and antibody purification teams (Division of Signal Transduction Therapy, University of Dundee), coordinated by Hilary McLauchlan and James Hastie, for generation and purification of antibodies.

E. Pozo-Guisado was supported by a Spanish Ministry of Education and Science fellowship. A. Zagorska and F.H. Rafiqi are the recipients of a 4-yr Wellcome Trust Studentship. A.C. Vitari was supported by a Pfizer-sponsored studentship. We thank the Association for International Cancer Research, Diabetes UK, the Medical Research Council, the Moffat Charitable Trust, and the pharmaceutical companies supporting the Division of Signal Transduction Therapy Unit (AstraZeneca, Boehringer-Ingelheim, GlaxoSmithKline, Merck & Co., Inc., Merck KgaA, and Pfizer) for financial support.