Chloride sensing by WNK1 kinase involves inhibition of autophosphorylation

Abstract

WNK1 [with no lysine (K)] is a serine-threonine kinase associated with a form of familial hypertension. WNK1 is at the top of a kinase cascade leading to phosphorylation of several cotransporters, in particular those transporting sodium, potassium, and chloride (NKCC), sodium and chloride (NCC), and potassium and chloride (KCC). The responsiveness of NKCC, NCC, and KCC to changes in extracellular chloride parallels their phosphorylation state, provoking the proposal that these transporters are controlled by a chloride-sensitive protein kinase. Here, we found that chloride stabilizes the inactive conformation of WNK1, preventing kinase autophosphorylation and activation. Crystallographic studies of inactive WNK1 in the presence of chloride revealed that chloride binds directly to the catalytic site, providing a basis for the unique position of the catalytic lysine. Mutagenesis of the chloride binding site rendered the kinase less sensitive to inhibition of autophosphorylation by chloride, validating the binding site. Thus, these data suggest that WNK1 functions as a chloride sensor through direct binding of a regulatory chloride ion to the active site, which inhibits autophosphorylation.

Introduction

Chloride ion is an important electrolyte involved in blood pressure maintenance, neuronal excitability and nociception, transcellular electrolyte transport, cell volume control, and airway fluid balance (1, 2). Chloride is uniquely and precisely regulated in diverse cell types; it is maintained at modest (30–60 mM) concentrations in most cells, and dips very low (10 mM) in actively transporting epithelia and neurons (3, 4). Misregulation of chloride is associated with diseases, such as hypertension and epilepsy. Chloride concentration can change radically as a function of osmotic stress (5), and low chloride induces cell cycle arrest (6). The main transmembrane proteins that set chloride concentrations are members of the solute carrier electroneutral cation-Cl⁻ cotransporters (SLC12A family, also known as CCCs). Other molecules that influence cellular chloride concentration include members of the SLC26 gene family of exchangers, and chloride channels, such as the cystic fibrosis transmembrane conductance regulator (CFTR) and the γ-aminobutyric acid-gated chloride channel (GABA_A) (7–9). However, the molecules that sense and respond to changes in intracellular chloride concentrations are largely unknown.

How chloride binds to proteins in any context is also poorly understood. Known chloride-binding sites tend to fall into two structural categories. One category is typified by the ClC family of chloride channels, which bind chloride through backbone-amide and hydrophobic interactions (10). A similar binding site is observed in the atrial naturetic peptide (ANP) receptor, a guanyl cyclase involved in volume regulation (11). Angiotensin converting enzyme (ACE) is an example of a second structural class of chloride-binding sites (12). The chloride-binding site in ACE has both hydrophobic and positively charged amino acids. Other proteins that have this binding mode include the serotonin transporter (SERT) and α-amylase (13, 14).

The regulation of transport and other processes through chloride has been demonstrated in various cellular assays, but often the structural and biochemical mechanisms are poorly defined. Although the chloride-mediated regulation of a few soluble proteins has been defined at both a biochemical and structural level, for example α-amylase and hemoglobin (13, 15), how changes in chloride concentration regulate other proteins including those in cellular membranes is unknown. For example, chloride regulates the voltage-gated potassium channel KCNB1, increasing the K⁺ current as a function of chloride in patch-clamp studies (16). The chloride-dependent Na⁺/H⁺ exchanger (Cl− NHE) is activated by chloride, as determined by intracellular pH measurements on exchanger-transfected cells (17). Both of these processes occur through unknown mechanisms.

Chloride concentrations are set primarily by the action of the specific members of the CCC family, in particular the transmembrane cotransporters NKCC1 (sodium-potassium-chloride cotransporter 1) and KCC1 (potassium-chloride cotransporter 1) (1, 18). The cotransporters are passive, acting according to the electrochemical gradient: NKCC1 mediates influx of ions, KCC1 mediates efflux. The activities of these transporters are regulated. Low intracellular chloride concentrations activate NKCC1 in secreting epithelia (19). Each of the cation-chloride contransporters is also regulated by phosphorylation. NKCCs, as well as sodium-chloride cotransporters (NCCs), are activated by phosphorylation (20, 21). KCCs are inhibited by phosphorylation (22). As a natural consequence of these combined observations, it was hypothesized that there should be a chloride-sensitive kinase (19, 20, 23). This putative kinase might act both as a primary sensor of chloride and transducer of this information to the cotransporters. NKCCs and KCCs also regulate cell volume, which has led to the further hypothesis that there should also be a chloride- and volume-sensitive protein kinase (24, 25).

Kinases of the WNK [with no lysine (K)] family are serine/threonine protein kinases noted for the unique placement one of the catalytic lysine residues (26, 27). Shortly after these kinases were first cloned, mutations in WNK1 or WNK4 genes were found to underlie pseudohypoaldosteronism II, a familial disease characterized by hypertension, hyperkalemia, and hyperchloremia (28). This disorder is often treated with thiazide diuretics, which are inhibitors of NCCs. Pseudohypoaldosteronism II is symptomatically the inverse of Gitelman’s syndrome, which is due to loss of the SLC12A3 gene encoding an NCC (29). This connection suggested that WNKs are involved in the pathway of SLC12A cotransporter phosphorylation. The linkage was established in studies of NKCC1 (SLC12A2). First, the kinases OSR (oxidative stress responsive kinase) and SPAK (sterile-20 proline alanine-rich kinase) were shown to interact with and phosphorylate the cotransporters NKCC1 and KCCs (30, 31). Subsequent studies showed that WNK1 interacts with and phosphorylates OSR and its close homolog SPAK (32–36). Apparently, there is a two-tiered kinase cascade; WNK phosphorylates and activates OSR and SPAK, which phosphorylate and regulate CCCs.

WNK isoforms exhibit pleiotropic involvement in ion transport regulation. WNK1 and WNK4 inhibit the epithelial sodium channel (ENaC), which is a major mediator of aldosterone-induced salt reabsorption (37, 38), by promoting ubiquitin-mediated degradation of the channel (39). The regulation of ENaC appears to be mediated by SGK1 (aldosterone-regulated serum- and glucocorticoid-induced kinase), which is activated by WNK1 (40). WNK1 and WNK4 also inhibit the renal outer medullary potassium channel (ROMK), a channel involved in salt reabsorption (41, 42). In contrast to the ENaC regulation, WNK-mediated inhibition of ROMK is due to clathrin-mediated endocytosis. Both effects on ENaC and ROMK appear to require regions outside the kinase domain of WNK, and the involvement of kinase activity is unknown. WNK1 and WNK4 also affect paracellular chloride flux (43) and maintenance of vascular tone (44).

WNK isoforms are activated in response to reduction in intracellular chloride concentrations (36, 45). Activation of NKCC2 by reduced intracellular chloride is blocked by WNK mutants that affect WNK-OSR docking interactions (45). Further, overexpression of WNK3 in HEK293 cells increases intracellular chloride concentrations. WNKs have been proposed to be chloride-sensitive kinases on the basis of these observations (45–49). Here, we show that the kinase domain of WNK1 directly binds a chloride ion, and we locate the chloride-binding site in the previously determined crystal structure of the inactive WNK1 kinase domain (27). Chloride binding inhibited the autophosphorylation of WNK1, thereby inhibiting kinase activity. Furthermore, we confirmed the residues involved in chloride binding through identification of chloride-insensitive mutants. Our results suggest that WNKs function as direct chloride sensors.

Results

WNK1 is stabilized by chloride salts

Because WNK1 is a regulator of CCCs and because those transporters are salt sensitive, we explored the relationship between WNK1 stability and salt concentration in vitro. We examined the stability of two forms of the WNK1 kinase domain: the inactive form of the kinase domain (WNK1-KDm/SA, in which activation loop phosphorylation site Ser³⁸² was mutated to alanine) (27, 50), and the active kinase domain (WNK1-KDm/S*, where * denotes phosphorylation on Ser³⁸², on the activation loop). We estimated stability from the temperature of protein melting (T_m) as indicated by binding of the dye Sypro Orange (fig. S1)(51). Both the active and inactive proteins exhibited an ~10°C increase in T_m with increasing NaCl concentration (Fig. 1A). These results indicated that that the kinase domain of WNK1 may interact directly with either the sodium or chloride ion, or both, and that the interaction was independent of Ser³⁸².

Fig. 1. Kinase stability, activity, and autophosphorylation as a function of NaCl concentration.

(A) Change in melt temperature (T_m) of WNK constructs as a function of [NaCl]. T_m is taken at the point of greatest dRFU/dt during differential scanning fluorimetry from 0 to 80 °C, where RFU is relative fluorescence units. WNK1 193–482 (WNK1-KDm/S*) and inactive WNK1 193–482 S382A (WNK1-KDm/SA) are plotted. (B) and (C) Change in T_m of WNK1-KDm/S* upon exposure to different cation and anion salts. Stability was analyzed as in A. (D) Change in T_m of three control kinases (p38α, MEK6, and ASK1-KDm) as a function of [NaCl]. Stability was analyzed as in A. (E) Activity of WNK1 as measured by incorporation of ³²P onto a model substrate (MBP) as a function of [NaCl]. The highest measured activity was normalized to 1; a control measured in the absence of MBP was taken as 0. (F) Activity of the indicated kinases against MBP as a function of [NaCl]. Activity was analyzed as in E. (G) Autophosphorylation by WNK1-KDm and control kinases as a function of [NaCl] measured by ³²P autoincorporation. Solid lines and points indicate data normalized with 1 set to the highest recorded signal and 0 set based on analysis in the absence of kinase. Dotted lines and circles indicate data normalized to the number of phosphate atoms incorporated per protein molecule. In all panels, data are shown as the mean and standard error from three independent experiments.

To determine which ion WNK1-KDm bound, we measured the change in T_m as a function of different concentrations of cation-chloride salts (LiCl, NaCl, KCl) (Fig. 1B) and sodium-anion salts (NaCl, NaBr, NaI) (Fig. 1C). NaBr exposure stabilized WNK1-KDm, whereas NaI was destabilizing. LiCl and KCl had similar effects to that of NaCl on the stability of WNK1.

We also tested three kinases or catalytic domains from kinases that have not been implicated in chloride signaling: full-length mitogen- and extracellular kinase-regulated kinase kinase 6 (MEK6) and p38α and the catalytic domain of apoptosis signal-regulated kinase 1 (ASK1-KDm). None of these exhibited an increase in stability in response to increased NaCl; although the ASK1-KDm was destabilized as the NaCl concentration increased (Fig. 1D). Each of the salts destabilized MEK6, p38α, and ASK1-KDm to varying degrees, with iodide causing the greatest destabilization (fig. S2).

For the WNK1-KDm, halide ions (Cl⁻ and Br⁻) stabilized the catalytic domain; whereas iodide had a destabilizing effect (Fig. 1C). Because any of the cation-chloride salts stabilized WNK1-Dm, cation binding appeared less important than chloride binding.

WNK1 activity is inhibited by salts

If WNK1 activity is required to initiate the WNK1 cascade, then we would expect chloride binding to have some effect on WNK1-KDm/S* activity. We measured the activity of WNK1 toward the model substrate, myelin basic protein (MBP), as a function of NaCl concentration and found that WNK1-KDm/S* activity decreased as a function of NaCl, with an apparent IC₅₀ of 40 mM (Fig. 1E). However, two of the chloride-insensitive control kinases tested (ASK1-KDm and MEK6) also exhibited a decrease in activity in response to increasing NaCl concentrations with similar IC₅₀ [30 mM for both (Fig. 1F)]. WNK1-KDm/S* activity toward the kinase domain of the physiological WNK1 substrate OSR (OSR 1–295) was also inhibited by increasing concentrations of NaCl, but the IC₅₀ (530 mM) was greater than for activity toward MBP (fig. S3).

Because this inhibition of the active kinase was not unique, this mechanism seems unlikely to be responsible for the chloride-sensitive responsiveness of the WNK1 to CCC pathway.

Chloride inhibits WNK1 autophosphorylation

WNK1 requires phosphorylation at the activation loop site for activity (52). In the previous experiment, we used WNK1-KDm/S*, which is autophosphorylated as produced in E. coli. To determine if NaCl inhibited WNK1 autophosphorylation, we generated the unphosphorylated species, WNK1-KDm/S, by dephosphorylating WNK1-KDm/S* with protein phosphatase 1 (PP1cγ) and verified the dephosphorylated protein by mass spectrometry (fig. S4A). NaCl inhibited the autophosphorylation of WNK1-KDm/S, as assessed by autoradiography, with an apparent IC₅₀ of 20 mM (Fig. 1G). This IC₅₀ is consistent with the observed [Cl]_Internal, which varies from 5 to 60 mM depending on the cell type (3, 4). We verified the inhibition of autophosphorylation by chloride by mass spectrometry (fig. S4A–C). Neither of the control kinases MEK6 nor p38α autophosphorylated appreciably under these conditions (Fig. 1G). Although ASK1-KDm can autophosphorylate (53), the effect of NaCl on ASK1 autophosphorylation was weak (apparent IC₅₀ of 1 M) and opposite from the effect on WNK1. Thus, chloride has a strong, inhibitory effect on the autophosphorylation of WNK1-KDm, suggesting that changes in cellular chloride concentrations could modulate the rate of WNK1 autophosphorylation and thus its activity.

Halide ion binds the kinase domain of WNK1

Because bromide ion bound the WNK1-KDm (Fig. 1C) and because bromide gives a unique anomalous X-ray diffraction signal that can be used to identify its binding site (54), we crystallized the kinase-dead WNK-KDm/SA in the presence of 300 mM NaBr. Data were collected at the bromine K absorption edge (0.92 Å) (Table 1).

Table 1.

Data and refinement statistics for WNK1-KDm/SA

	WNK1-KDm/SA (3FPQ Re-refinement)	WNK1-KDm/SA Chloride Refinement	WNK1-KDm/SA Bromide
Space Group	P 1	P 1	P 1 21 1
Unit Cell	a=38.3 Å, b=57.8 Å, c=65.7 Å α=91.3°, β=90.0°, γ=90.9°	a=38.3 Å, b=57.8 Å, c=65.7 Å α=91.3°, β=90.0°, γ=90.9°	a=38.0 Å, b=58.3 Å c=65.0 Å, β=91.35°
Resolution (Å)	50-1.8 (1.9-1.8)†	50-1.8 (1.9-1.8)†	43-3.5 (3.56-3.5)†
Observed Reflections	217,158	217,158	47,703
Unique Reflections	50,614	50,614	6,887
Completeness	97.6 (96.9)	97.6 (96.9)	98.8 (96.1)
I/σI	27.9 (10.7)	27.9 (10.7)	44.3(4.3)
Redundancy	4.3 (4.3)	4.3 (4.3)	6.9 (6.6)
Rwork(%)‡	16.7	16.7	23.0
Rfree(%)	20.8	20.8	28.8
Rmerge (%)†	0.077 (0.144)	0.077 (0.144)	0.075 (0.517)
Number of groups
Protein Atoms	4,706	4,706	2,180
Water	345	344	10
Anion (Occupancy)	None	Chloride (1.0)	Bromide (0.48)
RMSD
Bond Length, Å	0.02	0.02	0.09
Bond Angle, °	1.97	1.98	5.22
Average B factor, Å2	27.2	27.3	24.3
Ramachandran Map (%)
Favorable	90.7	96.1	93.7
Allowed	8.6	3.0	4.5
Disallowed	0.6	0.9	1.8

^†Values in Parentheses indicate the highest resolution shell

^‡R_work = Σ ||F_obs − |F_calc||/Σ |F_obs|, where F_obs and F_calc are the observed and calculated structure factors, respectively.

The previously determined structure of WNK1-KDm/SA (PDB ID: 3FPQ) was solved in the presence of 300 mM NaCl (27) (Fig. 2A). We compared our bromide-containing structure to the published chloride-bound structure. The WNK1-KDm/SA bromide crystals adopted a similar space group to the chloride-bound WNK1-KDm/SA (Table 1). The dataset contained 47,703 reflections with a maximum resolution of 3.5 Å, We refined the bromide-bound structure using 3FPQ as a starting model.

Fig. 2. WNK1 halide binding is observed crystallographically.

(A) Cartoon representation of WNK1-KDm/SA, PDB ID: 3FPQ. α-helices are pink, β-sheets are cyan, and loops are grey. The chloride position (green) was determined from a 5.5 σ peak in the |F_o − F_c| map. The 3/10 helix and hydrogen-bonding distances are shown in the inset. (B) Cartoon representation of WNK1-KDm/SA (blue) crystallized in the presence of NaBr (with the ion shown in yellow). Data were collected at the bromine K-edge (0.92 Å). The anomalous difference Fourier map contoured at 3.5 σ is shown in yellow. (C) Top-down view of the 3/10 helix from (B) with a bromide anion (yellow) modeled into the anomalous peak to show hydrogen-bonding distances. (D) Alignment of PDB ID: 3FPQ (Pink) and WNK1-sKDm/S* (Blue). The chloride of PDB ID: 3FPQ is shown in green. A depiction of the disordered activation loop of WNK1-sKDm/S* is shown using blue dots, with one dot for each Cα. Pymol was used to generate the figures (77).

The refined bromide structure superimposed on the previously defined WNK1 structure (PDB ID: 3FPQ) with an all-atom alignment RMSD of 0.5 Å. The anomalous difference Fourier map, indicative of a bromide ion, revealed a single strong peak, 5.2 σ, near the backbone amides of Leu³⁶⁹ of the DLG sequence and the adjacent residue Gly³⁷⁰ (Fig. 2B). The distances to the backbone amide nitrogen atoms were 4.0 and 3.5 Å, indicative of NH-Br⁻ hydrogen bonds (55) (Fig. 2C). The bromide ion refined to an occupancy of 0.48 (Table 1).

To identify chloride ions in the previously determined structure of WNK1-KDm/SA (PDB ID: 3FPQ) (27), we calculated the map with coefficients |F_o − F_calc| (λ= 1.0 Å) and identified a 5.5 σ peak in a position equivalent to the bromide ion found in the bromide-containing crystals. No other peak of similar intensity was observed in the map. We modeled this peak, which was originally assigned as water, as chloride (Fig. 2A, inset). Refined hydrogen-bonding distances to the backbone amides of Gly³⁷⁰ and Leu³⁷¹ are 2.9 Å and 3.3 Å, typical of chloride hydrogen bonds (55). Assuming a chloride ion rather than water in that structure raised the B-factor from 14 Ų to 28 Ų, more in line with the B-factors of the rest of structure (Table 1).

This comparison indicated that the chloride binds WNK1-KDm/SA through the same backbone amides as the bromide binds. These residues, Gly³⁷⁰ and Leu³⁷¹, form the N-terminus of a short 3/10 helix, encompassing residues Gly³⁷⁰-Lys³⁷⁵. The 3/10 helix is at the N-terminus of the activation loop and continues from the DLG sequence (subdomain VII) (56). The helix contributes to the inactive configuration observed in the unphosphorylated molecule (27). Similar helices have been found in the inactive conformations of several kinases, and this helix has been noted as a common mechanism for building an inactive configuration (57). None of the other 3/10 helices, however, are capped by an anion, or even a water molecule. The chloride ion additionally forms hydrophobic interactions with the following residues: Phe²⁸³ in β3, Leu²⁹⁹ in β4, Leu³⁶⁹ in the DLG sequence, and Leu³⁷¹ in the 3/10 helix (fig. S5).

The structure of the phosphorylated form of WNK1 reveals conformational changes

To better understand the structural basis of the chloride inhibition, we determined the structure of the doubly phosphorylated kinase domain. A shorter fragment was used, eliminating residues missing in the electron density of the WNK1-KDm/SA, to form WNK1 (209–483)/S382* (WNK1-sKDm/S*). The WNK1-sKDm/S* crystals produced a 1.8 Ǻ structure (table S1). The WNK1-sKDm/S* structure is similar to WNK1-KDm/SA, with an overall RMSD of 0.55 Å. However, several differences were observed. Although a small amount of chloride (50 mM) was in the crystallization buffer, the chloride-binding site found in WNK1-KDm/SA is empty in WNK1-sKDm/S*. The 3/10 helix is present, but exhibits increased disorder and conformational changes, which are largest at the C-terminus (Lys³⁷⁵) (Fig. 2D and table S2). Disorder is also observed along helix αC, and in the activation loop. The N-terminal domain is rotated approximately 8° relative to the inactive structure (PDB ID: 3FPQ).

Several structural cues indicated that WNK1-sKDm/S* adopted an inactive configuration in the crystal, in particular the partial maintenance of the 3/10 helix and mispositioning of two catalytic residues Thr³⁸⁶ and Glu²⁶⁸. Thr³⁸⁶ at the C-terminus of the activation loop (58) is not present in the active site. Glu²⁶⁸ in helix αC which forms a hydrogen bond with the catalytic lysine in subdomain III in the active conformation (59), is displaced out of the active site. The 3/10 helices found in the activation loop of other inactive structures also disrupt this interaction (57). Disorder in the activation loop could also indicate an inactive conformer, but may be influenced by the lattice. A set of hydrophobic residues has been identified as a hallmark of active kinases, and referred to as a ‘spine’ (60). The corresponding residues in WNK1 are Phe²⁸³, Leu²⁷², Leu³⁶⁹, and His³⁴⁷. All contacts between these residues are maintained in both WNK1-KDm/SA and WNK1-sKDm/S*, although Leu²⁷² is rotated out with the displacement of helix C, such that the spine is not straight. Thus, WNK1-sKDm/S* adopts an inactive configuration despite the activating phosphorylation.

WNK1 with mutations in the chloride-binding site autophosphorylate

If chloride binding inhibited WNK1-KDm autophosphorylation as proposed, mutations in the chloride-binding site may reduce chloride binding and increase autophosphorylation. We introduced mutations into the hydrophobic residues lining the WNK1-KDm chloride-binding site as observed in WNK1-KDm/S* (L299F, L371F, or L369F) (Fig. 3A). These mutants were active and phosphorylated when expressed in and purified from E. coli [70–100% phosphorylation as determined by mass spectrometry (table S3). Increasing NaCl concentration had the least stabilizing effect on the mutant L369F (2.6 °C at 500 mM NaCl compared with 6.8 °C for WNK1-KDm/SA) (Fig. 3B). Using PP1cγ-treated proteins, we assessed the effect of increasing NaCl concentrations on autophosphorylation and the L369F mutant (in the sequence DLG) was less inhibited by NaCl, with an apparent IC₅₀ in the autophosphorylation experiment of ~50 mM (Fig. 3C). Even at concentrations in excess of 1M, we only observed 50% inhibition. The L371F mutant (in the 3/10 helix) also exhibited an increase in IC₅₀; whereas the L299F mutant (in β4) exhibited only slight changes when compared with WNK1-KDm/SA. None of the mutations substantially affected the ability of increasing concentrations of NaCl to inhibit activity toward MBP (Fig. 3D). Thus, the effects of these mutations on the inhibition of autophosphorylation and kinase activity toward substrates by chloride-containing salt substantiate the crystallographic data for the chloride-binding site and provide a context for understanding effect of chloride binding on autophosphorylation.

Fig. 3. WNK1-KDm/S* mutants.

(A) F283, L299, L369, and L371 form a hydrophobic pocket surrounding the bound chloride. (B) Change in melt temperature (T_m) of WNK1 mutants as a function of [NaCl]. T_m is taken at the point of greatest dRFU/dt during differential scanning fluorimetry from 0 to 80 °C, where RFU is relative fluorescence units. WNK1-KDm (Red), WNK1-KDm/L371F (Green), WNK1-KDm/L299F (Purple), and WNK1-KDm/L369F (Blue) are plotted. (C) Autophosphorylation of WNK1-KDm mutants in response to varying NaCl concentration as measured by ³²P incorporation. Autoincorporation normalized with 1 being the highest signal and a kinase-free background of 0. WNK1-KDm (Red), WNK1-KDm/L371F (Green), WNK1-KDm/L299F (Purple), and WNK1-KDm/L369F (Blue) are plotted. (D) Activity of WNK1 and WNK1 mutants as measured by incorporation of ³²P onto the generic kinase substrate myelin basic protein (MBP) as a function of [NaCl]. The highest measured activity was normalized to 1, and an MBP-deprived control was taken as 0. WNK1-KDm (Red), WNK1-KDm/L371F (Green), WNK1-KDm/L299F (Purple), and WNK1-KDm/L369F (Blue) are plotted. Data in B–D are shown as mean and standard error from three independent experiments.

Discussion

Here, we showed that the kinase domain of WNK1 binds chloride and that chloride has an inhibitory effect on the autophosphorylation and activation of WNK1. Early analysis of NKCCs postulated the existence of a chloride-sensitive kinase to explain the dual activation of these transporters by phosphorylation and low chloride. (19, 20, 23). More recently, it has been suggested that WNKs perform this role (47, 49). Our data validated both the concept of a chloride-sensitive kinase and that WNK1 is such a kinase. We also showed that chloride inhibited autophosphorylation with an observed IC₅₀ of approximately 20 mM. This is slightly lower than observed concentrations for chloride in the kidney (~50 mM), and slightly higher than those present in neurons (~10–15 mM) (3, 4). Thus, chloride appears to regulate WNK1 within a concentration milieu appropriate to the functions of distinct tissues. The results suggest that the chloride concentration in cells may be set in part by the chloride sensitivity of WNK isoforms or splice variants. Further data are required to test the chloride-based inhibition of other WNK isoforms. The observed chloride-binding site, comprised of a combination of main chain interactions and hydrophobic contacts, is typical of the first of the two classes of chloride-binding sites described above, which utilizes backbone-amide and hydrophobic interactions. The chloride-binding site is similar to that of the ClC-family chloride cotransporter (10). In ClC, the chloride ion caps the N-terminus of a helix, making both two 3.5 Ǻ hydrogen bonds and hydrophobic contacts with a phenylalanine, an isoleucine, and a tyrosine. The ANP receptor also binds chloride through backbone amides (at a conserved glycine-proline turn) and has a hydrophobic flank formed of a disulfide bond (11).

The structural basis for the chloride regulation of WNK1 makes WNKs unique among metabolite or ion-sensing kinases and can explain the special placement of the catalytic lysine in WNKs. Most metabolite-sensing kinases have separate domains or subunits that bind the ligand. Protein kinase A (PKA), for example, possesses cAMP-binding subunits, whereas calcium calmodulin-dependent kinase binds calcium through the calmodulin subunit (61, 62). In contrast, the WNK1 chloride-binding site is directly in the kinase domain (Fig. 4A). This design may contribute to the fast cellular responses to changes in chloride concentration (47). The location of the chloride-binding site overlaps with the canonical placement of catalytic lysines found in other kinases, such as Lys⁷² in PKA (Fig. 4B); the catalytic lysine (Lys²³³) in WNK1-KDm/SA (Fig. 4B) does not overlap, thus offering a possible reason for the unique position of the catalytic lysine in WNKs.

Fig. 4. WNK1 serves as an intracellular salt sensor.

(A) WNK1 autoactivation is inhibited by chloride. In low chloride conditions, WNK1 autophosphorylates and becomes active. Phosphatase activity inactivates WNK1, but it can reactivate itself if the chloride concentration remains low. (B) WNK1-KDm/SA (PDB: 3FPQ) and PKA (PDB: 1ATP) were aligned in Pymol (77). The placement of L371 in WNK1-KDm/SA (cyan) overlaps spatially with the canonical kinase Lysine (Lys 72 of PKA) (gold, PDB: 1ATP).

By mutagenesis, we validated the chloride-binding site, which is situated at the DLG motif and the N-terminus of the activation loop. The mutant L369F, which is in the DLG motif, was the least sensitive to the inhibition of autophosphorylation by chloride. Most protein kinases have the sequence DFG (subdomain VII) at this position. Our data suggest that this leucine residue may destabilize the active structure. The question then arises whether kinases that contain a leucine in that position are regulated by ions or adopt inactive configurations similar to that adopted by WNK1. The β-adrenergic receptor kinase (βARK), among others, possesses a DLG motif (63). βARK negatively regulates the β-adrenergic receptor, a stimulator of chloride flux, through phosphorylation (64). Thus, βARK might be similarly regulated.

The present data offer a structural rationale for the low chloride activation of NKCCs by WNK1. However, the mechanism of hypertonic stress activation of NKCCs remains unclear (24). The dual activation of NKCCs by low chloride and extracellular hypertonicity led to the hypothesis that there exists a kinase sensitive not only to chloride but also to volume (23). The activation of this kinase in response to decreased cellular volume is thought to be mediated by macromolecular crowding (65). WNK1 is activated by hypertonic stress induced by either high external salt or sorbitol, as well as by low concentrations of chloride (26, 32, 36, 50, 66). The effects of hypertonic conditions on WNKs could occur on the kinase directly or on another factor that regulates the kinase. Here, we observed a phosphorylated form of WNK1-KDm that did not adopt a completely active configuration. This suggests a possible mechanism by which crowding effects could compete with chloride to promote an autophosphorylation-competent conformation. The observation that autophosporylation only potentiates activation has been made in other protein kinases (27). How a fully active conformation is induced and how hypertonicity activates WNK1 require further investigation.

WNK1-KDm/S* was dephosphorylated by PP1cγ in vitro. The activities of NKCCs and other CCC family members are likely regulated by the balance of kinases and phosphatases targeting these proteins. Because WNK1 can autophosphorylate and can be dephosphorylated suggests that the cellular balance of phosphatase and kinase activities may contribute to WNK1 regulation. Thus, we propose that as the chloride concentration decreases, autophosphorylation outstrips dephosphorylation and WNK1 activity increases, thereby stimulating downstream signaling (Fig. 4A). As the chloride concentration increases, dephosphorylation outstrips autophosphorylation, and WNK1 activity is inhibited, decreasing downstream signaling. The cellular phosphatase acting on WNKs is unknown, but both protein phosphatase-1 and 4 have been implicated (67, 68). The discovery that WNK1 is controlled by autophosphorylation, together with a method to induce constitutive activity of WNK1, may prove useful in elucidating the chloride sensitivity and signaling pathways involving WNK1 and other WNK isoforms.

Materials and Methods

Expression and purification of WNK1 and WNK1 mutants

Cloning, expression, and purification of the WNK1 kinase domain constructs of 194–483 and 209–483 was as described in Min et al. (27). The expression and purification of p38α was described in (69), MEK6 in (70), and ASK1-KDm in (53). Mutants were created using a Stratagene QuikChange® Site-Directed Mutagenesis kit and purified as WNK1 194–483.

Kinase function assays using MBP

WNK1-KDm/S*, p38α, MEK6, or ASK-KDm, or the WNK1 mutants was added at a final concentration of 2 μM to 10 molar excess of MBP in 30 μl of kinase reaction buffer (30 mM HEPES pH 8.0, 5 mM MgCl₂, 1 mM DTT, 100 μM ATP) containing 10 μCi of [γ32P]-ATP. Increasing amounts of NaCl were added to the reaction mixture to make a series of final chloride concentrations in the range of 10–1000 mM. Reactions were incubated for 1 hour at 37 °C. Reactions were stopped by blotting 20 μL of each reaction on filter paper squares (1 cm × 1 cm) and plunging into a solution of 10% trichloroacetic acid (TCA). The squares were washed three times in TCA, blotted, and transferred to scintillation vials. The vials were filled with 5 ml Complete Counting Cocktail 3a70B (Research Products International Corp.) and counted on a Beckman LS 3801. Reactions were normalized to 100% highest signal and 0% for the sample lacking substrate (MBP). To confirm that all counts were the result of MBP phosphorylation, autophosphorylation was monitored by separation of the reaction mixture on an SDS-PAGE gel followed by scintillation counting of the band containing WNK1-KDm/S*. No autophosphorylation of the WNK1 kinase domain was detected in the MBP substrate assay. Mutants were assayed similarly.

Phosphorylation of OSR by WNK1-KDm/S*

WNK1-KDm/S* was added at a final concentration of 2 μM to 5 molar excess of OSR1 1–295 in 30 μl of kinase reaction buffer (30 mM HEPES pH 8.0, 5 mM MgCl₂, 1 mM DTT, 100 μM ATP) containing 10 μCi of [γ32P]-ATP. Increasing amounts of NaCl were added to the reaction mixture to make a series of final chloride concentrations in the range of 30–1000 mM. Reactions were incubated for 1 hour at 37 °C. Reactions were stopped by blotting 20 μL of each reaction on filter paper squares (1 cm × 1 cm) and plunging into a solution of 10% trichloroacetic acid (TCA). The squares were washed three times in TCA, blotted, and transferred to scintillation vials. The vials were filled with 5 ml Complete Counting Cocktail 3a70B (Research Products International Corp.) and counted on a Beckman LS 3801. Reactions were normalized to 100% highest signal and 0% for the sample lacking substrate (OSR1 1–295).

Kinase Autophosphorylation

Autophosphorylation was measured similarly, using WNK1-KDm/S, p38α, MEK6, or ASK-KDm, or the WNK1 mutants generated from exposure of the protein to the serine-phosphatase PP1Cγ conjugated with a hexahistidine tag (5 μM) for 30 minutes at 37 °C in a solution of 100 mM Tris pH 8.0 and 10 mM MnCl₂ in a total reaction volume of 4.5 ml. The reaction mixture was purified by passing over a Ni-agarose column (GE) to remove the phosphatase from kinase. The kinase was buffer exchanged into 50 mM Tris pH 8.0, 50 mM NaCl, and 1 mM DTT and assayed for phosphorylation by either ³²P incorporation followed by scintillation counting as described above or by mass spectrometry. No other substrates except the dephosphorylated kinase were added to this reaction mixture, and the final WNK1-KDm/S or WNK1 mutant concentration was 2.5 μM for ³²P incorporation or 55 μM for mass spectrometry. The final concentration of p38α, MEK6, or ASK-KDm was 2.5 μM for ³²P incorporation. This solution was allowed to react for 30 minutes at 21 °C.

Mass spectrometry

WNK1 autophosphorylation reactions were performed and stopped by the addition of guanidine to 1 M final concentration at the time points indicated. WNK1 was digested in an 8:1 molar ratio with sequence-grade chymotrypsin (Roche) in the presence of 100 mM Tris pH 8.0, and 25 mM CaCl₂ (100 μL total reaction volume) at 30 °C overnight. Following digestion, the peptide mixture was separated by HPLC (Agilent 1100) on a RP-C18 column (Phenomenex Aeris Widepore 150 × 2.1 mm) using an acetonitrile-water gradient from 4% to 28% with 0.2% formic acid. Mass spectrometric analysis was performed on an LCQ DECA XP ion-trap mass spectrometer (ThermoFinnigan) with the HPLC coupled inline to an orthogonal electrospray ionization source. MS detector responses were obtained by integration under ion traces corresponding to m/z ranges for activation loop peptides, then scaled to the integration of the total ion current for all eluted peptides. MS/MS spectra were acquired in a data dependent mode and analyzed using MASCOT software (Matrix Science Ltd.) (71, 72). Synthetic peptide standards (21st Century Biochemicals) were used to confirm HPLC elution times and for quantitation.

Differential scanning fluorimetry assays

5 μM of the kinase was incubated with 10 mM Tris pH 8.0, 5X SYPRO Orange, water, and the specified salt concentration to make a 20 μl reaction volume for a well on a 96-well clear-bottom plate. The 96-well plate was centrifuged at 805 × g to remove bubbles created by pipetting. The temperature in each well was cycled from an initial 4 °C to 80 °C using 0.5 °C increments in a BioRad CFX96 real-time PCR machine. The fluorescence intensity of SYPRO orange was probed using the fluorescein amidite channel of the PCR machine.

Crystallization and X-ray data collection

Crystallization of WNK1-KDm/SA (194–483 S382A) was performed with a variation on the condition used by Min et al (70). A mixture of 24% w/v PEG 2000 monomethyl ether (MME), 100 mM HEPES pH 8.0, and 300 mM NaBr was used to set up Nextal hanging drop vapor diffusion crystal trays (Qiagen). A 1 to 1 (2 μl total) ratio of protein to precipitant solution was used, at 7 mg/ml protein concentration. Crystals grew to 200 μm in one week at 16 °C. Five crystals from these plates were cryoprotected using a solution of 20% glycerol, 24% PEG 2000, MME, 100 mM HEPES (pH 8.0), 300 mM NaBr. Data were collected at the APS beamline 19-ID at the anomalous edge of bromine (0.92 Å). Integration and scaling were performed using the HKL2000 software suite (73).

Crystals of WNK1-sKDm/S* (209–483 S382*) were obtained by hanging drop vapor diffusion using a protein concentration of 8 mg/mL. Protein buffer was 50 mM NaCl, 50 mM HEPES 8.0, 1 mM DTT. A 1:1 ratio of protein and precipitant solution were mixed (2 μL each). Precipitant solution was 0.35 M KH₂PO₄ and 19% PEG 3350. Crystals were grown to a size of 200 μm at 16 °C. Crystals were cryoprotected in 0.35 M KH₂PO₄, 19% PEG 3350, and 30% PEG 400. 1.8 Ǻ resolution X-ray data were collected at the APS beamline 19-ID. Integration and scaling were performed using the HKL2000 software suite (73).

Structure determination and refinement

A 2.8 Ǻ resolution X-ray structure of WNK1-KDm/SA in the presence of NaBr was solved by molecular replacement utilizing the A subunit of the previously solved WNK1-KDm/SA [NaCl, PDB ID: 3FPQ, (27)] structure as a starting model with the MOLREP module of the CCP4 program suite (74). Anomalous difference Fourier maps generated using the CCP4 suite revealed the position of the bromide ion. Restrained refinement using REFMAC5 in the CCP4 suite for the WNK1-KDm/SA and including the bromide ion coordinates obtained above gave an initial R-factor of 0.22 and R_free of 0.30. Model building using COOT (75) and subsequent rounds of restrained refinement in combination with TLS (Translation-Libration-Screw) and refinement of the occupancy of the bromide ion in PHENIX crystallographic software (76) yielded an R-factor and R-free of 0.20 and 0.29, respectively. The bromide ion occupancy was 0.48 with a 5.5 σ peak. All figures were made using Pymol (77).

The |F_o − F_c| map revealed a 5.5 σ peak in the active site. The original refinement based on the idea that water occupied this site resulted in a lower B-factor (14 Ǻ²) than the surrounding water molecules (30 Ǻ² and 22 Ǻ²). Using the assumption that the peak corresponded to a chloride ion, the structure was rerefined, and the B-factor of the chloride ion rose to 28 Ǻ², in line with the surrounding structure.

To determine the structure of WNK1-sKDm/S*, molecular replacement was carried out employing PDB ID: 3FPQ as a search model in the MOLREP module of the CCP4 program suite. One cycle of restrained refinement using REFMAC5/CCP4 program produced an R-factor of 0.24 and an R_free of 0.30. Subsequent cycles of model building and restrained refinement in combination with TLS produced an R and R_free of 0.23 and 0.27, respectively. The final model contained 233 water molecules, a phosphate ion, and a chloride ion.

Supplementary Material

Supplementary Data

Figure S1: Raw trace of WNK1-KDm/S* DSF using Sypro Orange

Figure S2: Control kinase stability as a function of salt concentration

Figure S3: Phosphorylation of OSR1 by WNK1-KDm/S*

Figure S4: WNK1 autophosphorylation measured by mass spectrometry

Figure S5: Interaction of bound chloride with hydrophobic residues

Table S1: Data and refinement statistics for WNK1-sKDm/S*

Table S2: B-Factors of the 3/10 helix residues in WNK1-KDm/SA and WNK1-sKDm/S*

Table S3: Phosphorylation state of WNK1-KDm/S* mutants after purification from E. coli