WNK1, a novel mammalian serine/threonine protein kinase lacking the catalytic lysine in subdomain II.

We have cloned and characterized a novel mammalian serine/threonine protein kinase WNK1 (with no lysine (K)) from a rat brain cDNA library. WNK1 has 2126 amino acids and can be detected as a protein of approximately 230 kDa in various cell lines and rat tissues. WNK1 contains a small N-terminal domain followed by the kinase domain and a long C-terminal tail. The WNK1 kinase domain has the greatest similarity to the MEKK protein kinase family. However, overexpression of WNK1 in HEK293 cells exerts no detectable effect on the activity of known, co-transfected mitogen-activated protein kinases, suggesting that it belongs to a distinct pathway. WNK1 phosphorylates the exogenous substrate myelin basic protein as well as itself mostly on serine residues, confirming that it is a serine/threonine protein kinase. The demonstration of activity was striking because WNK1, and its homologs in other organisms lack the invariant catalytic lysine in subdomain II of protein kinases that is crucial for binding to ATP. A model of WNK1 using the structure of cAMP-dependent protein kinase suggests that lysine 233 in kinase subdomain I may provide this function. Mutation of this lysine residue to methionine eliminates WNK1 activity, consistent with the conclusion that it is required for catalysis. This distinct organization of catalytic residues indicates that WNK1 belongs to a novel family of serine/threonine protein kinases.

The protein kinase superfamily contains over a thousand members that share a catalytic core of approximately 300 residues organized in two domains (1)(2)(3). Conserved structural motifs within the core sequence maintain the basic fold of the catalytic domain, and fewer than 10 highly conserved residues create the functional elements of the active site (4,5). Prior to the solution of the three-dimensional structure of cAMP-dependent protein kinase (PKA) 1 by Knighton et al. (4), several of the residues essential for the integrity of the structure and the active site were identified primarily by a combination of multiple sequence alignment (1,6), chemical modifications (7), and alanine scanning mutagenesis (8). Among these a lysine residue near the N terminus of the kinase in protein kinase subdomain II (Lys 72 in PKA); this residue has frequently been mutated to eliminate the catalytic activity of protein kinases (9). This lysine functions to anchor and orient ATP through interactions with the ␣ and ␤ phosphoryl groups (4,5,10). Until recently all members of the protein kinase family were found to contain a lysine following a short string of hydrophobic residues in this conserved position. Kinase suppressor of Ras (KSR) contains arginine in place of this lysine but has not yet been shown to catalyze phosphorylation of protein substrates (11,12). Mutation of this arginine in KSR does impair its function in reconstitution assays, suggesting that it plays a significant role in KSR function (11). Structural analysis of the mitogen-activated protein (MAP) kinase ERK2 shows that substitution of the conserved lysine with arginine causes the phosphoryl groups of ATP to be rotated away from the position necessary for phosphoryl transfer (10). Thus, the function of arginine in this conserved position of KSR is uncertain.
We have identified a novel protein kinase WNK1 (with no lysine (K)), which contains cysteine in place of lysine at the usual conserved location but has kinase activity as deduced from its ability to autophosphorylate and to phosphorylate an exogenous substrate in vitro. We have investigated the basis for its catalytic activity using a structural model, mutagenesis, and protein expression.
WNK1 was isolated in a nested PCR cloning strategy aimed to identify novel members of the MAP/extracellular signalregulated protein kinase (ERK) kinase (MEK) family. MAP kinases are a family of protein kinases that have been utilized to varying degrees to regulate or modulate almost all signal transduction pathways in cells (13,14). These enzymes themselves are regulated by cascades of at least two upstream protein kinases, a MEK and a MEK kinase. Members of the MEK (or MKK) family display considerable selectivity for their particular MAP kinase targets, thereby contributing to signaling specificity (13). They activate MAP kinases by dual phosphorylation on a tyrosine and a threonine residue, and each MEK recognizes only a small subset of possible MAP kinase substrates.
After the purification and cloning of the first MEK family member, MEK1 (15,16), others (MKK2, MKK3, MKK4, MEK5, MKK6, and MKK7) were discovered through low stringency or PCR screens rather than by purification (17)(18)(19)(20)(21)(22)(23)(24)(25). Not only did this approach streamline the definition of the components of * This work was supported by National Institutes of Health Grants DK34128 and GM53032. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The known MAP kinase cascades, but it also uncovered new pathways. We continued to examine clones derived from the screen that led us to isolate cDNAs encoding MEK5 (22). As described in this report, one of these clones encoded the unusual protein kinase WNK1.

MATERIALS AND METHODS
Isolation of cDNA Clones Encoding WNK1-First strand cDNA isolated from nerve growth factor stimulated PC12 cells was used as the template in PCR reactions utilizing nested degenerate primers derived from MEK sequences (22). One PCR product (product 15) was used to screen a rat forebrain cDNA library (kindly provided by Jim Boulter) at low stringency, and a weakly hybridizing clone of approximately 900 base pairs (PC12 clone 2-3) was isolated that had a short region of identity to the PCR product. This PC12 clone 2-3 was then used to rescreen the rat forebrain cDNA library at high stringency, and a group of strongly hybridizing clones were isolated. Partial sequences that encoded the N terminus and the kinase domain of WNK1 were assembled from two of these clones. A 0.5-kb WNK1 3Ј probe was labeled with [␣-32 P]dCTP by random-priming (Amersham Pharmacia Biotech) and used to screen another rat brain cDNA library that contains longer inserts (also kindly provided by Jim Boulter). One of the clones isolated contained the complete 3Ј WNK1 sequence. The full-length WNK1 Two potential coiled-coil regions and two proline-rich regions are also shown. There are total of 24 potential SH3 domain binding motifs (PXXP) in WNK1 represented by black lines in the diagram. In addition, the two proline-rich regions contain 3 and 11 PXXP motifs, respectively. C, sequence alignment of the kinase domain of WNK1 and its orthologs. The roman numerals shown above the sequences indicate subdomains. The accession numbers for the sequences from top to bottom are WNK1 (AF227741), AJ242724, Z68296, AF080436, AL049659, and Z46636. cDNA was assembled from these clones.
Northern Blot Analysis-A rat adult multi-tissue Northern blot (CLONTECH) was hybridized with a random-primed (Amersham Pharmacia Biotech) 0.5-kb WNK1 5Ј probe according to the manufacturer's suggestions. The same blot was stripped and reprobed with a 2.8-kb WNK1 3Ј probe. This Northern blot was stripped again and hybridized with a ␤-actin probe to confirm the presence of mRNA in each lane.
Plasmids, Mutagenesis, and Proteins-A pSK-WNK1 full-length construct was created from three overlapping cDNA clones and was used as the template for the subsequent subcloning. A 1.6-kb WNK1 fragment encoding residues 1-555 was amplified by PCR with an EcoRI site incorporated at the 5Ј end and a HindIII site incorporated at the 3Ј end, and this fragment was digested with EcoRI-HindIII and ligated into pGEX-KG or pCMV5-Myc vectors that had been digested with EcoRI and HindIII to create pGEX-KG-WNK1 (1-555) and pCMV5-Myc-WNK1 (1-555). To make a pGEX-KG-WNK1 full-length construct, pGEX-KG-WNK1 (1-555) was digested with HindIII, filled-in with Klenow, and then digested with SacII; pSK-WNK1 was digested with SpeI, filled-in with Klenow, and then digested with SacII. The 6.5-kb insert fragment from pSK-WNK1 was gel purified and ligated into the 5.1-kb vector fragment from pGEX-KG-WNK1 (1-555). To generate a pCMV5-Myc-WNK1 full-length construct, pSK-WNK1 was digested with SacII-SpeI and the 6.5-kb insert fragment was gel purified and ligated into the 4.8-kb SacII-XbaI digested pCMV5-Myc-WNK1 (1-555) vector backbone. All constructs and mutants were sequenced to confirm that the sequences were correct. pCEP4HA-ERK2, pSR␣-HA-JNK1, pCEP4HA-p38␣, and pCEP4HA-ERK5 were as described (26,27).
Site-directed mutagenesis was carried out using the Quikchange kit (Stratagene) according to the manufacturer's recommendation. The WNK1 mutants used in this study include K233M, K256M, K259M, C250A, C250K, D368A, and the double mutant S378D/S382D.
GST-WNK1 (full-length and 1-555) proteins were expressed in and purified from Escherichia coli strain BL21DE3 using the standard protocol (28). The induction conditions were: 40 M (for 1-555) or 400 M (for full-length) isopropyl-␤-D-thiogalactopyranoside at 30°C for 5 h. The protein concentration was estimated by comparing to serial dilutions of bovine serum albumin on the same gel stained with Coomassie Blue. Myelin basic protein (MBP) was purchased from Calbiochem. GST-c-Jun and GST-ATF2⌬ were as described (29).
Monoclonal anti-HA antibody (12CA5) was obtained from Berkeley Antibody Company. Monoclonal anti-Myc antibody (9E10) was obtained from the Cell Culture Center. For immunoblotting, cell or tissue lysates were separated by SDS-polyacrylamide gel electrophoresis and electrophoretically transferred onto nitrocellulose paper. Blots were then developed using enhanced chemiluminescence.
Proteins were immunoprecipitated from 0.2 ml of cell lysates with 2 l of antibody and 40 l of protein A-Sepharose beads. Precipitates were washed three times with 20 mM Tris-HCl (pH 7.4), 1 M NaCl and once with 10 mM Hepes (pH 8.0) and 10 mM MgCl 2 .
Transfection, Preparation of Extracts, and Cell Fractionation-HEK 293 cells were maintained, transfected, and harvested as described (31). For cell fractionation, cells from each 60-mm dish were washed once with 1ϫ phosphate-buffered saline, scraped, and resuspended in 200 l of buffer A (10 mM Hepes, pH 7.6, 1.5 mM MgCl 2 , 10 mM NaCl, 1 mM EDTA, 1 mM EGTA, and protease inhibitors 1 mM dithiothreitol, 1 g/ml leupeptin, 10 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride). Cells were lysed with a Dounce apparatus, and the nuclei were collected by sedimentation at 4,000 rpm for 5 min in a microcentrifuge at 4°C. A particulate fraction was collected by sedimenting the supernatant at 55,000 rpm for 30 min in a TL-100 ultracentrifuge at 4°C. The nuclear pellet was washed with 200 l of buffer C (20 mM Hepes, pH 7.6, 2.5% glycerol, 0.42 M NaCl, 1.5 mM MgCl 2 , 1 mM EDTA, and 1 mM EGTA) at 4°C for 30 min. Both the washed nuclear pellet and the particulate fraction were resuspended in 200 l of lysis buffer (10 mM Tris-HCl, pH 7.6, 100 mM NaCl, 1% SDS, 1 mM EDTA, 1 mM EGTA, and protease inhibitors 1 mM dithiothreitol, 1 g/ml leupeptin, 10 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride).
Other Methods-Immunofluorescence was essentially as described (32). HEK 293 cells were grown on coverslips coated with collagen, fixed with formaldehyde, and treated with anti-Myc antibody. Cells were washed and treated with goat anti-mouse fluorescein secondary antibody followed by diamidinophenylindole staining. Coverslips were mounted onto slides, and cells were examined under a fluorescence microscope. In vitro kinase assays and phosphoamino acid analysis were performed as described (33). A WNK1 model was made using PKA as a template. An alignment of the kinase domains of WNK1 and PKA was generated. InsightII (Molecular Biosystems) was used to change residues 50 -100 of the structure of PKA to the corresponding residues found in WNK1 and create a ribbon diagram.

RESULTS AND DISCUSSION
Isolation of the Rat Full-length cDNA Encoding WNK1-In an attempt to isolate novel mammalian MEKs, nested degenerate PCR primers designed based on sequences conserved among MEK family members were used to amplify products from first strand cDNA isolated from PC12 cells (22). One product of 150 base pairs was used to probe a rat forebrain cDNA library. A clone isolated from a low stringency screen was used to screen two rat brain cDNA libraries and several positive clones encoding a novel protein kinase named WNK1  Ϫ Peptide, the antibody was not exposed to the antigenic peptide; ϩ Peptide, the antibody was preincubated with 50 g/ml of antigenic peptide overnight to demonstrate antibody specificity.
were isolated. The full-length WNK1 cDNA containing 7.2 kb was assembled from three overlapping clones. The sequence surrounding the ATG start codon matched the Kozak consensus sequence for translation initiation, and stop codons were present upstream in all three reading frames. Although there was no poly(A) track found downstream of the stop codon in the available WNK1 sequence, there is a polyadenylation signal sequence (AATAAA) at the end of the cDNA clone. The open reading frame encoded by the WNK1 cDNA contains 2126 amino acids with a serine/threonine protein kinase domain in the N-terminal 490 residues (Fig. 1, A and B). A partial clone encoding a closely related kinase (WNK2) was also isolated.
Homologs of WNK1 exist in Caenorhabditis elegans, Phycomyces, Arabidopsis, and Oryza as well as other mammals (Fig.  1C). A human expressed sequence tag containing an open reading frame lacking the kinase domain is almost identical to the WNK1 C-terminal sequence, and two human open reading frames containing partial kinase domains show strong similarity to the WNK1 catalytic domain, indicating that they may encode parts of human WNK family members. WNK1 and its homologs are characterized by the absence of the catalytic lysine found in subdomain II of almost all known protein kinases (1). In WNK1, a cysteine lies in the position of the usual lysine. WNK1 shares several sequence features with the MEK  5. Kinase activity of WNK1. A, 293 cells were transfected with either pCMV5-Myc without an insert or pCMV5-Myc-WNK1, tagged proteins were immunoprecipitated with the anti-Myc antibody followed by kinase assays using MBP as the substrate. Autoradiography is shown on the left, and an immunoblot of the immunoprecipitates is shown on the right. B, either wild type GST-WNK1 or D368A GST-WNK1 was mixed with MBP in an in vitro kinase assay. Autoradiography is shown on the left. Phosphoamino acid analysis of autophosphorylated GST-WNK1 and phosphorylated MBP is shown on the right. C, proteins were immunoprecipitated from 293 lysates with preimmune serum or with antibodies Q255, Q256, or Y691 and subjected to in vitro kinase assay. Autophosphorylation of endogenous WNK1 is shown. D, endogenous WNK1 was immunoprecipitated from 293 cells treated with various stimuli and assayed for autophosphorylation. Background phosphorylation of MBP was too high to be a reliable measure of the activity of the endogenous protein. One of four similar experiments is shown.
family, including the length of the activation loop (between subdomain VII and VIII), the position of potential activating phosphorylation sites (SFAKS; these sites are also conserved in the IB kinases IKK1 and 2 (34, 35)), conservation in the region of protein substrate binding, and a pattern of conserved hydrophobic residues near the C terminus. However, the kinase domain of WNK1 does not clearly fall into any kinase subgroup but shows the greatest similarity to MEKK-like kinases (ϳ30% identity) and, to a lesser extent, to Ste20p-like kinases. The N-terminal and C-terminal noncatalytic domains of WNK1 are not similar to any other proteins in the data base.
Expression of WNK1 in Rat Tissues and Mammalian Cell Lines-To examine the expression of WNK1 in various rat tissues, a rat multi-tissue Northern blot was hybridized with a 0.5-kb WNK1 5Ј probe. Two transcripts of 11 and 9.5 kb were detected in several tissues including lung, liver, and spleen (Fig. 2). The existence of two different species of mRNA suggests that an alternative splicing event might be involved in WNK1 mRNA processing. Similar results were obtained using a 3Ј 2.8-kb probe (data not shown). A polyclonal antibody raised against a WNK1 N-terminal peptide recognized a protein of ϳ230 kDa corresponding to the predicted size of WNK1 in rat brain and several mammalian cell lines including 293, COS-1, and INS-1 cells (Fig. 3), suggesting that it is widely expressed.
Localization of WNK1-To determine whether WNK1 is soluble or associated with nuclear or particulate fractions, HEK293 cells were fractionated and proteins in each fraction were separated by SDS-polyacrylamide gel electrophoresis and blotted with an anti-WNK1 antibody. The majority of endogenous WNK1 protein was found in the particulate fraction, suggesting that WNK1 is associated with either membranes or the cytoskeleton (Fig. 4A). To examine its cellular localization of further, a Myc-tagged WNK1 construct was transfected into 293 cells, and the expressed protein was visualized by immunofluorescence with an anti-Myc antibody. Most staining was observed outside of the nucleus, suggesting that WNK1 is cytoplasmic protein not restricted to the plasma membrane (Fig. 4B).
WNK1 Is a Serine/Threonine Protein Kinase-To examine the kinase activity of WNK1, Myc-tagged WNK1 was transfected in 293 cells and then immunoprecipitated from lysed cells with an anti-Myc antibody. The immunoprecipitate showed kinase activity toward MBP as well as itself (Fig. 5A). Although activity toward MBP was detected, autophosphorylation was more consistent than MBP phosphorylation in the immune complex kinase assay because of the high background contributed by other protein kinases. Wild type GST-WNK1 expressed in E. coli phosphorylated both itself and MBP. Mutation of the putative Mg 2ϩ binding residue aspartate 368 to alanine generated a WNK1 protein with no detectable kinase activity. Phosphoamino acid analysis revealed that phosphorylation occurs mainly on serine residues, indicating that WNK1 is a serine/threonine kinase (Fig. 5B).
To test the endogenous activity of WNK1, the endogenous protein was immunoprecipitated with anti-WNK1 antibodies and assayed by its ability to autophosphorylate. Immunoblotting confirmed that WNK1 was immunoprecipitated (not shown). The ability of WNK1 to autophosphorylate was apparent from the incorporation of labeled phosphate into a band of approximately 230 kDa. In contrast, neither preimmune serum nor the unrelated anti-ERK1 antibody immunoprecipitated an autophosphorylating band of this size (Fig. 5C). To identify regulators of WNK1, we tested a number of agents and stimuli to determine whether they could increase WNK1 activity in 293 cells. Endogenous WNK1 was immunoprecipitated following cell treatment, and autophosphorylation was assayed. Among the stimuli tested, 0.5 M NaCl (Fig. 5D) and less so 0.5 M sorbitol (not shown) caused a reproducible increase in WNK1 autophosphorylation, suggesting that WNK1 may be involved in osmosensing pathways. No effects were detected with epidermal growth factor, the microtubule disrupting agent nocodazole, anisomycin, lysophosphatidic acid (Fig. 5D), serum, heat shock, phorbol ester, H 2 O 2 , or okadaic acid (not shown).
The Known MAP Kinase Pathways Are Not Activated by Overexpression of WNK1 in 293 Cells-Because WNK1 was isolated as a possible MEK homolog and has modest similarity to the MEKK-like and Ste20p-like kinases within its catalytic domain, we examined the potential regulation of MAP kinase pathways by WNK1. The WNK1 kinase domain (residues 1-555) was expressed in mammalian cells and the activities of the known MAP kinases, HA-tagged forms of ERK2, ERK5, JNK1, or p38, were measured. Myc-tagged WNK1 (1-555) constructs, either wild type or kinase dead (D368A), were used in the majority of experiments because expression was to a much greater extent than for full-length WNK1. HA-tagged proteins were then immunoprecipitated with an anti-HA antibody and assayed using MBP, c-Jun, or ATF-2 as substrates. No obvious changes in activity of any of these kinases were observed with overexpression of either wild type or kinase-dead WNK1 (1-555) (Fig. 6) or with full-length WNK1 (not shown). In vitro, GST-WNK1 displayed neither MEK nor MEKK activity. It failed to phosphorylate recombinant ERK1, ERK2, ERK5, JNK1, or p38. WNK1 also did not phosphorylate MEK1, MKK2, MKK3, MKK4, MEK5, MKK6, or MKK7; nor did it phosphorylate IB, ribosomal protein S6, or fragments of MEKK1 (not shown). In addition, the two possible phosphorylation sites of WNK1 (SFAKS) that lie in the same relative positions as the activating sites of phosphorylation in the MEK family were mutated to aspartic acid (S378D/S382D). Although comparable mutations increase the activity of some MEK family members (e.g. MEK1), there was no detectable effect of these mutations on WNK1 activity (not shown). In summary, these results suggest that WNK1 does not directly regulate the MAP kinase pathways tested. WT, wild type. One of four experiments is shown. B, partial sequence alignment between PKA and WNK1. WNK1 residues that were mutated are underlined. Identical residues between the two enzymes are shown in between the two sequences. Roman numerals indicate the kinase subdomains. WNK1 Lacks the Conserved Catalytic Lysine Residue in Subdomain II-Several residues involved in catalysis of phosphoryl transfer are highly conserved in all protein kinases, notably the catalytic lysine residue present in kinase subdomain II, which binds to ATP. Surprisingly, this apparently invariant lysine residue is replaced by a cysteine (Cys 250 ) in WNK1. The lack of lysine at this position was confirmed in multiple independent clones of rat WNK1. More striking, this unusual difference is conserved across diverse species, suggesting functional relevance. Two possible explanations for this deviation in WNK1 were either that the catalytic lysine was located at a different position in the structure or that the WNK1 catalytic mechanism was distinct from other protein kinases. To distinguish between these possibilities, we first created a structural model of WNK1 based on the coordinates of PKA (Fig. 7). Based on this model, several candidate lysine residues that might potentially function in ATP binding, Lys 233 , Lys 256 , and Lys 259 , were mutated to methionine. In addition, Cys 250 was mutated to either alanine or lysine to determine whether it was required for catalysis. These mutants were expressed in bacteria as GST fusion proteins and assayed in vitro using MBP as substrate (Fig. 8). WNK1 C250K had greatly reduced kinase activity, suggesting that Cys 250 may play some catalytic role. However, WNK1 C250A had the same activity as wild type protein. This result demonstrated that Cys 250 is not required for WNK1 activity, but a lysine that has a larger and positively charged side chain may interfere with the folding or activity of the catalytic site. WNK1 K256M and K259M exhibited kinase activity similar to the wild type protein, indicating that Lys 256 and Lys 259 are not required for catalytic activity. In contrast, WNK1 K233M had no detectable kinase activity, indicating that Lys 233 , like Asp 368 , plays a critical role in WNK1 activity. This lysine residue, Lys 233 , is conserved in position in all the WNK homologs, consistent with its importance for catalytic activity. Interestingly, Lys 233 replaces a glycine residue in the glycine string of subdomain I that comprises the phosphate anchor ribbon as shown in Fig. 7. Mutation of this residue in PKA has relatively little impact on its kinetic properties (36), consistent with the idea that glycine, although it is usually present, is not required at this position for kinase activity. From this location in the primary sequence, the side chain of Lys 233 can apparently fill the position normally occupied by the lysine residue present in other protein kinases that corresponds to Cys 250 in WNK1 (Fig. 7).
In summary, we have isolated and characterized a protein kinase represented in diverse organisms with a previously unknown placement of a key catalytic residue. Mutagenesis studies indicates that a lysine, Lys 233 , in the glycine ribbon is required for activity of WNK1. All other residues conserved among the protein kinases are in the expected locations in WNK1. The impaired activity of a mutant in which the putative magnesium binding residue, aspartate 368, was replaced with alanine supports the conclusion that the active site is otherwise typical of protein kinases. Modeling experiments suggest that the active site structure can be preserved despite this significant sequence deviation. Interestingly, a key lysine present in GTPases and other nucleotidases often lies quite near the glycine string (37). Perhaps the altered organization of the catalytic residues in WNK1-like kinases reflects a yet-to-bediscovered regulatory or functional adaptation better served by this particular sequence arrangement.