Characterization of OSR1, a member of the mammalian Ste20p/germinal center kinase subfamily.

In examining the protein kinase components of mitogen-activated protein (MAP) kinase (MAPK) cascades that regulate the c-Jun N-terminal kinase (JNK) in Drosophila S2 cells, we previously found that distinct upstream kinases were involved in responses to sorbitol and lipopolysaccharide. Here we have extended that analysis to the possible MAPK kinase kinase kinases (MAP4Ks) in the JNK pathway. Fray, a putative Drosophila MAP4K, provided a major contribution to JNK activation by sorbitol. To explore the possible link to JNK in mammalian cells, we isolated and characterized OSR1 (oxidative stress-responsive 1), one of two human Fray homologs. OSR1 is a 58-kDa protein of 527 amino acids that is widely expressed in mammalian tissues and cell lines. Of potential regulators surveyed, endogenous OSR1 is activated only by osmotic stresses, notably sorbitol and to a lesser extent NaCl. However, OSR1 did not increase the activity of coexpressed JNK, nor did it activate three other MAPKs, p38, ERK2, and ERK5. A two-hybrid screen implicated another Ste20p family member, the p21-activated protein kinase PAK1, as an OSR1 target. OSR1 phosphorylated threonine 84 in the N-terminal regulatory domain of PAK1. Replacement of threonine 84 with glutamate reduced the activation of PAK1 by an active form of the small G protein Cdc42, suggesting that phosphorylation by OSR1 modulates the G protein sensitivity of PAK isoforms.

Ste20p is the yeast MAP4K protein that activates the MAP3K in the pheromone-responsive MAPK cascade of the budding yeast mating pathway (5,6). In the past several years, numerous protein kinases with catalytic domains closely related to that of Ste20p have been identified and constitute the Ste20p family. Based on structure and regulation, two subfamilies have been defined, the p21-activated kinase (PAK) subfamily and the larger germinal center kinase (GCK) subfamily (7). The PAKs include the six enzymes nearest in characteristics to Ste20p itself; each contains a C-terminal catalytic domain and an N-terminal regulatory domain with a small G protein binding motif. PAKs have been shown not only to activate MAPKs (primarily JNK and p38) but also to influence disassembly of the actin cytoskeleton and apoptosis (8 -14). The GCKs are distinct from PAKs in that they have N-terminal catalytic domains followed by C-terminal putative regulatory regions without conserved G protein binding sites. The 28 known human GCK-related kinases are classified in eight subdivisions and have diverse and much less well characterized functions (7). Some, like Ste20p and PAKs, regulate the JNK and p38 MAPK pathways. Those reported to activate JNK include the GCK-IV subfamily members, MINK, NIK, HGK, TNIK; GCK-I subfamily members, GCK, HPK1, GLK; and the GCK-V subfamily member, SLK (15)(16)(17)(18)(19)(20)(21)(22)(23)(24). Those reported to activate p38 include the GCK-VI subfamily member SPAK and the GCK-VIII subfamily members TAO1 and TAO2 (25)(26)(27). However, some have no apparent connection to known MAPK pathways. These include the GCK-II subfamily member MST1, the GCK-III subfamily members MST3 and MST4, the GCK-V subfamily member LOK, and the GCK-III subfamily member SOK-1 (28 -32). Similar to PAKs, some GCKs have been reported to regulate F-actin structure, cell spreading, and apoptosis (21,24,(33)(34)(35)(36). Among novel functions that have been found, SPAK is reported to regulate the Na-K-Cl cotransporter (NKCC1), and a role in cell cycle control has been inferred for Stk10, which is a novel polo-like kinase (PLK) kinase (37)(38)(39).
We previously examined the MAPK cascade components used by two agents to stimulate JNK in Drosophila S2 cells (40). To extend these studies here, we have examined the potential involvement of putative MAP4Ks in regulating Drosophila JNK. We found that when the Ste20p relative Fray was knocked down using RNA interference, JNK activity stimulated by sorbitol decreased markedly, whereas ablation of other putative MAP4Ks decreased JNK activity little (CG4527) or not at all. These results suggested that Fray was the major MAP4K regulating the JNK pathway in response to sorbitol in S2 cells. We then wished to determine whether mammalian Fray homologs were MAP4Ks upstream of JNK. The kinases most closely related to Fray in the human genome are OSR1 (oxidant stress-responsive protein 1) and SPAK (Ste20/SPS1related, proline alanine-rich kinase). SPAK has been reported to activate p38 but not JNK (25). Human OSR1 had been isolated but not characterized. Here, we report the characterization of human OSR1 and an initial analysis of its biochemical functions.

MATERIALS AND METHODS
Cloning, Subcloning, Mutagenesis, and Plasmids-Total RNA prepared from HeLa cells was subjected to RT-PCR with a pair of primers spanning the complete human OSR1 cDNA synthesized based on the sequence in the NCBI data base. The 1.6-kb RT-PCR products were cloned into the GST-tagged bacterial expression vector pGEX-KG. This plasmid was used as the template for subsequent subcloning. Fulllength OSR1 cDNA was also subcloned into p3XFLAG-CMV and pCMV5-Myc. Fragments encoding OSR1-(1-433), OSR1-(1-344), OSR1-(1-291), and OSR1-(345-527) were amplified by PCR and subcloned into pGEX-KG, pRSET (His 6 tag), pCMV5-Myc, and pVJL11 as indicated. Kinase-dead mutants of OSR1 (OSR1KR) and fragments were generated by mutating lysine 46 in the ATP binding pocket to arginine. All constructs were transformed into the bacterial strain TG-1 and grown at 30°C to reduce the frequency of mutations. All clones were sequenced to confirm correct amplification.
The HA antibody (12CA5) was from Berkeley Antibody, and the anti-Myc antibody (9E10) was from the National Cell Culture Center, and both were used at a dilution of 1:1000 for immunoblotting. The monoclonal anti-FLAG antibody was from Sigma and was used at 1:4000. The anti-Lamin A/C antibody was from Santa Cruz Biotechnology and was used at 1:1000. The polyclonal anti-OSR1 serum (U5438) was raised against His 6 -OSR1-(345-527) and was used at 1:8000.
Fractionation and Immunofluorescence-Cell fractionation was as described (41). Immunofluorescence was as described (46) with the following changes. HeLa cells were grown to confluence and starved in medium with 0.5% FBS overnight. Cells on coverslips were fixed in 2% paraformaldehyde for 10 min, permeabilized in cold methanol at Ϫ20°C for 10 min, and then incubated with anti-OSR1 U5438 antibody (1:800). After incubation with anti-rabbit secondary antibody (Alexa, 1:3000), OSR1 localization was observed using a Zeiss Axioskop 2 plus fluorescent microscope.
Preparation of Tissue and Cell Lysates and Immunoblotting-Cultured cells or tissues from a 13-month-old mouse (provided by David Russell, Department of Molecular Genetics) were homogenized and lysed in Triton X-100 lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 0.5% Triton X-100, 0.5 mM sodium orthovanadate, 20 g/ml aprotinin, 10 g/ml pepstatin A, 10 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride). Insoluble material was sedimented in a microcentrifuge for 15 min at 4°C. Protein concentration was measured by Bradford assay using bovine serum albumin as standard. Thirty g of soluble protein from each sample was resolved by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5% non-fat milk (40) for 1 h at room temperature and then incubated with the appropriate antibody.
Immunoprecipitation, in Vitro Kinase Assays, and Phosphoamino Acid Analysis-Lysate protein (300 g) was incubated with the indicated antibody for 1 h at 4°C and then with 30 l of a 50% slurry of protein A-Sepharose beads for 1 h. After three washes with detergent buffer (0.25 M Tris, pH 7.4, 1 M NaCl, 0.1% Triton X-100, 0.1% sodium deoxycholate) and one with 10 mM HEPES (pH 7.6), beads were incubated with indicated substrates in 50 l of 1ϫ kinase buffer (20 mM HEPES, pH 7.6, 5 M ATP (5 Ci of [␥-32 P]ATP), 10 mM MgCl 2 , 10 mM ␤-glycerol phosphate) at 30°C for 30 min for kinase assays. Purified proteins were incubated with indicated substrates in 30 l of 1ϫ kinase buffer at 30°C for 30 min. One-dimensional phosphoamino acid analysis was performed as described (47).
Yeast Two-hybrid Analysis-A neonatal mouse brain cDNA library (gift from Mark Henkemeyer, Center for Developmental Biology) in plasmid pGADGH was screened as described. 2

RESULTS AND DISCUSSION
MAP4Ks and Activation of JNK in S2 Cells-In an earlier study, we examined the components of the protein kinase cascades that control JNK activity in response to lipopolysaccharide and sorbitol in Drosophila S2 cells (40). Both agents used both of the MAP2Ks MEK4 and MEK7 to activate JNK. Although lipopolysaccharide required a single MAP3K (DTAK), sorbitol employed four MAP3Ks to stimulate JNK activity. We have completed the examination of the likely kinase components of the MAPK module by testing the potential involvement of putative MAP4Ks in regulating Drosophila JNK. Based on sequence alignments and a consideration of the published classification of the fly protein kinases (48), six putative MAP4Ks, CG11228, DPAK, DPAK3, DMSN, CG4527, and Fray, were found to be expressed in S2 cells as determined by PCR analysis (data not shown). The expression of each of these was reduced using RNA interference, and the effect of the loss of each singly on activation of JNK by sorbitol was then examined ( Fig. 1, A and B; data not shown). When expression of Fray, a Drosophila GCK-VI kinase family member, was knocked down, JNK activity stimulated by sorbitol decreased significantly. Reduction in expression of one of the other putative MAP4Ks, CG4527, decreased sorbitol-stimulated JNK activation to a small but reproducible extent. Reducing expression of CG4527 caused a further reduction in the residual JNK activation remaining upon suppression of Fray (Fig. 1C). These results suggested that Fray was the major MAP4K regulating the JNK pathway in response to osmotic stress in S2 cells. Interestingly, Fray is required for normal axonal ensheathment during fly development (49).
Structure and Expression of OSR1-We wished to learn whether mammalian homologs of Fray, SPAK and OSR1, also regulate the JNK pathway. SPAK has been reported to activate p38 but not JNK (25). Thus, we focused on analyzing OSR1, which has not been characterized previously as a protein. Human OSR1 contains 527 amino acids with a predicted molecular mass of 58 kDa. We isolated a cDNA encoding OSR1 by RT-PCR using HeLa RNA. It has a conserved Ste20p-like protein-serine/threonine kinase domain at its N terminus with a C-terminal region of undefined function. The OSR1 kinase domain has the highest identity to the kinase domain of SPAK (89%), Drosophila Fray (74%), and the Caenorhabditis elegans Y59A8B.23 gene product (71%). These four enzymes comprise the GCK-VI subfamily of Ste20p protein kinases. Two small regions of similarity were found in the C-terminal regions of GCK-VI subfamily members, which were named PF1 and PF2 domains. They may represent regulatory or targeting elements (25). A putative caspase 3 cleavage site (DEFD) is present at the end of the PF1 domain of OSR1 (Fig. 2).
To detect the expression of OSR1, a rabbit polyclonal antibody was generated against the C-terminal, poorly conserved region. The specificity of the anti-OSR1 antibody was confirmed by immunoblotting and immunoprecipitating overexpressed Myc-OSR1 in HEK-293 cells (data not shown). With this antiserum, a 58-kDa protein was recognized in all mouse tissues examined except thymus, including heart, spleen, liver, kidney, lung, testis, large intestine, small intestine, and stomach. OSR1 was also detected in mammalian cell lines including HEK 293, HeLa, PC3, BT20, HI299, SW480, 2721, and Cos-1, which were derived from a variety of tissues including kidney, cervix, ovary, prostate, breast, lung, and colon (Fig. 3A). Interestingly, although the same amount of protein was analyzed from each lysate, little or no OSR1 was detected in the pancreatic beta cell line INS-1 or the mouse fibroblast lines C2C12 or 3T3, suggesting some tissue specificity to its expression. The wide expression of OSR1 is consistent with the broad transcription of OSR1 mRNA as deduced from Northern blotting (50).
To detect the subcellular localization of endogenous OSR1, proteins from soluble, particulate, and nuclear fractions derived from HeLa cells were immunoblotted with anti-OSR1 and anti-lamin A/C antibodies. OSR1 was detected in all three fractions, whereas lamin A/C was detected only in the nuclear fraction (Fig. 3B). Endogenous OSR1 detected by immunofluo- rescence was distributed throughout HeLa cells, consistent with the fractionation data (Fig. 3C).
Protein Kinase Activity of OSR1-To verify that OSR1 is a serine/threonine protein kinase, recombinant wild type GST-OSR1 and the kinase inactive mutant GST-OSR1KR were expressed in bacteria and assayed with MBP as substrate. Wild type GST-OSR1 phosphorylated both MBP and itself, but GST-OSR1KR showed no detectable activity (Fig. 4A). Phosphoamino acid analysis of autophosphorylated GST-OSR1 revealed primarily phosphothreonine (Fig. 4B). Equal amounts of Myc-tagged OSR1 and OSR1KR expressed in HEK 293 cells were immunoprecipitated with the anti-Myc antibody and assayed with MBP as substrate (data not shown). A phosphorylated band at 58 kDa, representing autophosphorylated Myc-OSR1, and phosphorylated MBP were detected in assays with the wild type OSR1 immunoprecipitate. However, the same bands were also detected in assays with OSR1KR. Similar experiments performed with different epitope tags (3XFLAG) and different cell types (Cos-1 and HeLa) yielded similar results (data not shown). Because OSR1KR had no activity when expressed in bacteria and because MBP is phosphorylated by many abundant protein kinases, we conclude that the phosphorylation comes from contaminating proteins co-purifying with or non-specifically trapped in OSR1 in immune complexes; some of these may be kinases that normally phosphorylate OSR1. As a consequence, MBP was not a suitable substrate to measure OSR1 activity in cell lysates or immunoprecipitates, although it was useful to characterize the activity of the protein expressed in bacteria.
The PF1 Domain Is Necessary for OSR1 Kinase Activity-To examine the contribution of the C-terminal region to the kinase activity of OSR1, truncated forms of OSR1 were expressed as GST fusions in bacteria, and the purified proteins were assayed with MBP as substrate. Full-length OSR1, OSR1-(1-433) and OSR1-(1-344) have nearly identical kinase activity toward MBP or themselves. In contrast, OSR1-(1-291), a truncated protein with intact, conserved kinase domain but without the PF1 domain, has no detectable activity toward MBP or itself (Fig. 5). Thus, the PF1 domain is essential for OSR1 kinase activity. Many Ste20p-related kinases contain autoinhibitory domains; removal of the regulatory domains results in a significant increase in kinase activity due to loss of autoinhibition. This has been shown for PAKs, MST1, MST2, TAOs, and SOK1, for example (10,26,32,51,52). In the case of OSR1 and other GCK-VI kinases, the PF1 domain may comprise an essential part of the kinase catalytic domain. Because there is no significant difference in the kinase activity of full-length OSR1 and OSR1-(1-433), the PF2 domain is apparently not involved in regulating catalytic activity.
Cellular Stimuli That Activate OSR1-To identify possible regulators of OSR1, a number of stimuli were tested on HeLa cells. Autophosphorylation was used to assess activity of endogenous, immunoprecipitated OSR1. Treatment with 0.5 M sorbitol for 30 min caused an obvious increase in OSR1 autophosphorylation. OSR1 was activated in the range of 0.5-0.7 M, with activation detectable from 15 to 60 min of treatment. Higher concentrations of sorbitol resulted in reduced activation (Fig. 6A). In contrast to strong activation of OSR1 by sorbitol, modest or no activation was caused by NaCl, anisomycin, okadaic acid, serum, nocodazole, Taxol, H 2 O 2 , phorbol ester, and epidermal growth factor (see below and data not shown). A time course of activation by NaCl showed that the greatest effect was at 15 min of NaCl treatment (Fig. 6B); weaker activation was detected at 2, 5, 10, and 30 min. The effect of sorbitol on OSR1 activity was consistently greater than that of NaCl under all conditions examined.
OSR1 Does Not Activate Four Known MAPKs-To determine whether OSR1, like Fray, is an upstream regulator of JNK or other MAPKs, Myc-OSR1, OSR1KR, or the fragments OSR1- (1-433) or -(1-344) were co-expressed with HA-MAPKs, JNK, p38, ERK5, or ERK2 in HEK 293 cells. The HA-MAPKs were immunoprecipitated and assayed with GST-c-Jun, GST-MEF2C, or MBP as indicated. Neither JNK nor any of the other MAPKs were activated by coexpression with wild type or truncated OSR1 (Fig. 7). In addition, reduction of OSR1 expression by more than 75% with dsRNA oligonucleotides did not decrease JNK activation by sorbitol in HeLa cells (data not shown). Thus, it appears that OSR1 participates in distinct stress pathways from those controlling MAPKs. Alternatively, OSR1 may activate JNK or other MAPK pathways under conditions or in the presence of accessory proteins that we did not identify here.
The isolated N terminus of PAK1-(1-231) inhibits the activity of the PAK1 catalytic domain in vitro. Known activating mutations of PAK1, H83L/H86L and L107F, are located in the autoinhibitory domain. The inhibitory activity of GST-PAK1-(1-231) H83L/H86L and L107F mutants decreased markedly as compared with the activity of the wild type fragment (10,54,55). Because Thr-84 is also located in this region, we tested the idea that phosphorylation on this site may influence the autoinhibitory activity of the PAK1 N terminus. The inhibitory activity of GST-PAK1-(1-231) T84E, a possible phosphomimic of phosphorylated Thr-84, was measured. Results showed no difference in the inhibitory activity of wild type and T84E PAK1-(1-231) (Fig. 10A). PAK1-(1-231) previously phosphorylated by OSR1 to a stoichiometry of 0.4 mol of phosphate/ mol inhibited the activity of PAK1 catalytic domain, as well as did unphosphorylated PAK1-(1-231) (data not shown). To determine whether the activity of PAK1 was influenced by OSR1 in cells, 3XFLAG-OSR1 and Myc-PAK1 were coexpressed. PAK1 was immunoprecipitated with the anti-Myc antibody and assayed with MBP as substrate. However, no change in PAK1 kinase activity was detected (data not shown).
The autoinhibitory and small G protein binding domains of PAK1 overlap, and Thr-84 is in the region of overlap. Thus, the activities of PAK1, PAK1T84E, and PAK1T84A were compared in the absence (Fig. 10B) and presence (Fig. 10C) of increasing concentrations of the activated Cdc42 mutant G12V. PAK1 proteins were immunoprecipitated, and assayed with a PAK1selective substrate, the proline-rich insert domain, residues 265-301, from MEK1. PAK1H83L/H86L was little further activated by Cdc42 G12V, consistent with its previously described behavior (data not shown) (10). Little or no difference in the activities of PAK1 and the Thr-84 mutants was detected without Cdc42 G12V (Fig. 10B). On the other hand, the activity of PAK1T84E coexpressed with Cdc42 G12V was less sensitive to the small G protein. At the highest amount of Cdc42 G12V tested, the activity of PAK1T84E ranged from 20 to 60% of wild type PAK1. With expression of less Cdc42 G12V, the activity of the T84E mutant was consistently 20 -30% of wild type PAK1 (Fig. 10C). In five experiments, activation of PAK1 T84A was not significantly less than wild type PAK1, although at the lower amount of Cdc42 G12V, a small reduction in activity was usually observed. These findings support the conclusion that phosphorylation of PAK1 by OSR1 desensitizes PAK1 to activation by small G proteins, providing a modulatory input to PAK1 activity. This may serve to slow disassembly of the cytoskeleton under conditions of osmotic stress.
In summary, we have characterized a novel Ste20p family kinase OSR1. OSR1 requires the accessory PF1 domain for activity. This domain is present only in the four GCK-VI members of the Ste20p family; thus, these enzymes apparently have a regulatory mechanism markedly different from many other Ste20p family members. In contrast to our expectations based on studies in S2 cells, OSR1 did not activate mammalian JNK or other MAPKs. It is, nevertheless, in a stress-sensitive pathway, because it is activated by sorbitol and NaCl. Interestingly, OSR1 phosphorylates another Ste20p PAK1 on a single site within the PAK1 regulatory domain. This residue is present in FIG. 10. Effects of OSR1 phosphorylation on PAK1 activity. As shown in A, GST-PAK1-(232-544) (1 mg) was assayed with MBP as substrate, either alone or in combination with increasing amounts of GST-PAK1-(1-231) wild type (WT) or the mutants T84E, H83L/H86L, or L107F (top panel). Coomassie staining of GST-PAK1-(1-231) and mutants is shown (bottom panel). As shown in B, cells were transfected with plasmids (3 mg) encoding Myc tag vector (V), wild type PAK1, or mutants T84E, T84A, H83L/H86L, or L107F. Overexpressed proteins were immunoprecipitated with anti-Myc antibody and assayed with 2 g of GST-MEK1 insert as substrate (top panel). Immunoblotting of the immune complexes confirmed equal amounts of PAK1 and mutant proteins in each lane (bottom panel). C, as in B, except the indicated PAK1 constructs were coexpressed with the indicated amounts of Cdc42 G12V. The activation of PAK1 proteins relative to background by Cdc42 G12V is indicated below the autoradiogram of the kinase assay. PAKs 1, -2, and -3, the three PAK family members that contain conserved small G protein binding domains, suggesting that OSR1 has the capacity to modulate the activation of all three of these PAKs. We have found that the C terminus of OSR1 binds to gelsolin in a yeast two-hybrid assay. 3 Gelsolin is a well characterized actin-binding protein. It binds to growing ends of actin filaments, nucleates actin assembly, and severs actin filaments. Because OSR1 can phosphorylate PAK1 and bind to gelsolin, we suggest that OSR1 may also have a function in regulating the actin cytoskeleton.