Cloning and Characterization of a Human STE20-like Protein Kinase with Unusual Cofactor Requirements*

We cloned and characterized a novel human member of the STE20 serine/threonine protein kinase family named mst-3. Based on its domain structure, mst-3 belongs to the SPS1 subgroup of STE20-like proteins, which includes germinal center (GC) kinase, hematopoietic progenitor kinase (HPK), kinase homologous to STE20/SPS-1 (KHS), kinases responsive to stress (KRS1/2), the mammalian STE20-like kinases (mst1/2), and the recently published STE20/oxidant stress response kinase SOK-1. mst-3 is most closely related to SOK-1, with 88% amino acid similarity in the kinase domain. The similarity of the mst-3 kinase domain to STE20 is 42%. The mst-3 transcript is ubiquitously expressed, and the protein was found in all human, mouse, and monkey cell lines tested. An in vitro kinase assay showed that mst-3 can phosphorylate basic exogenous substrates as well as itself. Interestingly, mst-3 prefers Mn2+ to Mg2+ as a divalent cation and can use both GTP and ATP as phosphate donors. Like SOK-1, mst-3 is activated by autophosphorylation. However, a physiological stimulus of mst-3 activity was not identified. mst-3 activity does not change upon exposure to several mitogenic and stress stimuli. Overexpression of mst-3 wild-type or kinase dead protein affects neither the extracellular signal-regulated kinases (ERK1/2 or ERK6), c-Jun N-terminal kinase (JNK), p38, nor pp70S6 kinase, suggesting that mst-3 is part of a novel signaling pathway.

Eukaryotic cells are able to couple extracellular signals to specific biological processes such as cell growth, differentiation, and stress responses through the activation of distinct evolutionarily conserved intracellular signaling cascades, collectively known as mitogen-activated protein kinase (MAPK) 1 cascades. In both mammals and lower eukaryotes, the core of MAP kinase cascades is a three-component module consisting of a generic MAPK kinase kinase, which phosphorylates and activates a dual-specificity MAPK kinase, which in turn activates the MAPK. In many cases the activated MAP kinases translocate to the nucleus where they phosphorylate transcription factors, thus eliciting the biological response (1).
In the budding yeast Saccharomyces cerevisiae, at least six MAPK pathways have been identified. They regulate diverse biological processes such as mating and invasive growth, cell wall integrity, and the response to high osmolarity as well as pseudohyphal development and spore formation in diploid cells (2)(3)(4)(5). As in yeast, several mammalian MAPK pathways have been identified (6,7). The best characterized pathway is the mitogenic signaling pathway, which leads to the activation of the extracellular signal-regulated kinases 1/2 (ERK1/2) through both growth factor and G protein-coupled receptors (3,8). ERKs 1/2 are activated by the MAPK kinases, MEK1/2 (9,10). On the level of the MAPK kinase kinase, MEK1 can be activated by several kinases, among them Raf (11), Mos (12), and to a lesser extent mitogen-activated protein kinase kinase kinase 1 (13). Other mammalian MAPK pathways can be activated by a variety of stress agents like tumor necrosis factor ␣, interleukin-1, UV light, and osmotic shock and lead to the activation of the MAPKs c-Jun N-terminal kinase (JNK) and p38 (14 -16). Like the ERKs, JNK and p38 are activated by distinct upstream kinases. MAP kinase kinase 4 (MKK4) (17), also called stress-activated protein kinase kinase 1 (SEK) (18), activates both JNK and p38 (17), whereas MKK3 and MKK6 specifically activate p38 (17, 19 -22).
First, like STE20 and Cla-4, the mammalian PAKs consist of a C-terminal kinase domain and an N-terminal regulatory domain and have a small GTPase Rac1/Cdc42 binding region. PAKs bind to GTP-Cdc42 and Rac but not Rho (28 -31). Upon binding, they autophosphorylate and are activated (28 -31).
Second, all other STE20-like kinases identified thus far re-* This work was supported in part by Public Health Service Research Grant CA46595 from the National Institutes of Health (to J. B). 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 Coexpression studies have demonstrated that some mammalian STE20 homologues are able to activate mammalian MAPK pathways. Activated PAKs can specifically activate the JNK and p38 pathways but not the ERK pathway. GCK, HPK, and KHS can also activate JNK but not p38 or the ERKs (35,36,39). These kinases do not contain the Rac1/Cdc42-binding domain, and JNK activation is thus presumed to be Rac1/Cdc42-independent.
In this paper we describe the cloning and initial characterization of a human STE20 homologue (mst-3) with unusual kinase activity characteristics. mst-3 belongs to the Sps1 subfamily of STE20-like kinases with an N-terminal kinase domain and a C-terminal regulatory domain. mst-3 mRNA and protein is ubiquitously expressed. mst-3 kinase prefers manganese to magnesium as a co-factor and can use both ATP and GTP as phosphate donors. mst-3 has high basal activity and does not activate any of the known mammalian MAPK pathways. In vitro, mst-3 kinase activity is positively regulated by autophosphorylation. Immunolocalization analysis demonstrates that mst-3 is predominantly localized in the cytoplasm. . These fragments were subcloned into pBluescript (Promega) and sequenced using the dideoxy chain termination method with Sequenase 2.0 (U. S. Biochemical Corp.). Sequence was analyzed using DNASTAR. Data base searches were done using BLAST. Several clones encoded kinase domain fragments of novel kinases related to the yeast STE20 protein. A STE20like clone was used to screen 10 6 plaques of a human oligo(dT)-primed T cell library (courtesy of Dr. F. McKeon). Four plaques were purified, subcloned into pBluescript, and partially sequenced (Applied Biosystems Inc.). One clone contained full-length mst-3 but no stop codon upstream of the putative start site, and both strands were sequenced. A 5Ј probe was generated and used to screen 10 6 plaques of a randomprimed human HeLa cell library (courtesy of Dr. R. Reed). Three plaques with identical sequence were purified and isolated; one contained stop codons in all three reading frames upstream of the putative start codon.
Plasmids-Full-length mst-3, cloned into pBluescript, was used as a template for polymerase chain reaction reactions to create in-frame constructs for further subcloning. pJ3H (courtesy of Dr. J. Chernoff) was used to create hemagglutinin (HA)-tagged mst-3 (with an N-terminal stuffer of 26 amino acids). The HA-tagged mst-3 was excised from pJ3H and cloned into pcDNA for mammalian expression. pGEX4T (Pharmacia Biotech Inc.) was used to create a glutathione S-transferase fusion to full-length mst-3 for bacterial expression. A kinase-dead mst-3 was created by changing lysine 53 to arginine using site-directed polymerase chain reaction-based mutagenesis.
Northern Blot Analysis-Hybridized poly(A ϩ ) RNAs from human tissues were obtained (CLONTECH, courtesy of L. Cantley). A probe was generated encompassing nucleotides 1-350, labeled with [ 32 P]dCTP, and spin column-purified (Pharmacia). The probe was used to screen the RNA blot according to manufacturer's instructions.
mst-3 Kinase Assays-Assays for mst-3 kinase activity were performed as described previously (1). However, unless indicated otherwise, kinase assays were performed in the presence of 5 or 10 mM MnCl 2 for 3-5 min at 22-24°C to ensure linear assay conditions in a 30-l volume with 4 g of histone H3 (Boehringer Mannheim) as a substrate. The reaction was stopped by the addition of 30 l of 2 ϫ sample buffer and analyzed by SDS-PAGE (14%). The gel was then fixed and autoradiographed. 32 P incorporation was quantitated using a PhosphorImager with software from Molecular Dynamics.
Biosynthetic Labeling-For [ 35 S]methionine/cysteine biosynthetic labeling, subconfluent COS cells were labeled overnight with 1.5 mCi [ 35 S]methionine/cysteine (NEN Life Science Products) per 10-cm tissue culture dish in 4 ml of methionine/cysteine-free medium (ICN Biomedicals, Inc.). Cells were lysed in 0.8 ml of lysis buffer as described before, and lysates were incubated with 10 ml of ␣mst1, preimmune serum, or antiserum preadsorbed with an excess of 2 g of antigen for 1 h at 4°C followed by incubation with protein A-Sepharose 4b beads (Sigma) for 1 h at 4°C. After centrifugation, pellets were washed in buffer A, B, and ST, resuspended in 500 l of radioimmune precipitation buffer, and immunoprecipitated and washed as described as before. The immunoprecipitated proteins were separated on SDS-PAGE (10%), the gel was fixed, and the precipitated proteins were visualized by autoradiography.
Cell Culture and Transfections-All cell types used (HeLa, A431, 293, NIH3T3, Swiss3T3, COS) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Only PC12 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% horse serum and 5% fetal bovine serum. For transfections, COS cells were seeded at a concentration of 4 ϫ 10 5 cells per 6-cm dish 1 day before transfection and then transfected for 4 h using the DEAEdextran method (47). A total of 10 g of DNA was used per 6-cm plate. NIH3T3 cells were seeded at 8 ϫ 10 4 cells per 35-mm dish and transfected the following day with lipofectAMINE (Life Technologies, Inc.) according to manufacturer's instructions.
Immunofluorescence-NIH3T3 cells were seeded on coverslips placed in 35-mm tissue culture dishes and transfected as described above. Cells were washed in PBS, fixed (3.7% formaldehyde, 10 mM Tris-Cl (pH 7.2), 0.1 M NaCl) for 10 min, permeabilized with 0.2% Triton X-100 in PBS for 5 min, blocked in 2% bovine serum albumin in PBS for 10 min, and stained with ␣HA monoclonal antibody (1:1,000 dilution) for 1 h. Slides were then washed in PBS, 0.2% Triton. The secondary antibody, rhodamine-conjugated goat anti-mouse immunoglobulins, was used at a 1:600 dilution for 30 min. Cells were washed with PBS, 0.2% Triton X-100, incubated with Hoechst dye (1 mg/ml), and washed again. Slides were analyzed using a Zeiss microscope.

RESULTS
Cloning and Sequence of mst-3-Degenerate oligonucleotides corresponding to conserved regions in the kinase domains of yeast STE20 and p65PAK were used to amplify a 200-nucleotide fragment from a HeLa cell cDNA library. This fragment was then used as a probe to screen a T cell library (courtesy of Dr. F. McKeon). A clone encoding what appeared to be a full-length cDNA was isolated and contained a perfect Kozak consensus sequence (GCCATGG) but no stop codon upstream of the putative start codon. Therefore, a 5Ј mst-3 mst-3, a Mn 2ϩ -regulated STE20-like Protein Kinase probe (encompassing nucleotides 1-350) was generated and used to screen a random-primed HeLa cell library. One clone was isolated that contained a stop codon upstream of the putative start codon and no other candidate initiation start codon in between. However, no putative polyadenylation site was identified (Fig. 1).
Amino Acid Sequence Analysis and Similarity to Other Kinases-The 2.0-kilobase cDNA contains an open reading frame of 431 amino acids with a predicted molecular mass of 48 kDa (Fig. 1). mst-3 contains all subdomains of serine/threonine kinases (40). Using the BLAST program to determine sequence homologies to other kinases, mst-3 was identified as a member of the STE20 family of protein kinases. mst-3 consists of an N-terminal kinase domain and a 142-amino acid C-terminal regulatory domain. The overall structure of mst-3 more closely resembles the SPS-1 group of the STE20 family of protein kinases (see Fig. 2C). mst-3 is highly related to the recently published SOK-1 kinase (38), with 88% identity in the kinase domain and 68.8% overall amino acid identity. The mst-3 kinase domain is related to the kinase domains of mst-1/KRS1 (37,48), mst-2/KRS2 (41,48), GCK (33), HPK (35), SPS1 (27), hPAK65 (30), and STE20 (23,24) with 56, 56, 48, 44, 48, 45, 42% amino acid similarities, respectively (Fig. 2B). The mst-3 C-terminal regulatory domain does not contain a Cdc42/Rac binding region and does not have any identifiable sequence motifs other than several acidic regions.
Analysis of mRNA and Protein Expression-Northern blot analysis using poly(A ϩ ) RNA from multiple human tissues demonstrated that the mst-3 transcript is ubiquitously ex-pressed (Fig. 3). Under high stringency conditions, using a unique region upstream of the kinase domain as a probe, a single 2-kilobase mRNA was detected that is expressed at highest levels in heart, skeletal muscle, and pancreas. To confirm equal loading of mRNAs, the blot was re-probed using a human ␤-actin probe (data not shown).
The expression of mst-3 protein was examined using Western blot analysis. A polyclonal antiserum (␣mst-3) raised against a glutathione S-transferase fusion of the full-length mst-3 protein recognized a protein of about 52 kDa in all human (Hela, C2C12, 293, A431), mouse (NIH3T3), and monkey cells (COS) tested (Fig. 4A). The increased apparent molecular mass could be due to posttranslational modifications. Using either preimmune serum or antibody preadsorbed with antigen abolished the signal, demonstrating the specificity of the ␣mst-3 serum (data not shown). An HA-tagged recombinant mst-3 (HAmst-3) that contained a 26-amino acid extension at the N terminus could be detected as a slower migrating band on Western blots when transfected into COS cells (Fig. 4A).
␣mst-3 also immunoprecipitated a single protein of the same size from 35 S-labeled COS cell lysates under denaturing conditions (Fig. 4B, lane 2) that was absent from immunoprecipitates performed using identical conditions but with either preimmune serum (Fig. 4B, lane 1) or antigen-preadsorbed ␣mst-3 antiserum (Fig. 4B, lane 3) for the immunoprecipitation.
Catalytic Activity of mst-3-␣mst-3 was used to immunopreciptate endogenous mst-3 from COS lysates to develop an in vitro kinase assay. No phosphotransferase activity above the mst-3, a Mn 2ϩ -regulated STE20-like Protein Kinase control preimmune precipitates was detected using histones H1, H2, H4, casein, and phosvitin as substrates (data not shown). However, mst-3 readily phosphorylated MBP and histone H3 at levels significantly above control (Fig. 5A). A phosphoprotein migrating at the size of mst-3 was also detected, suggesting that mst-3 autophosphorylates (Fig. 5A, lanes 2 and  4). Although the presence of a protein in the immunocomplex that phosphorylates mst-3 cannot be excluded, it is unlikely since the phosphoprotein was still present after immunoprecipitates were washed in buffer C containing 0.1% SDS and absent in immunoprecipitates from HA-tagged kinase-dead mst-3 (HAmst-3-KR)-transfected COS cell lysates (Fig. 5C,  lanes 6 and 8). Substrate phosphorylation was abolished when ␣mst-3 was preadsorbed with antigen before the immunoprecipitation (Fig. 5B, lane 9).
The intracellular concentration of Mn 2ϩ is in the M range, whereas the Mg 2ϩ concentration is in the mM range (42). To determine whether Mn 2ϩ is able to activate mst-3 at physiological concentrations, mst-3 activity was tested using a range of Mg 2ϩ and Mn 2ϩ concentrations (Fig. 5, D and E). Endogenous mst-3 was immunoprecipitated from HeLa cells, and in vitro kinase assays were performed using the indicated Mn 2ϩ and Mg 2ϩ concentrations. Mn 2ϩ at M concentrations could activate mst-3 to higher levels than Mg 2ϩ at mM concentrations (compare mst-3 activity at 100 M Mn 2ϩ to 10 mM Mg 2ϩ ). The minimum Mn 2ϩ concentration at which a small increase in mst-3 activity was seen, was 50 M (Fig. 5D). Optimal activation was achieved at 1 mM. Above 5 mM, Mn 2ϩ was inhibiting, and mst-3 activity decreased. Mg 2ϩ did not activate mst-3 at M concentrations (Fig. 5D). Optimal Mg 2ϩ -regulated mst-3 activity was observed with 10 mM Mg 2ϩ , and 20 mM was not inhibiting (Fig. 5E). These results suggested that Mn 2ϩ may play a role in mst-3 regulation under physiological conditions. Phosphoamino acid analysis of in vitro phosphorylated MBP and histone H3 revealed that mst-3 is a serine/threonine protein kinase. Mst-3 phosphorylated MBP on Ser and Thr and histone H3 on Thr. Mst-3 also autophosphorylated on threonine (Fig. 5F).
A small number of protein kinases utilize GTP as well as ATP as a phosphate donor. To determine the mst-3 phosphate donor specificity, increasing amounts of unlabeled GTP were used to compete with ␥-32 P-labeled ATP in the phosphotransferase assay. Cold GTP successfully competes with ATP (Fig.  6A). To further confirm the specificity of GTP as a phosphate donor, in vitro kinase assays were performed in the presence of either [␥-32 P]ATP or [␥-32 P]GTP (Fig. 6, B and C). Endogenous mst-3 was immunoprecipitated with ␣mst-3 antiserum. COS cells transfected with either HAmst-3wt or HAmst-3KR were immunoprecipitated with polyclonal ␣HA antibody. In vitro kinase assays were performed in the presence of equal amounts of [␥-32 P]GTP or [␥-32 P]ATP at identical specific activities. mst-3 could use GTP and ATP as phosphate donors; however, ATP was preferred (Fig. 6, B and C).
Autophosphorylation and Activation of mst-3-As described above (Fig. 5, A and C), a phosphoprotein of the size of mst-3 appeared after in vitro kinase assays and was presumed to be autophosphorylated mst-3. To determine whether mst-3 autophosphorylation could lead to activation, mst-3 activity toward MBP was tested after a 20-min preincubation period with ATP and either Mn 2ϩ or Mg 2ϩ . In the presence of Mn 2ϩ but not Mg 2ϩ , this consistently resulted in a 3-6-fold increase in kinase activity compared with samples that were not preincubated (Fig. 7) and a marked shift in electrophoretic mobility (data not shown). Similar results were obtained using histone H3 as a substrate (data not shown), suggesting that although mst-3 had high basal activity, it was not constitutively active.
Mst-3 activity was also elevated 1.5-3-fold by incubating cells with calyculin A, a potent inhibitor of type 1c and 2a serine/threonine phosphatases but not by incubation with vanadate, a general tyrosine phosphatase inhibitor (Fig. 7B), suggesting that mst-3 is only partially regulated by phosphorylation.
Regulation of mst-3 Kinase Activity-To determine the position of mst-3 in signaling pathways, we coexpressed HAmst-3wt and HAmst-3KR with several MAP kinases (ERK-1, JNK, p38, ERK-6) and pp70-S6 kinase. In unstimulated versus stimulated cells, neither HAmst-3wt nor HAmst-3KR activated or inhibited ERK-1, JNK, p38, ERK-6, and pp70-S6 kinase. 2 Furthermore, we tested numerous agonists representing different classes of stimuli for their ability to activate or inhibit mst-3 activity in various cell types (Table I). In view of the high basal activity of mst-3, all time course assays were done in parallel with assays to ensure linear kinase assay conditions. No activating or inhibitory stimuli were found. Surprisingly, inducers of oxidative stress (H 2 O 2 and menadione), which have been shown to activate the closely related SOK kinase (38), did not activate mst-3. 2 K. Schinkmann, unpublished information.

FIG. 5. Characterization of mst-3 kinase activity.
A, endogenous mst-3 was immunoprecipitated from COS cells with either preimmune serum (lanes 1 and 3) or immune serum (lanes 2 and 4) and subjected to immunocomplex kinase assays in the presence of the indicated substrates and manganese. B, COS cells were transfected with HAmst-3-wt or HAmst-3-KR. The lysates were immunoprecipitated with either preimmune serum (lanes 1, 3, 5, and 7) or ␣HA antiserum (lanes 2, 4, 6, and 8). COS lysates from untransfected cells were immunoprecipitated with either preimmune serum (lanes 10 and 12), ␣mst-3 antiserum (lanes 11 and 13), or ␣mst-3 antiserum preadsorbed with antigen (lane 9). Immunocomplex kinase assays were performed in the presence of either manganese (5 mM) or magnesium (10 mM) and 1 M ATP using histone H3 as a substrate. The reactions were subjected to SDS-PAGE and visualized and quantitated using a PhosphorImager. C, the identical HAmst-3-transfected COS lysates as in B, depicting the putative HAmst-3 autophosphorylation activity. D and E, concentration-dependent activation of mst-3 kinase activity using either manganese or magnesium at M (D) or mM (E) concentrations. Endogenous mst-3 was immunoprecipitated from COS cells using ␣mst-3 antiserum, and immunocomplex kinase assays were performed under different concentrations of divalent cations as indicated using histone H3 as a substrate. After separation on SDS-PAGE, the activities were visualized and quantitated as above. All results shown in A-E are representative of at least three independent experiments. F, phosphoamino acid analysis of in vitro phosphorylated histone H3, MBP, and the putative mst-3 autophosphorylation. The relative positions of phosphoamino acids are indicated on the left. kD, kilodalton.

Subcellular Distribution of mst-3-
To determine the subcellular distribution of mst-3, NIH3T3 fibroblasts were transiently transfected with HAmst-3wt, and the expressed protein was visualized using ␣HA monoconal antibody. Nuclei were visualized with Hoechst (Fig. 8, B and D). HAmst-3wt was localized predominantly in the cytoplasm (Fig. 8A). No staining was visible in the control untransfected cells (for an example see Fig. 8A) or HAmst-3-transfected cells stained with secondary antibody only (Fig. 8C). DISCUSSION We have cloned and characterized a novel human member of the growing family of STE20-related protein kinases. Pending the identification of a physiological function, this kinase was named mst-3. Based on its structure, mst-3 belongs to the SPS1-like subfamily of STE20-like kinases with an N-terminal catalytic and a C-terminal regulatory domain. mst-3 is most closely related to the recently identified SOK-1 kinase (38) with amino acid similarities of 88 and 68.8% in the kinase domain and overall protein sequence, respectively. However, mst-3 is not regulated in the same fashion as SOK-1 and exhibits distinct kinase activity characteristics.
An in vitro kinase assay using HA-tagged recombinant as well as endogenous mst-3 was developed and revealed that mst-3 possesses unusual cofactor requirements. The ability to phosphorylate itself and exogenous substrates is consistently 20 -50-fold higher in the presence of Mn 2ϩ versus Mg 2ϩ (Fig. 5,  B and C). This property is mostly associated with the kinase activity of tyrosine kinases. At the M ATP concentrations used in in vitro kinase assays, Mn 2ϩ is also the most potent divalent cation activator of autophosphorylation for the tyrosine kinases of the insulin (43) and epidermal growth factor (44) receptors. Interestingly, a 54-kDa serine protein kinase associated with Varicella-Zoster virus ORF47 also exhibits a preference for Mn 2ϩ (45).

mst-3, a Mn 2ϩ -regulated STE20-like Protein Kinase
To examine the role of Mn 2ϩ versus Mg 2ϩ , we studied the effect of different Mn 2ϩ and Mg 2ϩ concentrations on mst-3 activation. Mn 2ϩ within the concentration interval of 0.05-5 mM stimulates mst-3 activity (Fig. 5D), whereas higher concentrations inhibit it (Fig. 5E). At a concentration of 0.5 mM Mn 2ϩ and 50 M ATP, 97-98% of the ATP is present as MnATP, and increasing the Mn 2ϩ concentration does not lead to a significant MnATP increase (43). The increase of mst-3 activity observed in response to increasing Mn 2ϩ concentrations ( Fig. 5D and E) thus suggests that besides formation of the substrate MnATP, binding of free Mn 2ϩ to a distinct site on mst-3 may be required for full kinase activation. The inhibitory effect of Mn 2ϩ concentration above 5 mM (Fig. 5E) might be due to an inhibitory feedback mechanism.
Under physiological conditions, the concentration of Mn 2ϩ is in the M range, about 3 orders of magnitude below the intracellular Mg 2ϩ concentration, which, like the ATP levels, lies in the mM range. The fact that the Mn 2ϩ concentration required for mst-3 activation was about 2 orders of magnitude lower than the Mg 2ϩ concentration at the M ATP levels used in our assays combined with the observation that Mn 2ϩ could activate mst-3 to much higher levels than Mg 2ϩ suggested that Mn 2ϩ is a physiological regulator of mst-3. However, the physiological role of Mn 2ϩ in mst-3 activation is unclear.
Subdomains VI and VIII of protein kinases contain residues that are conserved among members of either the serine/threonine kinases or the tyrosine kinases (40). The most striking indicator of amino acid specificity lies in subdomain VI. Tyrosine kinases contain either DLRAAN (src family) and DLAARN (all others), whereas serine/threonine kinases contain DLKPEN. In mst-3, this indicator region resembles more the tyrosine kinases (DIKAAN), whereas the region in subdomain VIII classifies it as a serine/threonine kinase GTPFW-MAPE (40). Phosphoamino acid analysis demonstrates that mst-3 is a serine/threonine kinase (Fig. 5F).
To place mst-3 into one of the existing signaling pathways or a novel pathway, we attempted to stimulate or inhibit its activity using several mitogenic and stress stimuli (Table I). Due to the high basal activity of immunoprecipitated endogenous and recombinant mst-3, immunocomplex kinase assays performed under standard kinase assay conditions would not be in the linear range and would thus mask activation. Therefore, all stimulation experiments were performed under optimized kinase assay conditions. However, no consistent stimulus of mst-3 activity was identified. Mst-3 activity does not change upon growth arrest due to serum deprivation or growth promoting agents such as serum, lysophosphatidic acid, epidermal growth factor, platelet-derived growth factor, and phorbol 12-myristate 13-acetate. The stress stimuli anisomycin, sorbitol, staurosporine, and UV exposure did not increase mst-3 activity in the indicated cell lines (Table I). Unexpectedly, mst-3 could not be further activated by oxidative stress due to H 2 O 2 or menadione, which have been reported to activate the closely related SOK-1 kinase (38). One explanation for this discrepancy is that mst-3 and SOK-1 have distinct biochemical properties, that Mn 2ϩ is not a cofactor for SOK-1, and that despite their closely related sequence, mst-3 and SOK-1 are regulated differentially and are part of separate signaling pathways.
mst-3 activity does not change upon raising the intracellular Ca 2ϩ or cAMP concentrations. We also tested serine/threonine and tyrosine phosphatase inhibitors on mst-3 activity, achieving partial activation of mst-3 activity with calyculin A, suggesting that full mst-3 activation might also be regulated through other mechanisms, i.e. dimerization and phosphorylation. Preliminary data indicated that mst-3 can dimerize (data not shown). In this respect mst-3 resembles mst-1, which has been shown to dimerize (46). mst-3 wild-type or kinase-inactive protein expression did not activate or inhibit any of the four MAPK cascades (ERK-1, JNK, p38, ERK-6) nor pp70-S6 kinase in coexpression experiments (data not shown), suggesting that mst-3 is part of a novel signaling pathway. Alternatively, mst-3 activity might be necessary but not sufficient for the activation of the pathways tested.
mst-3 is not a constitutively active kinase, since immunoprecipitated mst-3 autophosphorylates and, like SOK-1 and PAK, is activated by autophosphorylation. Depending on cell type and the substrate used, a 20-min incubation in the presence of ATP without substrate followed by incubation with substrate results in a 2-8-fold increase in mst-3 activity. Thus, even though mst-3 possesses high basal activity, it is not a constitutively active kinase. At this point the physiological role of mst-3 is unknown as is the significance of its similarity to STE20. Elucidation of these critical problems will have to await the identification of physiological substrates and effectors.
Acknowledgments-We thank members of the Blenis laboratory for helpful discussions and R. Tung for help with some experiments. We   FIG. 8. Subcellular localization of mst-3. HA-tagged mst-3 cDNA was transfected into HeLa cells, and HAmst-3 was visualized using ␣HA as a primary antibody followed by cy3 (goat anti-mouse conjugated rhodamine) as a secondary antibody (A) or cy3 only (C) as described under "Experimental Procedures." Nuclei were visualized using Hoechst staining (B and D). HAmst-3 was localized predominantly in the cytoplasm (A). HeLa, Jurkat Forskolin mst-3, a Mn 2ϩ -regulated STE20-like Protein Kinase also thank Dr. F. McKeon and Dr. R. Reed for the T cell library and the HeLa cell library, respectively.