INTRODUCTION

Ste20, a component of the pheromone-response pathway in budding yeast, represents the first identified member of a new family of serine/threonine protein kinases (1, 2). Recently, several mammalian and yeast homologs to Ste20 have been identified which appear to fall into two classes, those that bind and are activated by the small guanosine triphosphatases Cdc42 and/or Rac1, and those that do not appear to be regulated in this manner. Members of the former class are involved in a variety of cytoskeletal effects and include Ste20, Cla4, and the Paks¹ (3, 4, 5, 6, 7, 8). Those that do not have recognizable GTPase binding sites but are quite conserved throughout the catalytic domain to Ste20 and Paks include Sps1, and three human kinases, GC kinase, Mst1, and Mst2 (9, 10, 11, 12). Sps1 is involved in the activation of a mitogen-activated protein kinase pathway in Saccharomyces cerevisiae, which regulates spore formation (9), and recent work has demonstrated that GC kinase activates a stress-activated protein kinase pathway (13). The pathway in which Mst1 functions is not known; however, it does not appear to activate either the mitogen-activated protein kinases Erk1 and -2 or the stress-activated protein kinases Jnk and p38 (11).²

Here we report a structure/function analysis of Mst1. We find that there are two distinct functional domains within the COOH-terminal half of Mst1. The extreme COOH-terminal 57 amino acids encode a dimerization domain, while a 63-amino acid region, spanning amino acids 331-394, contains an inhibitory domain. The implications of these findings are discussed.

EXPERIMENTAL PROCEDURES

A 1.6-kb BamHI-EcoRI fragment from pJ3H-Mst1 (11) was subcloned into pBluescriptII-KS and pEG202-92 (a gift from T. Yen, FCCC, in which the pEG202 polylinker has been altered such that the BamHI site is in-frame) to create pBS-Mst1-BE and pEG202-Mst1, respectively. The plasmid pBS-3PE contains a 0.8-kb PstI-EcoRI fragment from pJ3H-Mst1 cloned into pBluescriptII-KS. Carboxyl-terminal deletions were made using the polymerase chain reaction (PCR) with pJ3H-Mst1 (11) or pBS-3PE as templates. The 5 primer was either the T7 primer or 5-ATGTACCCATACGATGTTCCAGATTACGCT-3 (HA primer), which hybridizes to the sequences encoding the HA tag, and the 3 primer was specific for the indicated end point and contained a stop codon and an EcoRI site. Internal deletions and point mutations were made using a two-step PCR (14) in which pBS-Mst1-BE was used as the template in the first reaction with the M13 reverse primer and an internal primer spanning the indicated deletion or point mutation. A portion of the product from this reaction was used together with pJ3H-Mst1 as the template and the M13 reverse and HA primers to amplify the full-length product. All amplified fragments were digested with BamHI-EcoRI and cloned into pJ3H (15). To create plasmid pJG4-5-Mst1, Mst1 was amplified from pJ3H-Mst1 using 5-CGGAATTCGATTACGGAGGAT-3 and 5-CCGCTCGAGGGTACCATCGATAAATTC-3, digested with EcoRI and XhoI, and subcloned into pJG4-5 (16, 17). All PCRs were performed using Deep Vent DNA polymerase (New England Biolabs) and an Idaho technologies thermal cycler. pLexA-RPB7 contains S. cerevisiae RNA polymerase II subunit RPB7 fused in-frame to LexA and pJG4-5-RPB4 is an in-frame fusion of an activation domain RPB4 fusion (18, 19). Plasmid, pGST-N-Mst1 contains a 0.8-kb BamHI-EcoRI fragment from pBS-3PE subcloned into pGEX-KT (19) to create an in-frame fusion between GST and amino acids 276-487 of Mst1.

COS1 cells were grown in Dulbecco's modified Eagle medium (DMEM), 10% fetal bovine serum containing 50 units/ml penicillin, 50 µg/ml streptomycin, 100 µg/ml kanamycin. Where indicated, cells were transfected using LipofectAMINE (Life Technologies, Inc.) according to manufacturer's protocol. Cells were washed twice in phosphate-buffered saline (PBS) followed by lysis and immunoprecipitation 48 h after transfection as described previously (11). Where indicated cells were incubated in DMEM without methionine for 30 min followed by incubation with 50 µCi of [³⁵S]methionine and cysteine (DuPont NEN) per ml of 90% DMEM without methionine, 10% DMEM with methionine, 10% fetal bovine serum for 4 h prior to cell lysis. Lysates from cotransfections were split in half for immunoprecipitation with anti-HA (12CA5) (Berkeley Antibodies Inc.) and anti-Myc (9E10) antibodies (20). Immunoblotting was performed essentially as described previously using either m1-45 (1:7500), 9E10 (1:2500), or 12CA5 (1:2500) antibodies (11).

Equal amounts of various forms of immunoprecipitated Mst1, as visualized by Western blot, were incubated with 5 µg of myelin basic protein (Sigma) in 20 µl of kinase buffer (40 m Hepes, pH 7.5, 10 m MgCl₂) containing 20 µ ATP and 2.0 µCi of [-³²P]ATP for 10 min at 30 °C. Reactions were terminated by the addition of 2 × SDS sample buffer and boiling for 5 min. A portion of the sample (15 µl) was separated on a 12% SDS-polyacrylamide gel, transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore) followed by exposure to x-ray film.

COS1 cells were scraped from culture dishes and lysed by Dounce homogenization in a buffer containing 50 m Tris-HCl, pH 8.0, 137 m NaCl, and 10% glycerol. The lysate was filtered through a 0.2 µ filter (Whatman), and approximately 1 mg of lysate in 200 µl of buffer was loaded onto a Superose 6 column (Pharmacia Biotech Inc.) that had been equilibrated with lysis buffer. The flow rate was 0.4 ml/min, and 250-µl aliquots were collected. The fractions were concentrated with Nanosep 10K concentrators (Filtron) to a volume of 50 µl. SDS sample buffer was added, and one third was applied to a 10% SDS-polyacrylamide gel followed by transfer to PVDF membrane and immunoblotting as described previously (11).

Escherichia coli strain DH5 was transformed with pGST-N-Mst1 or pGEX-KT. An overnight culture was diluted 1:10 with LB containing ampicillin (100 µg/ml), grown at 37 °C with shaking for 1 h. Isopropylthiogalactoside was added to 0.125 m. After a 3-h induction, cells were pelleted, resuspended in 0.04 volume of PBS containing 150 µg/ml lysozyme, and placed on ice for 5 min. Dithiothreitol was added to 5 m, and incubation was continued on ice for 5 min followed by the addition of Sarkosyl to 1.5%. The lysate was sonicated for 1 min, then spun at 12,000 × g for 5 min. Triton X-100 was added to the supernatant to 4% and mixed with 0.34 volume of 50% glutathione-agarose beads for 25 min at 4 °C. The beads were washed three times in PBS, 1% Triton X-100. Approximately 10 µg of GST and GST-N-Mst1 were eluted from glutathione-agarose beads with four washes in 25 µl of 30 m Tris, pH 7.5, 5 m glutathione. The eluted material was passed over a G-25 Sephadex (Sigma) column which had been equilibrated with 50 m triethanolamine, pH 8.2, 100 m NaCl. Reactions containing ~0.15 µ GST or GST-N-Mst1, 50 m triethanolamine, pH 8.2, 100 m NaCl were incubated either with various concentrations of glutaraldehyde (0.01-100 µ) for 15 min at room temperature or with 1.0 µ glutaraldehyde for the indicated amount of time. Cross-linking was terminated by the addition of SDS sample buffer and boiling for 5 min. GST-N-Mst1 and GST samples were applied to 6 and 10% SDS-polyacrylamide gels, respectively, followed by transfer to a PVDF membrane and immunoblotting as described previously (11). The COOH terminus of Mst1 was not cleaved from the GST moiety with thrombin since Mst1 is rapidly degraded with this procedure.

Deletions of the Mst1 carboxyl terminus were made via PCR, and each was cloned in-frame to lexA into pEG202-92, a high copy yeast vector containing HIS3 as the selectable marker and lexA under control of the constitutive ADH1 promoter (16, 17). Each construct together with pSH18-34 (contains 8 lexA operators fused to a lacZ reporter) and pJG4-5-Mst1 (a high copy yeast vector containing full-length Mst1 fused to an acidic activation domain and under control of the inducible GAL1 promoter) were used to transform yeast strain EGY48 (ura3 trp1 his3 lexA operator, LEU2) (16, 17). Transformants were grown in minimal medium with galactose as the carbon source. Dimerization was assessed by two methods; -galactosidase activity (21) and the ability of transformants to grow in the absence of leucine.

RESULTS

Previously we had noticed that a proteolytic product of Mst1 had more kinase activity than the full-length molecule (11). To map the potential inhibitory region, a series of COOH-terminal truncations were created. Deletions lacking the last 57 amino acids or less of Mst1 had activity similar to that of the full-length molecule (Fig. 1). However, deletion of an additional 70 amino acids (1-360) resulted in an ~3-fold increase in kinase activity, and deletion of 30 more amino acids (1-330) increased activity another 3-fold, suggesting that within amino acids 330-430 there exists an inhibitory region (Fig. 1). More substantial deletions of Mst1 were no more active than the 1-330 construct (data not shown). An internal deletion of amino acids 331-360 was only twice as active as the wild type kinase, and a larger internal deletion of amino acids 331-394 caused a ~9-fold increase in Mst1 kinase activity. These results suggest that the inhibitory element is localized to amino acids 331-394. To demonstrate that the activity observed with all Mst1 constructs was due to Mst1 kinase activity, a catalytically inactive form of Mst1 (K59R) is shown as a control.

In order to determine whether Mst1 might regulate a known kinase cascade, we tested whether the most active form of Mst1 (1-330) could activate the mitogen-activated protein kinases ERK1 and -2 or the stress-activated protein kinases Jnk and p38. While treatment with epidermal growth factor activated the ERKs, and anisomycin activated the stress-activated protein kinases, we failed to detect any activation of these kinases in extracts from unstimulated Mst-transfected COS cells (data not shown). Similarly, activated Mst1 failed to transform NIH-3T3 cells, as assessed by focus formation and by anchorage independence assays (data not shown).

We had noticed that immunoprecipitates of epitope tagged Mst1 from transfected cells contained a band that comigrated with endogenous Mst1, suggesting that Mst1 self-associates (11). To examine the ability of Mst1 to self-associate directly, COS cells were cotransfected with HA and Myc-tagged Mst1 and lysates subject to coimmunoprecipitation assays. One half of each lysate was immunoprecipitated with anti-HA antibodies and the other half with anti-Myc antibodies. Immunocomplexes were separated by SDS-PAGE in duplicate and blotted with anti-Myc or anti-HA antibodies. When both tagged forms of Mst1 were coexpressed, HA-Mst1 was detected in immunocomplexes with Myc-Mst1 (Fig. 2, lane 6), and conversely, Myc-Mst1 was detected in immunocomplexes with HA-Mst1 (Fig. 2, lane 9).

To localize the multimerization region, coimmunoprecipitation assays were performed using Myc-tagged Mst1 (M-Mst1) and either the first 300 amino acids of Mst1 (H1-300, catalytic domain) or amino acids 276-487 (H276-487) fused to the HA epitope. To enhance the sensitivity of the assay in this series of experiments the cells were first labeled with [³⁵S]methionine and cysteine prior to cell lysis, immunoprecipitated as before, and proteins were detected by autoradiography. As shown in Fig. 3, full-length Mst1 was unable to associate with amino acids 1-300; however, an interaction was detected between full-length Mst1 and amino acids 276-487. To map the self-association domain further, COOH-terminal truncations of H276-487 were made. Removal of the last 32 amino acids (H276-455) resulted in loss of multimerization, and it was not restored by further COOH-terminal truncations (Fig. 3, top panel).

Internal deletion analysis was performed to identify internal regions required for self-association and to determine whether the regions which function to inhibit Mst1 catalytic activity map to these same regions. In this series of experiments, the construct containing amino acids 276-487 is Myc-tagged, and the constructs containing internal deletions are HA-tagged. Both internal deletions lacking regions required for inhibition of Mst1 activity (H331-360 and H331-394) are able to multimerize indicating that the inhibitory and self-association domains are distinct (Fig. 3, middle panel). In addition, constructs lacking amino acids 391-409 (H391-409) and amino acids 411-430 (H411-430) are still able to multimerize. However, without amino acids 431-455 self-association does not occur. These results together with those from the COOH-terminal truncations indicates that the extreme COOH-terminal 56 amino acids encompass a multimerization domain.

Computer analysis using two different algorithms predicts that amino acids 431-487 form an -helix. To determine if the -helix is required for self-association, the putative -helix was disrupted by changing amino acid 444 from a leucine to a proline, a residue predicted to disrupt helix formation. As shown in Fig. 3 (middle panel) this single amino acid substitution disrupted the ability of Mst1 to multimerize. Wild-type Mst1 (11), as well as the L444P dimerization mutant, are located in the cytosol, as assessed by subcellular fractionation and by immunofluorescence (data not shown).

To further demonstrate the ability of Mst1 to self associate and map the region required for self association, the yeast two-hybrid assay was employed. Mst1 was subcloned into two yeast expression vectors. One such that Mst1 was in-frame to lexA (pEG202-Mst1) and another such that it was in-frame with an acidic activation domain (pJG4-5Mst1). These two vectors together with pSH18-34, a reporter vector containing 8 lexA operators fused to lacZ, were used to transform yeast strain EGY48 (16, 17). An interaction, as assessed by the ability of transformants to both grow in the absence of leucine and produce -galactosidase in a galactose dependent manner, was detected only when Mst1 was present as both a DNA binding domain and acidic activation domain fusion in the same cell (Table I). Mst1 was unable to interact with either control proteins, RPB7 or RPB4 (subunits of RNA polymerase), demonstrating that the Mst1-Mst1 interaction is specific. In addition, an interaction was detected only when the multimerization domain was intact. The Mst1 COOH terminus was also independently isolated as an interactor with full-length Mst1 in a yeast two-hybrid screen using a HeLa cDNA library (data not shown). We believe these results together with the coimmunoprecipitation assays demonstrate that Mst1 exists as at least a dimer.

Table I. Self-association of MST1 in yeast Bait Interactor Growth on Leu-minus medium  -Galactosidase activitya MST1 MST1 + 329.4  ± 52 MST1 1-455   23.2  ± 20 MST1 1-330   3.45  ± 3.29 MST1 276-487 + 402.2  ± 66 MST1  431-455   2.0  ± 1.36 MST1 L444P   5.3  ± 6.09 MST1 RPB4   3.99  ± 1.79 RPB7 MST1   3.25  ± 2.29 a  -Galactosidase activity is an average obtained with two isolates at two different time points.

Lysates prepared from asynchronous COS cells were subjected to gel filtration chromatography using a Superose 6 column. Fractions were analyzed by immunoblotting with anti-Mst1 antibodies. Mst1 eluted in a broad peak ranging from approximately 145 to 443 kDa with an average of 200 kDa, suggesting that Mst1 may exist as a multimer (Fig. 4). Mst1 was not detected at significant levels in higher or lower molecular weight fractions (data not shown). The inability to detect Mst1 in lower molecular weight fractions indicates that it may not exist as a monomer.

While coimmunoprecipitation assays and the yeast two-hybrid assay indicate that Mst1 is able to self associate, these experiments do not discriminate the ability to dimerize from the ability to form higher order multimers. In addition, gel filtration analysis indicates that Mst1 migrates with proteins that are more than twice the molecular mass of Mst1, suggesting that Mst1 may exist as more than a dimer. To explore this possibility, the COOH terminus of Mst1 (amino acids 276-487) was expressed and purified as a fusion to GST. After elution from glutathione beads, the purified protein was cross-linked with a constant amount of glutaraldehyde for various lengths of time or constant time with various concentrations of glutaraldehyde. GST alone was cross-linked as a control to demonstrate that GST does not dimerize under these conditions. While GST alone was unable to be cross-linked, Mst1 was cross-linked as a dimer very rapidly (Fig. 5). The dimer was not replaced with higher order species indicating that Mst1 exists predominantly as a dimer.

GST-N-Mst1

DISCUSSION

Two distinct functional domains have been localized within the COOH-terminal half of Mst1, one that affects kinase activity and one that mediates dimerization. A series of COOH-terminal and internal deletions indicates that amino acids 331-394 function to inhibit Mst1 catalytic activity. There are at least four mechanisms by which this region may inhibit kinase activity. First, an inhibitory molecule may bind this region, as has been suggested for the related GC kinase (13). However, our present and previous data argue against this scenario, as a COOH-terminal truncation of Mst1 is much more active than the full-length molecule, whether kinase activity is measured using immunoprecipitates (this study) or an in-gel assay (11). Associated proteins are lost in this latter assay since the kinase is denatured, separated by SDS-PAGE, then renatured. Second, this region may contain regulatory phosphorylation sites. Several serine and threonine residues lie within this region. Threonine is a potential phosphorylated residue within two consensus protein kinase C sites, one within amino acids 331-360 and another between 360 and 394. However, we do not believe these sites are utilized since phorbol ester treatment does not affect Mst1 activity, and in vivo ³²P-labeling indicates that Mst1 is phosphorylated only on serine residues (data not shown). In addition, serine phosphorylation is not lost when amino acids 331-394 are absent (data not shown). Third, the inhibitory domain may contain pseudosubstrate sites. We do not believe this region contains inhibitory autophosphorylation sites since amino acids 276-487 are unable to serve as a substrate for Mst1 in an in vitro kinase assay (data not shown). Nevertheless, it does not rule out the possibility that there are pseudosubstrate sites within this region. Fourth, loss of the inhibitory domain may simply result in a conformational change such that the active site is more accessible.

Using a combination of techniques we have also established that Mst1 dimerizes and have localized the dimerization domain to the extreme COOH-terminal 57 amino acids. This region is predicted to be highly -helical and a point mutation that disrupts the putative helix abolishes dimerization in vivo. While gel filtration chromatography indicates that Mst1 exists in a large molecular weight complex greater than that of a Mst1:Mst1 dimer, in vitro cross-linking demonstrates that Mst1 does not form high order multimers; therefore, Mst1 may be in association with additional proteins. Deletion of the dimerization domain (Fig. 1) has little effect on Mst1 kinase activity and the L444P point mutation had no effect on kinase activity (data not shown) indicating that the inhibitory and dimerization domains are distinct. Mst2, a very close homolog to Mst1, contains nearly identical sequences in these regions, indicating that both the inhibitory and dimerization domains are likely to be conserved between the two proteins (12).

Recently, it has been suggested that the structurally related GC kinase may oligomerize; however, this has not been demonstrated directly (13). Sequence comparison of the Mst1 dimerization and inhibitory domains to other kinases in this family indicates that these domains are not conserved at the amino acid level. The only family member with any potential similarity is Sps1, which is predicted to have a highly -helical COOH terminus. The ability of Sps1 to dimerize has not been examined (9).

While dimerization of receptor protein kinases is quite common, dimerization of cytosolic protein kinases is unusual and can be either inhibitory or activating. Kinases that dimerize include the serine/threonine protein kinases Akt, cGMP, and cAMP-dependent kinases, and smooth muscle myosin light chain kinase (22, 23, 24, 25). While removal of the dimerization domain had no effect on the activity of Mst1 unless the inhibitory domain was also removed, this domain may be important for the recognition by an effector molecule or may allow it to phosphorylate its natural substrates. Phosphorylation by cGMP kinase illustrates this latter possibility. While monomers of cGMP kinase are able to phosphorylate histone and peptides, only the dimer is able to phosphorylate vimentin, a protein suspected to be an in vivo substrate (23). Therefore, the role of dimerization with respect to Mst1 activity may become clear following the identification of natural substrates.

Although effectors of Mst1 have not been identified, mapping of a region important in the inhibition of Mst1 kinase activity suggests that this kinase, and by homology Mst2 as well, are highly regulated. Work is currently underway to determine the biological effects of Mst1 kinase activity.