Isolation of Hyperactive Mutants of the MAPK p38/Hog1 That Are Independent of MAPK Kinase Activation*

Mitogen-activated protein kinases (MAPKs) play pivotal roles in growth, development, differentiation, and apoptosis. The exact role of a given MAPK in these processes is not fully understood. This question could be addressed using active forms of these enzymes that are independent of external stimulation and upstream regulation. Yet, such molecules are not available. MAPK activation requires dual phosphorylation, on neighboring Tyr and Thr residues, catalyzed by MAPK kinases (MAPKKs). It is not known how to force MAPK activation independent of MAPKK phosphorylation. Here we describe a series of nine hyperactive (catalytically and biologically), MAPKK-independent variants of the MAPK Hog1. Each of the active molecules contains just a single point mutation. Six mutations are in the conserved L16 domain of the protein. The active Hog1 mutants were obtained through a novel genetic screen that could be applied for isolation of active MAPKs of other families. Equivalent mutations, introduced to the human p38α, rendered the enzyme active even when produced in Escherichia coli, showing that the mutations increased the intrinsic catalytic activity of p38. It implies that the activating mutations could be directly used for production of active forms of MAPKs from yeasts to humans and could open the way to revealing their biological functions.

MAPK 1 is a generic term for a large family of enzymes, which function in a variety of signal transduction pathways. Mammalian MAPKs are divided into at least three subfamilies (ERKs, p38s, and JNKs) based on degree of homology, biological activities, and phosphorylation motif (1)(2)(3)(4)(5)(6). Although highly homologous in structure and in pattern of activation, each MAPK is activated in response to a specific battery of signals and in turn phosphorylates a particular array of substrates. As a result, each MAPK imposes specific effects on the cell. For example, in many cell lines (e.g. fibroblasts), the ERK MAPKs are activated when cells are exposed to growth factors, and their activation is important for enhancement of cell proliferation (7,8). In other cells (neuronal and myogenic cell lines), however, activation of ERKs is associated with growth arrest and differentiation (7,9). In contrast to the ERK enzymes, the activity of p38 and JNK MAPKs is only slightly induced by growth factors. These enzymes are strongly activated in response to stress signals such as heat shock, osmotic shock, UV radiation, cytokines, and metabolic inhibitors. JNK and p38 MAPKs seem to be responsible mainly for protective responses, stress-dependent apoptosis, and inflammation (3,10). In some cell types, however, p38 and JNK may play a role in differentiation and development (3,10,11). Studies with knockout mice and knockout cell lines revealed essential roles for MAPKs in various aspects of embryonal development (12)(13)(14)(15)(16)(17).
Although many aspects of MAPK biology have been revealed, the exact role of each MAPK in a given biological system is not fully understood and is difficult to study. The main reason for this difficulty is our inability to activate a given MAPK in vivo and to follow the biochemical and physiological consequences. Currently, a MAPK is experimentally activated in vivo using extracellular stimuli or through expression of an active form of a component that functions upstream to that MAPK (7,8). Each of these treatments activates more than one MAPK and evokes many cellular responses. Activation of a given MAPK per se could be obtained theoretically by expressing an active form of this kinase, which would be active independently of external signals and upstream components. However, the catalytic activity of MAPKs is tightly regulated and strictly dependent on upstream activation (2,18,19). Although the mechanisms of MAPK activation have been revealed (2,20,21; see below), this knowledge could not be applied for the production of active forms of MAPKs. Here we describe a novel genetic screening system, which produces and isolates such active kinases.
Following exposure of cells to an extracellular ligand, the activity of the relevant MAPK increases ϳ1000-fold (18). This activation is mediated through a complex signal transduction net that culminates in phosphorylation of the MAPK by a MAPK kinase (MAPKK). MAPKKs are dual specificity kinases that phosphorylate MAPKs on particular Thr and Tyr residues. This dual phosphorylation is the basis for the dramatic increase in MAPK activity. MAPKs mutated in either the Thr or Tyr phosphoacceptors cannot be activated (18,22). The unusual mode of MAPK activation (through dual phosphorylation) underlies the difficulties in producing active forms of these enzymes, because a PO 4 -Thr-X-PO 4 -Tyr structure is difficult or impossible to mimic by mutagenesis (18).
Comparison of the crystal structure of ERK2 with that of dually phosphorylated ERK2 (21) shows that phosphorylation of the activation loop induces conformational changes in both the activation loop itself and in another domain at the COOHterminal extension (Pro 309 -Arg 358 ) known as L16. These conformational changes induce tight interactions between the phosphorylation lip and the L16 domain. A recent report sug-gested that these interactions create an interface for homodimerization of the kinase molecules. The putative dimerization is stabilized by hydrophobic contacts involving mainly leucine residues located at L16 of the two monomers and by an ion pair involving Phe 329 and Glu 343 (located in L16) of one monomer and His 176 of the other monomer (20). It is not known whether other MAPKs (e.g. JNKs and p38s) are dimerized upon phosphorylation and activation. Although the crystal structures of ERK2 and phospho-ERK2 revealed the conformational changes that activate the enzyme, they did not suggest a strategy for producing a constitutively active MAPK by mutagenesis.
We have devised a novel protocol for isolation of active forms of MAPKs using a genetic screen in yeast. The yeast Saccharomyces cerevisiae possesses five different MAPKs, which are highly homologous to their mammalian counterparts (1,23). The yeast MAPKs Fus3, Kss1, and Mpk1 are close to the ERK subfamily. The yeast Hog1 MAPK has a QM*TG*YVSTR phosphorylation motif (almost identical to that of p38) and is functionally replaced by either JNKs (24) or p38s (25). Hog1 is phosphorylated and activated by the MAPKK Pbs2 (26), a functional homolog of JNKK1/MKK4 (27). The Pbs2/Hog1 MAPK cascade is essential for survival of yeast cells under high osmotic conditions. Yeast cells lacking either PBS2 or HOG1 cannot grow on media supplemented with high concentrations of sugar or salt (26). Our effort to produce active forms of MAPKs took advantage of the Pbs2/Hog1 pathway. We reasoned that it should be possible to screen a library of randomly mutated HOG1 genes in cells lacking Pbs2 activity. The premise is that a HOG1 clone that allows pbs2⌬ cells to grow on salt should encode an active, MAPKK-independent Hog1 (see Fig. 1).
Previous efforts to obtain active forms of MAPKs were only partially successful. Several gain-of-function mutations that were identified in the S. cerevisiae FUS3 (28,29) and in the Rolled MAPK of Drosophila melanogaster (30,31) did not render the kinases catalytically hyperactive. The use of MAPKK-MAPK hybrids (32, 33) seems more useful. Yet, in the in vivo situation, MAPKKs and MAPKs are not colocalized in the cell and are differently controlled; thus, the use of the chimeric proteins might be problematic for physiological studies. Hence, despite efforts through various approaches, active forms of MAPKs, which are independent of MAPKK activity, are not available.
The approach taken in this study is most stringent, because it screens for MAPKs that are active in the complete absence of their relevant MAPKK. In addition, the design of the screen forces just minor changes in the MAPK (point mutations). We describe the isolation of nine different point mutations in the Hog1 kinase. Each mutation is sufficient to render Hog1 catalytically and biologically active, independent of upstream regulation. Remarkably, insertion of equivalent mutations to the human p38␣ rendered this enzyme catalytically active also.
Mutagenesis Procedures-The library of HOG1 mutants was produced in bacterial strain LE30 according to Silhavy et al. (34). A plasmid carrying the HOG1 and URA3 genes (pRS426-HOG1 obtained from M. Gustin) was introduced into LE30 cells. About 50,000 colonies were obtained and allowed to grow on LB ampicillin plates for 24 h. All colonies were collected in a pool into 1 liter of LB medium (containing ampicillin) and further grown for 12 h. Then, the culture was diluted 1:5 and further grown for 15 h prior to plasmid preparation.
Site-directed mutagenesis was performed by polymerase chain reaction or with the QuickChange kit (Stratagene) according to the recommendations of the manufacturer. The sequences of the primers used for mutagenesis are listed in Table I.
Screening of the HOG1 Mutant Library-Transformation of the library of HOG1 mutants into yeast was performed as described by Schiestl and Gietz (35). Transformed cells were plated on selective YNB-URA plates. Colonies that grew (about 10,000 per 100-mm plate) were replica-plated onto YPD plates containing different concentrations of NaCl (0.9, 1.1, and 1.3 M).
Plasmid loss assay, for positive colonies, was performed by streaking patches of positive colonies on YPD plates, allowing the colonies to grow for 24 h, and isolating single colonies. These single colonies were replica-plated onto YNB-URA plates as well as onto NaCl-containing YPD plates.
RNA Preparation and Analysis-Cells were grown in 100 ml of YNB-URA to an A 600 of 0.5-0.9, split, and collected by centrifugation. Half of the cells were induced for 90 min with YNB-URA containing 1 M NaCl, whereas the other half of the cells were resuspended in YNB-URA. After induction, cells were centrifuged and frozen in liquid nitrogen. RNA preparation, primer extension analysis, and sequences of the primers were previously described (36,37).
Preparation of Cell Lysates and Western Blot Analysis-50-ml cell cultures were grown to an A 600 of 0.5-1.0. Half of the culture was induced with 1 M NaCl for 10 min, as described above. Cultures were pelleted and resuspended in 8 ml of water and 10 ml of 20% trichloroacetic acid. After the samples were repelleted, they were resuspended in 200 l of 20% trichloroacetic acid at room temperature, and 650-mg of glass beads were added. Each sample was vortexed twice for 4 min each time. Supernatants were transferred to new Eppendorf tubes, and glass beads were rinsed twice with 200 l of 5% trichloroacetic acid (the final concentration of trichloroacetic acid was 10%). Following centrifugation, pellets were resuspended in 200 l of 2ϫ Laemmli sample buffer followed by addition of 100 l of 1 M Tris base. Samples were vortexed for 30 s and boiled for 3 min prior to centrifugation. Supernatant was used.
The SDS-polyacrylamide gel electrophoresis, Western blot, and ECL reaction for identification of the HA-Hog1 protein and the phosphorylated Hog1 were performed as described in Sambrook et al. (38). The antibody for the HA tag, 12CA5, is monoclonal. A dilution of 1:1000 was used. For identification of the double phosphorylated Hog1, ␣-P-p38 antibodies (New England Biolabs) were used (diluted 1:2,500). The secondary antibodies, anti-mouse and anti-rat, respectively, were diluted 1:10,000.
Preparation of Native Cell Lysates and in Vitro Kinase Assay-200-ml cell cultures were grown to an A 600 of 0.5-1.0. After a 10-min induction with 1 M NaCl, as described above, cells were centrifuged at 4°C, and the pellet was washed with 50 ml of ice-cold water. Following centrifugation the pellet was resuspended in 1-2 volumes of ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 0.25 M NaCl, 0.1% Nonidet P-40, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml trypsin inhibitor, 10 g/ml pepstatin A, 313 g/ml benzamidine, 1 mM sodium vanadate, 10 mM NaF, 1 mM p-nitrophenyl phosphate, 10 mM ␤-glycerol-P). 600 mg of glass beads were added, and the pellet was vortexed eight times for 1 min each time. Samples were centrifuged at 800 ϫ g for 5 min, and the supernatant was centrifuged again at 15,000 ϫ g for 15 min at 4°C. Samples were aliquoted to small volumes (ϳ200 l) and frozen immediately in liquid nitrogen.
For immunoprecipitation of the Hog1 protein, 40 l of 50% protein A were washed with ice-cold phosphate-buffered saline buffer and preconjugated with 5 l of 12CA5 antibody overnight at 4°C. Following two washes with phosphate-buffered saline buffer and then one with lysis buffer, the beads were resuspended in lysis buffer, and 300 g of the lysate were added. Samples were incubated for 1 h at 4°C. Then, three washes with lysis buffer followed by three washes with kinase buffer (25 mM HEPES, pH 7.5, 20 mM MgCl 2 , 20 mM ␤-glycerol-P, 10 mM p-nitrophenyl phosphate, 0.5 mM sodium vanadate, 1 mM dithiothreitol) were performed. The kinase reaction was performed by resuspending the beads in 30 l of kinase buffer containing 20 M ATP, 5 Ci of [␥-32 P]ATP, and 9 g of GST-ATF2 as a substrate. The experimental details on optimizing the conditions for the Hog1 in vitro kinase assay, kinetic parameters, and affinities to substrates are described elsewhere. The reaction took place at 30°C for 30 min and was stopped by transferring the samples to ice and adding 10 l of Laemmli loading buffer ϫ4 prior to boiling. Kinase reactions were separated through SDS-polyacrylamide gel electrophoresis. The gels were dried under vacuum, and exposed to x-ray film.
Expression, Purification, and Analysis of Human p38␣-Wild type and active mutants of p38␣ were expressed in Escherichia coli using the pET15b expression vector (Novagen) so that all proteins contained a polyhistidine tag in their NH 2 terminus. Growth conditions, induction of expression, and protein purification were performed according to Wang et al. (39). Purified proteins were assayed as previously described (40) using GST-ATF2 (20 g) as a substrate.

Rationale and Design of the Genetic Screen for Active Forms
of Hog1-Our screen for active forms of MAPKs takes advantage of the pbs2⌬ strain, which lacks the MAPKK Pbs2 and cannot grow under high osmotic conditions. Overexpression of the MAPK Hog1 (the Pbs2 target) did not enable pbs2⌬ cells to grow on salt (see Fig. 2A; second row in each plate), demonstrating the absolute dependence of Hog1 activity on Pbs2-mediated phosphorylation. This result forms the basis for our premise that only an active, independent form of Hog1 might enable pbs2⌬ cells to grow on salt. The idea is therefore to produce a library of HOG1 mutants, hoping that one mutant or a few mutants would gain intrinsic catalytic activity, which is independent of Pbs2 activation. The rare active mutants will be identified by introducing the library to pbs2⌬ cells and selecting colonies on salt. Each colony that grows should harbor an active, Pbs2-independent Hog1 kinase ( Fig. 1).
Isolation of Point Mutations in HOG1, Which Render It Independent of Pbs2 Activation-A library of HOG1 mutants was produced (34,41) and introduced into a pbs2⌬ strain (26). 5 ϫ 10 5 transformants were obtained and allowed to grow on selective media (YNB with no uracil) with no salt. Then colonies were replica-plated to plates supplemented with NaCl (0.9 -1.3 M). 150 positive colonies (colonies that grew on salt concentrations of 1.1 M NaCl or higher) were collected. The linkage between growth on salt and the library plasmid was verified through a plasmid loss assay and then by retransfection (following purification from yeast) to pbs2⌬ cells. 41 clones passed these tests (data not shown) and were considered true positives (Table II). Fig. 2A shows that the Hog1 mutants isolated, but not wild type Hog1, enable pbs2⌬ cells to grow on salt.
The 41 true positive clones were sequenced. Each clone was found to contain a single point mutation in the coding sequence of Hog1. Many of the clones harbor an identical mutation (Table II), suggesting that the screen was saturated. Altogether, nine different point mutations were identified (Table II; Figs. 2A and 3). To show unequivocally that each of these point mutations is sufficient to render Hog1 independent of Pbs2, we introduced, through site-directed mutagenesis, each mutation to a wild type HOG1 gene and expressed the resulting mutants in pbs2⌬ cells. All nine mutations allowed growth on salt, showing that indeed each point mutation is sufficient to make Hog1 independent of Pbs2.
The results shown so far demonstrate that the HOG1 variants isolated in the screen are independent of Pbs2. Because MAPKs may be active as dimers (2,20), the possibility remains that the mutants are not exclusively independent but dimerize with the endogenous Hog1 that is expressed in the pbs2⌬ strain. We therefore tested the ability of the active mutants to allow the pbs2⌬hog1⌬ double knockout strain to grow on salt. Obviously, this strain is not rescued by overexpression of either Hog1 or Pbs2 (Fig. 2B). Yet, it is rescued by the Hog1 mutants (Fig. 2B), showing that they are active biologically, functioning alone, independently of Pbs2 and endogenous Hog1. To verify that the active Hog1 mutants rescue pbs2⌬ cells by activating the authentic biochemical pathway downstream to Hog1, we measured mRNA levels of GPD1 and GPP2 genes. The products of these genes are involved in glycerol biosynthesis, and their expression is Hog1-dependent (1). Our analysis revealed that all the Hog1 active mutants, but not wild type Hog1, induced transcription of these genes in pbs2⌬ cells (data not shown).
Six of the Activating Mutations Are Located in the Dimerization Domain- Fig. 3A shows the location of the mutations within the HOG1 linear sequence. Strikingly, six of the nine mutations are located in a short stretch of residues between amino acids 314 and 332, which is part of the L16 domain. These mutated residues appear to be in close proximity to (but not within) the docking domain suggested recently by Nishida and co-workers (42). The other two mutations are located at the NH 2 terminus. One mutation (D170A) is located just four amino acids from the phosphoacceptor Thr 174 . The second mutation in the NH 2 terminus is Y68H. Alignment of the HOG1 sequence with the sequences of mammalian MAPKs (Fig. 3B) reveals that all mutations but one, W320R, occur in residues that are conserved in at least one subfamily.
Hog1 Variants Are Catalytically Active in Vivo in the Absence of Salt Induction-Having isolated Hog1 variants that function biologically independently of Pbs2, we wished to reveal the biochemical basis of their unusual property. The most attractive explanation would be that these molecules have acquired intrinsic catalytic activity. To directly test its kinase activity, each mutant was expressed in hog⌬ cells, immunoprecipitated, and assayed in vitro for its ability to phosphorylate ATF2 (Fig.  4). Strikingly, most mutants exhibited very high catalytic activity in the absence of any stimulation (Fig. 4). Under these conditions, the activity of wild type Hog1 molecules was barely detectable (Fig. 4, upper panel, lane 3). Particularly high basal kinase activity was measured for Hog1 F318L and Hog1 F318S . The activity of these molecules did not change when cells were treated with salt (Fig. 4). Thus, Hog1 F318L and Hog1 F318S man- a For unknown reasons, two mutants, N391D and A314T, were not able to rescue pbs2⌬ cells when expressed from a strong promoter and containing an HA tag. Therefore these mutants were studied only in some of the further experiments. FIG. 2. HOG1 active mutants, but not HOG1 wild type, are able to rescue pbs2⌬ cells and hog1⌬pbs2⌬ cells from high salt concentrations. A, Two rows of plates are shown. On the left-hand side of each row is a master plate (not containing salt), and the other two plates are replicas (containing salt). Each plate is divided into two parts; the right side contains hog1⌬ cells (used as controls (22)), and the left side contains pbs2⌬ cells (26). Note that replica plates are supplemented with two different concentrations of NaCl. PBS2 and vector, written in white on the plates, denote the plasmid identity in the cells of the upper row. In all other rows, the same plasmids (wild type HOG1 or mutants) are expressed in both pbs2⌬ and hog1⌬ cells. ifest their maximal catalytic activity under any growth conditions and could be regarded as constitutively active molecules. Because equal amounts of Hog1 proteins were used in all assays (Fig. 4), it is clear that the specific activity of the mutants is much higher than that of wild type Hog1. Not only Hog1 F318L and Hog1 F318S but also all the other mutants showed very high catalytic activity in the absence of any stimulation. Yet, the activity of the other mutants further increased when salt was added to the culture. Namely, Hog1 Y68H , Hog1 D170A , Hog1 W320R , Hog1 F322L , and Hog1 W332R mutants are also hyperactive under any growth conditions but manifest their maximal activity after salt induction. We believe that the further increase in activity of some mutants reveals a new mechanism of control of Hog1 (perhaps a removal of a repressor; see "Discussion").
In summary, all Hog1 mutants isolated in the screen are hyperactive independently of any induction. Their specific activity is way above the maximal activity of wild type Hog1, but the activity of some mutants could be even further induced by salt via an unknown mechanism.
Insertion of Equivalent Mutations into the Human p38␣ Renders the Kinase Catalytically Active-To obtain another indication that the mutants have acquired intrinsic, independent catalytic activity, we attempted to produce them as recombinant proteins in E. coli and to measure their activity. Yet, we were not able to produce soluble Hog1 proteins. Because soluble recombinant versions of Hog1 are difficult to obtain (other laboratories were also not successful in producing recombinant Hog1), we decided to test the activating mutations in the human p38␣ protein, which is known to be soluble when expressed in bacteria. This approach is validated by the fact that the activating mutations of HOG1 occur in residues that are conserved in mammalian MAPKs (Fig. 3B). Hence, we replaced Phe 327 of the human p38␣ gene (homolog of Phe 322 in HOG1) with Leu or Ser. The p38 Phe327Leu and p38 Phe327Ser , as well as a wild type p38 protein, were expressed in E. coli utilizing the pET15 vector and purified using Ni 2ϩ -agarose. The recombinant purified proteins were tested for their kinase activity in vitro (Fig. 5A). As expected, wild type p38 exhibited very low catalytic activity. The mutants, however, exhibited significant kinase activity (at least 70-fold higher than wild type; Fig. 5A). Because the mutated p38 enzymes are active as recombinant proteins expressed in bacteria, it is clear that they have ac-quired intrinsic catalytic activity, which is independent of activation by any upstream MAPKK. Furthermore, the result with the p38 mutants validates the notion that mutations identified in the HOG1 screen could be used to produce active forms of mammalian MAPKs.
Analyzing the results of the in vitro kinase assay (Fig. 5A), we noticed a phosphoprotein of 38 kDa. This phosphoprotein appeared only when active p38 mutants were used (Fig. 5A). Because this protein is almost certainly an autophosphorylated p38, we examined the autophosphorylation activity of the mutant enzymes. We found that p38 Phe327Leu and p38 Phe327Ser have acquired the intrinsic capability for autophosphorylation (Fig. 5B). The autophosphorylation activity of the mutants is directed mainly toward Thr 180 , as determined by Western blot analysis (data not shown). Some phosphorylation on Tyr 182 was detected on both wild type and mutant recombinant p38. Thus, an increase in Thr 180 autophosphorylation is most probably the mechanism responsible for the independence of the active variants of upstream MAPKKs. Autophosphorylation activity of wild type p38 was barely detectable (Fig. 5B).
Most of the Active Hog1 Mutants Are Not Phosphorylated on Thr 174 and Tyr 176 in pbs2⌬ Cells-The results shown above suggest that the Hog1 active mutants acquired an increased intrinsic catalytic activity. This idea is most strongly supported by the results with the active recombinant p38 proteins, which certainly were not activated by E. coli proteins.
Although it seems that the mutants have acquired intrinsic FIG. 4. Hog1 mutants are catalytically active in the absence of salt induction. Hog1 wild type or active Hog1 mutants were expressed and immunoprecipitated from hog1⌬ cells and assayed in vitro using GST-ATF2 as a substrate. Cells were either exposed or not exposed to FIG. 5. Human p38␣ proteins carrying mutations equivalent to those that activate Hog1 possess an intrinsic catalytic activity and the capability to autophosphorylate. Wild type (w.t.) p38 protein and mutants p38 F327L and p38 F327S were expressed in E. coli and purified to near homogeneity. Purified proteins were tested for their kinase activity in vitro using GST-ATF2 as a substrate (A) and for autophosphorylation activity (B). Assay mixtures were separated through SDS-polyacrylamide gel electrophoresis. Gels were stained with Coomassie Brilliant Blue (upper panels) and exposed to x-ray film (lower panels). activity, we wished to rule out the possibility that the Hog1 mutants are active in vivo because they acquired an affinity to another MAPKK, which phosphorylates and activates them. Because at least five MAPK cascades function in yeast (1,23), there should be at least four intact MAPKKs in the pbs2⌬ cells. To address this possibility, we measured, by Western blot analysis, the phosphorylation status of the Hog1 mutants. We used anti-phospho-p38 antibodies that recognize dually phosphorylated Hog1 (43). Whole cell extracts were prepared from pbs2⌬ and hog1⌬ cells expressing wild type Hog1 or the various mutants. Extracts were prepared from cells exposed, or not exposed, to salt. The Western blot analysis revealed several phosphorylation patterns (Fig. 6). Most significantly, when expressed in pbs2⌬ cells, the active Hog1 mutants F318S, W320R, and W332R show no detectable phosphorylation (Fig.  6A). Hog1 mutants Y68H, D170A, and F318L show low levels of phosphorylation in pbs2⌬ cells. This level of phosphorylation is even below the level measured in wild type Hog1 expressed in pbs2⌬ cells exposed to osmotic shock (Fig. 6A, upper panel, lane  8). A possible explanation for the low level of Hog1 phosphorylation in pbs2⌬ cells could be autophosphorylation activity (28,40). It should be stressed that this low phosphorylation of Hog1 in pbs2⌬ cells was insufficient to allow growth on salt ( Fig. 2A). Taken together, these results suggest that the dual phosphorylation state of the active mutants is not related to their intrinsic catalytic activity. This notion has been further addressed by mutagenizing the phosphoacceptors Thr 174 and Tyr 176 in all the mutants (see below). Hog1 mutants showed a different pattern of phosphorylation when expressed in hog1⌬ cells (Fig. 6B). First, mutants Y68H, D170A, A314T, and W320R are phosphorylated in the absence of salt stimulation, at a significantly higher level than that of the wild type Hog1. Second, the level of their phosphorylation following induction with salt is similar, if not higher, than that of wild type. Thus, the presence of Pbs2 was important for increased phosphorylation of these mutants (compare Fig. 6, A and B), although Pbs2 was not required for their biological activity (Fig. 2).
Strikingly, mutants F318L and F318S expressed in hog1⌬ seem to be labile. As shown in Fig. 6B (upper panel, lanes 11  and 12 and lower panel, lanes 5 and 6), in lysates prepared from hog1⌬ cells expressing Hog1 F318L or Hog1 F318S , a 35-kDa protein reacted with the anti-HA antibodies. Normally, the apparent molecular mass of native Hog1 is about 50 kDa. The 35-kDa peptide appeared only when some lysis protocols were used (trichloroacetic acid protocol) and not others (native lysis; see Fig. 4, for example; for protocols, see "Material and Methods"). Thus, Hog1 F318L and Hog1 F318S are intact proteins in vivo but are susceptible to cleavage when vigorous lysis procedures are used. It is clear that these mutations induced dramatic structural changes, which are probably responsible for the high catalytic activity of these two mutants. Note that Hog1 F318L and Hog1 F318S molecules are fully functional in hog1⌬ cells and allow growth on salt (Fig. 2). In summary, the Western blot analysis suggests that in pbs2⌬ cells most mutants are not activated through dual phosphorylation, supporting the results above (Figs. 4 and 5) that strongly suggested that the mutants acquired an intrinsic independent catalytic activity.
The Activity of the Active Hog1 Variants Depends on Thr 174 but Not on Tyr 176 -The Western blot analysis strongly suggests that the active Hog1 mutants do not require phosphorylation for their activity. Yet, because some studies suggested that very low phosphorylation, which may not be detected in the Western analysis, could induce significant activation (44), we decided to prove this unequivocally. To this end we mutated in each of the active variants the phosphoacceptors Thr 174 and Tyr 176 . We expected that all mutants (and in particular F318S, W320R, and W332R, which are not phosphorylated at all in pbs2⌬ cells) would tolerate replacement of Thr 174 with Ala and Tyr 176 with Phe and would remain active. As can be seen in Fig.  7, replacement of Tyr 176 with Phe had no effect on the ability of most mutants to allow growth of both hog1⌬ cells and pbs2⌬ cells (see Fig. 8 for Hog1 D170AY176F ). Only Hog1 A314T and Hog1 W320R were affected by the Y176F mutation. This result suggests that phosphorylation of Tyr 176 is dispensable for the activity of the mutants and further supports the notion that they are independent of any MAPKK.
Unexpectedly, when we mutated Thr 174 to Ala the mutants lost their ability to rescue either hog1⌬ or pbs2⌬ cells (Fig. 7). Similarly, when both Thr 174 and Tyr 176 were changed to Ala and Phe, respectively, none of the mutants could support growth on salt (Fig. 8). These findings demonstrate that most Hog1 mutants do not require dual phosphorylation for their activity (Fig. 6A) but depend on Thr 174 . It is possible that phosphorylation of Thr 174 is not required, but this residue plays an essential role in maintaining the catalytic core of the enzyme. Indeed, not only did Robbins et al. (18), who replaced the Thr 183 phosphoacceptor in ERK2 with Glu, not observe an increase in activity, but in fact they measured a dramatic reduction in the maximal kinase activity. Alternatively, it is possible that the Hog1 mutants acquired autophosphorylation activity that selectively phosphorylates Thr 174 .
Active Hog1 Molecules Reveal Novel Biological Effects of Hog1-Because the active Hog1 mutants were isolated as molecules that allow pbs2⌬ cells to grow on salt, they are by definition, biologically active. In addition to their ability to allow growth under osmotic stress, we noticed that expression of the active forms in wild type cells significantly affected important biological properties of the cells. The growth rate under optimal growth conditions of cells expressing the mutants was dramatically reduced. The generation time of these cultures was about 2 times longer than that of cells expressing wild type Hog1. In addition, the active mutants increased cell aggregation and flocculation (data not shown). A detailed description of the phenotypes induced by the mutants will be provided elsewhere. 2 These effects disclose novel roles for Hog1 in growth arrest and cell-cell interaction and provide strong indications for the potential uses of the active molecules for biological studies. DISCUSSION Although MAPKs are involved in pivotal biological processes and are therefore extensively studied, it has been difficult to address the exact role of a given MAPK in a particular biological system. Such a question could be approached using active forms of MAPKs that were hitherto not available. This study describes a novel genetic screen in yeast that provides active MAPK molecules that function independently of their MAPKK.
The basic rationale behind this screen is that only an active form of a MAPK would induce the appropriate respective phenotype in a MAPKK null strain. This rationale was applied here for isolation of active Hog1 molecules. The mutants isolated could execute all Hog1 biochemical and biological activities independently of Pbs2 activity. Furthermore, the active Hog1 variants dramatically affected the growth rate and other properties of the cell, disclosing novel biological activities of Hog1. Their ability to perform all these functions stems from the intrinsic catalytic activity they acquired. The importance of the mutants for the studies of MAPKs in general is manifested by the fact that insertion of similar mutations into the human p38␣ kinase rendered it catalytically active. Preliminary studies with mammalian cells in culture show that the active p38 mutants are highly active in vivo, similar to the Hog1 mutants in yeast.
Thus, the results with p38␣ open the way to producing more active forms of various MAPKs based on the mutations identified in Hog1. Insertion of appropriate mutations into JNK (equivalent to the HOG1 mutations Y68H and W322R (Fig. 3)) and into ERK2 (equivalent to the HOG1 mutations D170A, A314T, F318S, and F318L (Fig. 3)) may render these MAPKs active. A systematic effort that will test these mutations in a variety of MAPKs could provide a useful battery of active forms of each MAPK. In addition to suggesting particular mutations that may render MAPKs constitutively active, the mutations identified in our study point out the domains that should be mutated in an effort to activate MAPKs. Six of the nine mutations were found in L16 between residues 314 and 332. This domain is equivalent to residues 323-341 in ERK2, which was suggested to be important in forming the interface for dimerization (20). It is currently not known whether Hog1 or p38 are active as dimers. It could be, however, that the activating mutations (in particular, mutations that replaced a charged amino acid with a hydrophobic residue) support formation of intermolecular contacts and consequently the formation of an active dimer. Additional support for this notion comes from the D170A mutation. Asp 170 is homologous to Asp 173 in ERK2, which resides near His 176 , which forms an ion pair with Glu 343 of another monomer (20). The possibility that each mutant stabilizes a dimer is exciting, because it suggests that combin-  7. Active Hog1 mutants containing the mutation Y176F can rescue yeast from high salt concentrations but cannot when containing the T174A mutation. Each plate is divided into two. Hog1 wild type and mutants on the left side contained the mutation T174A, and on the right side they contained the mutation Y176F. A, Hog1 molecules were expressed in hog1⌬ cells. The positive control (first row, left side) was a wild type Hog1 expressed in hog1⌬ cells; the negative control (first row, right side) was a wild type Hog1 expressed in pbs2⌬ cells. B, proteins were expressed in pbs2⌬ cells. The positive control was Hog1 F318S expressed in pbs2⌬ cells (first row, right side); the negative control was wild type Hog1 expressed in pbs2⌬ cells (first row, left side).

FIG. 8. Elimination of both phosphorylation sites (indicated by two asterisks) in Hog1 wild type and mutants abolishes their activity in both pbs2⌬ cells (A) and in hog1⌬ cells (B).
Replacement of only Tyr 176 with Phe does not abolish the activity of Hog1 D170A (indicated with one asterisk; A, left plate, row 5). The upper plates are master plates containing YNB-URA media that were replica-plated onto YPD ϩ 1.3 M NaCl plates.
ing several mutations on the same MAPK molecule would generate an even more stable and more active dimer. We are currently combining mutations on the same Hog1 molecule, and preliminary results suggest that indeed certain combinations result in superactive kinases. Another possibility to explain the intrinsic activity of the mutants is that the mutations strengthen the contacts between L16 and the phosphorylation lip of the same molecule. Such contacts may result in refolding the phosphorylation lip from the closed conformation to the open structure (21).
Some of the mutants isolated (e.g. Hog1 F318S and Hog1 F318L ) are stronger than others and show very high basal activity in vivo. Strikingly, the activity of other mutants was further increased by salt. What mechanism could be responsible for this increase in activity of some mutants when salt is provided? It may be that activation of Hog1 requires the removal of an inhibitory component, in addition to MAPKK-dependent phosphorylation. In such a case, the salt-dependent increase in the activity of the mutants is a result of this molecular event. This possibility may be supported by studies that identified several inhibitors of JNK1, including JNK-interacting protein (45) and GST (46). It may be that the mechanism of action of some active mutants involves their release from an inhibitory complex in combination with an increase in their catalytic activity.
We thus believe that more than one mechanism underlies the activity of the mutants. Some mutants may form spontaneous dimers, whereas others may use another mechanism (escaping inhibition?). The differences in the intrinsic catalytic activities of the various mutants support the idea of different mechanisms. Clearly, many mechanistic studies are required to fully explain the mechanism of action of each mutant. In addition, structural studies, which are in progress, will reveal the changes in protein folding induced by each mutation.
We may already speculate, however, that combining, on the same Hog1 molecule, a mutation that uses one mechanism with a mutation that uses another may generate an enzyme molecule that is even more active in vivo than the molecules described here. As already stated, our preliminary results in this direction support this notion.
Finally, the rationale that was successfully applied here for isolation of active Hog1 molecules could be used to screen for other active MAPKs. For example, active forms of Fus3 could be isolated in cells lacking the MAPKK Ste7, using the mating phenotype as the biological assay. Analogously, constitutively active forms of Kss1 could induce a pseudohyphae phenotype in ste7⌬ diploid strains. Furthermore, one can envisage a screen for active forms of mammalian MAPKs in yeast. For example, because p38 and JNK are biologically functional in yeast (24,25), it should be possible to screen for active forms of these enzymes directly in pbs2⌬ cells. However, the results with p38␣ (Fig. 5) suggest that screening for active forms of other MAPKs may not be necessary. The activating mutations identified in Hog1 could be applied directly in many other MAPKs.