Two Adjacent Docking Sites in the Yeast Hog1 Mitogen-Activated Protein (MAP) Kinase Differentially Interact with the Pbs2 MAP Kinase Kinase and the Ptp2 Protein Tyrosine Phosphatase

Abstract

Functional interactions between a mitogen-activated protein kinase (MAPK) and its regulators require specific docking interactions. Here, we investigated the mechanism by which the yeast osmoregulatory Hog1 MAPK specifically interacts with its activator, the MAPK kinase Pbs2, and its major inactivator, the protein phosphatase Ptp2. We found, in the N-terminal noncatalytic region of Pbs2, a specific Hog1-binding domain, termed HBD-1. We also defined two adjacent Pbs2-binding sites in Hog1, namely, the common docking (CD) domain and Pbs2-binding domain 2 (PBD-2). The PBD-2 docking site appears to be sterically blocked in the intact Hog1 molecule, but its affinity to Pbs2 is apparent in shorter fragments of Hog1. Both the CD and the PBD-2 docking sites are required for the optimal activation of Hog1 by Pbs2, and in the absence of both sites, Hog1 cannot be activated by Pbs2. These data suggest that the initial interaction of Pbs2 with the CD site might induce a conformational change in Hog1 so that the PBD-2 site becomes accessible. The CD and PBD-2 docking sites are also involved in the specific interaction between Hog1 and Ptp2 and govern the dynamic dephosphorylation of activated Hog1. Thus, the CD and the PBD-2 docking sites play critical roles in both the activation and inactivation of Hog1.

Mitogen-activated protein kinases (MAPKs) are a conserved family of protein kinases that serve major roles in intracellular signal transduction in eukaryotic cells (for a review, see reference 9). Five different MAPKs have been described in budding yeast (Saccharomyces cerevisiae), and more than 10 have been described in mammals. Different external stimuli, such as mitogenic growth factors, proinflammatory cytokines, and osmotic and oxidative stresses, activate distinct subsets of MAPKs. MAPKs are activated through a kinase cascade in which an activated MAPK kinase kinase (MAPKKK) phosphorylates and thus activates an MAPK kinase (MAPKK). An activated MAPKK then phosphorylates and activates an MAPK. Because there are also large numbers of MAPKKs and MAPKKKs in each organism, the potential number of MAPKKK-MAPKK-MAPK combinations is enormous. However, only a small subset of the potential combinations is actually activated upon a specific stimulus, suggesting that specific MAPKKK-MAPKK-MAPK interactions are tightly regulated.

Several mechanisms exist to ensure highly selective recognition among the three tiers of kinases belonging to a single MAPK cascade. Specific substrate-enzyme interaction, i.e., recognition of the substrate phosphorylation site(s) by the kinase catalytic site, must obviously be important. MAPKKs phosphorylate both the threonine and the tyrosine in the T-X-Y motif within the activation loop of substrate MAPKs. This specificity, however, is inadequate to select a unique substrate, because multiple species of MAPKs usually coexist in the same cell, and they all possess similar phosphorylation site sequences (9). Similarly, substrates of MAPKs have a relatively simple substrate phosphorylation site specificity. MAPKs phosphorylate a Ser or a Thr residue, followed by a Pro (the S/T-P motif), which is too simple to be selective. Thus, MAPKK-MAPK specificity, as well as MAPK-substrate specificity, must be enhanced by an additional mechanism, such as a specific docking interaction. In fact, each MAPK has a site termed the common docking (CD) domain, which is located immediately C-terminal to the kinase catalytic domain. The CD domain of an MAPK interacts with its specific activator (MAPKK), inactivator (phosphatase), and substrates (36). However, there appears to be much variation in the fine details of MAPKK-MAPK specificity determination.

The yeast osmoregulatory Hog1 MAPK cascade is an excellent model in which to study the mechanistic details of intracellular signaling (13, 16). The Hog1 MAPK is activated when cells are exposed to hyperosmotic extracellular environments (6). Activated Hog1 initiates an adaptive program that includes adjustments in cell cycle progression, regulation of protein translation, induction or repression of various genes, and synthesis and intracellular retention of the compatible osmolyte glycerol. The budding yeast has two, apparently redundant, osmosensing mechanisms. The first is called the SLN1 branch, in which a complex two-component system composed of the Sln1-Ypd1-Ssk1 multistep phosphorelay activates the redundant Ssk2 and Ssk22 MAPKKKs (19, 25, 27). The second is called the SHO1 branch, in which a tetraspan membrane protein with an intracellular SH3 domain (Sho1), together with two mucin-like membrane glycoproteins (Hkr1 and Msb2), generates an intracellular signal that leads to the activation of the Ste11 MAPKKK (18, 28, 38, 39). Signals emanating from either upstream branch converge at a common MAPK-MAPKK, Pbs2, which is the specific activator of the Hog1 MAPK (4, 6, 18, 19). Activated Hog1 is eventually inactivated by the concerted actions of Ser/Thr phosphatases and Tyr phosphatases, of which the Ptp2 tyrosine phosphatase is the most important (32).

In this work, we investigated the molecular mechanism by which the yeast Hog1 MAPK specifically interacts with the Pbs2 MAPKK and the Ptp2 protein phosphatase. The major conclusion from this study is that Hog1 recognizes Pbs2 and Ptp2 through two separate docking interactions: one via the CD domain and the other involving a nearby region, Pbs2-binding domain 2, termed the PBD-2 domain. These docking interactions govern not only their specific interaction but also the phosphorylation and dephosphorylation of the substrate sites. Interestingly, the roles of the CD and PBD-2 docking sites differ between the Hog1-Pbs2 interaction and the Hog1-Ptp2 interaction.

MATERIALS AND METHODS

Yeast strains.

The Saccharomyces cerevisiae strains used in this study are listed in Table 1.

TABLE 1.

Yeast strains used in this study

Strain	Genotype	Source (reference)
TM100	MATaura3 leu2 trp1	19
TM141	MATaura3 leu2 trp1 his3	27
TM157	MATaura3 leu2 trp1 ptp2::hisG	T. Maeda
TM261	MATα ura3 leu2 his3 pbs2::LEU2	18
QG137	MATaura3 leu2 trp1 his3 hog1::LEU2	39
QG144	MATaade2 ura3 leu2 trp1 his3 ptp2::hisG ptp3::HIS3	Q. Ge
KY458	MATaura3 leu2 trp1 his3 HOG1-4A	K. Yamamoto
SW110	MATaura3 leu2 trp1 ptp2::hisG hog1::LEU2	S. M. Wurgler- Murphy
YM105	MATaura3 leu2 trp1 his3 hog1::LEU2 pbs2::kanMX6	Y. Murakami
YM109	MATaura3 leu2 trp1 ptp2::hisG hog1::LEU2 pbs2::kanMX6	Y. Murakami
YM112	MATaura3 leu2 trp1 his3 hog1::LEU2 PBS2Δ(102-230)	Y. Murakami

Buffers and media.

Standard yeast medium and genetic procedures were as described previously (31). SRaf medium consists of 6.7 g/liter yeast nitrogen base (Sigma) and 20 g/liter raffinose with the appropriate yeast synthetic drop-out medium supplements. CAD medium consists of 6.7 g/liter yeast nitrogen base, 20 g/liter glucose, 5 g/liter Casamino Acids (BD), and when required, 40 mg/liter adenine, 40 mg/liter tryptophan, and/or 20 mg/liter uracil. Buffer A contains 50 mM Tris-HCl (pH 7.5), 15 mM EDTA, 15 mM EGTA, 2 mM dithiothreitol, 0.2% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 5 μg/ml leupeptin, 50 mM NaF, 25 mM β-glycerophosphate, and 150 mM NaCl. Z buffer contains 50 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, adjusted to pH 7.0. Lysis buffer for mammalian cells contains 20 mM Tris-HCl (pH 7.5), 1% Triton X-100, 10% glycerol, 137 mM NaCl, 2 mM EDTA, and immediately before use, 50 mM β-glycerophosphate, 2 mM sodium orthovanadate, 10 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 10 μg/ml leupeptin, and 10 μg/ml aprotinin were added to lysis buffer. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer (1×) contains 60 mM Tris-HCl (pH 6.8), 2% SDS, 700 mM 2-mercaptoethanol, 10% glycerol, and 25 μg/ml bromophenol blue.

Vector plasmids.

p426TEG1 (P_TEF2-GST URA3⁺ 2μm) is a multicopy vector that allows a constitutive expression of glutathione S-transferase (GST)-fusion proteins using the strong TEF2 promoter (27). p426GAG1 (P_GAL1-GST URA3⁺ 2μm) is a multicopy vector that allows a galactose-inducible expression of GST-fusion proteins using the GAL1 promoter (34). YCpIF15 and YCpIF16 (P_GAL1-HA TRP1⁺ CEN4) are single-copy vectors that allow a galactose-inducible expression of hemagglutinin (HA)-tagged proteins using the GAL1 promoter (12).

Expression plasmids.

p426TEG1-Hog1 (pYM28) was constructed for the constitutive expression of GST-tagged Hog1 (GST-Hog1-FL) by fusing the HOG1 open reading frame (ORF) to the C terminus of GST in p426TEG1. p426GAG1-Rck2 (pYM215) was constructed for the galactose-inducible expression of GST-tagged Rck2 (GST-Rck2) by fusing the RCK2 ORF to the C terminus of GST in p426GAG1. pGAL1-HA-Rck2 (pYM94) was constructed for the galactose-inducible expression of HA-tagged Rck2 by fusing the RCK2 ORF to the C terminus of the HA coding sequence in YCpIF16. An expression plasmid for green fluorescent protein (GFP)-tagged Hog1 (pHog1-GFP is pVR65Trp), in which the GFP ORF is fused to the C terminus of the HOG1 ORF, has been described previously (29). Galactose-inducible expression plasmids for GST-tagged Ptp2 (pGAL1-GST-Ptp2 is pSWM36) and Ptp2-C666S (Ptp2-C/S) (pGAL1-GST-Ptp2-C/S is pSWM37), in which the GST ORF is fused to the N terminus of the PTP2 ORF in a multicopy vector, were described previously (42). A galactose-inducible expression plasmid for HA-tagged Ptp2 (pGAL1-HA-Ptp2 is pSWM38) is based on the YCpIF15 vector (S. Wurgler-Murphy and H. Saito, unpublished). An expression plasmid for HA-tagged Pbs2-K389M [p423GAL1-Pbs2(K/M)-HA is pKT40], in which the galactose-inducible GAL1 promoter, the PBS2 ORF, and a sequence encoding the HA epitope are fused in this order and placed in the multicopy vector pRS423, has been described previously (37). The HOG-specific reporter plasmid pKY57 (8xCRE-lacZ TRP1 CEN) has been also described previously (39). Deletion and missense mutants were constructed by a PCR-based method and/or by oligonucleotide-based mutagenesis. All mutations were confirmed by nucleotide sequence determination.

Detection of phosphorylated Hog1.

Cells were grown in yeast-peptone-dextrose (YPD) or CAD medium until an optical density at 600 nm (OD₆₀₀) of ∼0.5 was attained. NaCl was added to a final concentration of 0.4 M, and the cells were harvested at the indicated times, collected, and frozen in liquid nitrogen. Cells were suspended in SDS loading buffer, immediately boiled for 5 min, and subjected to SDS-PAGE. Activated Hog1, doubly phosphorylated at Thr-174 and Tyr-176, was detected by immunoblotting the total lysate with anti-phospho-p38 antibody (Cell Signaling) suspended in the immunoreaction enhancer solution Can Get Signal (Toyobo). For the experiments using wild-type Hog1, expression levels were examined by reblotting the membranes with anti-Hog1 C-terminal antibody yC-20 (Santa Cruz Biotechnology). Hog1 C-terminal deletion mutants, which do not react with yC-20, were probed with the anti-Hog1 antibody y-145 (Santa Cruz Biotechnology). GST-Hog1 was detected by anti-GST antibody B-14 (Santa Cruz Biotechnology). Immunoblots were developed with horseradish peroxidase-conjugated whole antibody and an enhanced chemiluminescence reagent (GE Healthcare). Probes were removed from membranes by using WB stripping solution (Nacalai Tesque) before the membranes were reprobed with a different antibody.

In vivo immunocoprecipitation assay.

Cells were grown in SRaf medium until an OD₆₀₀ of ∼0.5 was attained. Galactose was then added to a final concentration of 2%, and the incubation was continued for an additional 4 h. The cells were collected, suspended in ice-cold lysis buffer A, and immediately frozen in liquid nitrogen. Cell extracts were prepared at 4°C by vortexing cell suspensions vigorously with glass beads and were collected by centrifugation at 9,000 × g for 10 min. For in vivo coprecipitation assays, cell extracts were incubated with glutathione-Sepharose beads (Pierce) in buffer A for 3 h with gentle rotation at 4°C. The precipitates were washed five times with ice-cold buffer A, resuspended in SDS loading buffer, and boiled for 5 min. In typical experiments using HA epitope-tagged and GST-tagged proteins, immunoblotting analysis was performed first with the anti-HA monoclonal antibody 12CA5 (Roche). After development, the membrane was stripped and reprobed with the anti-GST antibody.

The HOG-specific 8x CRE-lacZ reporter assay.

At least three independent single colonies were freshly grown in CAD medium until an OD₆₀₀ of ∼0.5 was attained. Cells were collected following pretreatment with (or without) 0.4 M NaCl for 30 min, washed and suspended in Z buffer, and placed in liquid nitrogen. Cell suspensions were thawed in a 37°C water bath and immediately frozen again in order to permeabilize the cells. For the assay of β-galactosidase activity, the chromogenic substrate o-nitrophenyl-β-d-galactoside was added in excess, and the extracts were incubated at 37°C until a mid-yellow color had developed. The reactions were stopped by the addition of 1 M Na₂CO₃. Cell supernatants were collected by centrifugation at 18,000 × g for 10 min at 4°C. β-Galactosidase activities were calculated by determining the OD₄₂₀ of the supernatants and the OD₆₀₀ of cell cultures (22).

Tissue culture, transient transfection, and immunoprecipitation.

COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, l-glutamate, penicillin, and streptomycin. For the transient transfection assays, cells were grown in 35-mm dishes and transfected with appropriate expression plasmids using the Effectene transfection reagent (Qiagen). Cell lysates were prepared in lysis buffer with 0.5% deoxycholate (23) and incubated with anti-FLAG antibody (M2) for 2 h at 4°C with gentle rotation. Immune complexes were mixed with protein G-Sepharose beads and incubated for an additional 1 h at 4°C with gentle rotation. The beads were washed three times with lysis buffer with 10% deoxycholate and subjected to SDS-PAGE.

RESULTS

Mapping of an Hog1-binding domain in Pbs2.

Previously, we showed that the yeast Hog1 MAPK binds the Pbs2 MAPKK through as-yet-undefined binding sites (26). In order to identify the region in Pbs2 that is necessary for its binding to Hog1, we constructed plasmids that expressed HA-tagged full-length Pbs2 protein (Pbs2-HA) or various Pbs2 deletion mutants under the control of the inducible GAL1 promoter (see Fig. 1A for a schematic representation of the various Pbs2 constructs). Each of the Pbs2-HA plasmids was introduced into a pbs2Δ host strain together with a second plasmid that constitutively expressed either GST-tagged full-length Hog1 (GST-Hog1) or GST alone. The expression of Pbs2-HA was induced by galactose for 4 h before cell lysates were prepared. GST-Hog1 in the cell lysate was precipitated using glutathione-Sepharose beads, and coprecipitated Pbs2-HA was detected by immunoblotting with an anti-HA antibody.

FIG. 1.

Mapping of HBD-1 in Pbs2. (A) A schematic diagram of full-length Pbs2 (Pbs2-FL) and its various deletion derivatives. The kinase catalytic domain is indicated by gray shading. The ability of the constructs to bind full-length Hog1 in in vivo coprecipitation assays (shown in panel B) is indicated in the column at the right (+, binds Hog1; −, does not bind Hog1). (B) Results of in vivo coprecipitation assay to detect Hog1-Pbs2 binding. A plasmid expressing GST-Hog1 (+) or control GST (−) under the control of the constitutive TEF2 promoter (p426TEG1-Hog1 or p426TEG1, respectively) was cotransformed with another plasmid that expresses HA-tagged full-length Pbs2 (p423GAL1-Pbs2-HA) or its deletion derivatives under the control of the GAL1 promoter into the TM261 host strain (pbs2Δ mutant). Expression of the Pbs2-HA proteins was induced with galactose for 4 h. GST or GST-Hog1 was precipitated from cell lysates using glutathione-Sepharose beads as described in Materials and Methods. Coprecipitated Pbs2-HA proteins were detected by immunoblotting using an anti-HA antibody as shown in the top panels. The middle panels indicate the expression of Pbs2-HA in the total extract. The bottom panels show the levels of GST and GST-Hog1 in the precipitate. IP, immunoprecipitation; IB, immunoblotting.

Consistent with the previous observation, full-length Pbs2 binds Hog1 (Fig. 1B, lanes 1 and 2). Deletion of the Pbs2 N-terminal segment up to the 107th amino acid [Pbs2Δ(5-107)] did not affect its binding to Hog1 (Fig. 1B, lanes 3 to 6). In contrast, Pbs2Δ(5-150) completely failed to bind the coexpressed GST-Hog1 (Fig. 1B, lanes 7 to 8). Of the two short deletion mutations within this region, Pbs2Δ(107-150) and Pbs2Δ(107-135), the former could not bind Hog1, while the latter could bind it efficiently (Fig. 1B, lanes 11 to 14). Thus, at least residues 136 to 150 of Pbs2 are required for Hog1 binding. A set of Pbs2 C-terminal deletion mutations showed that residues beyond the 246th, including the kinase catalytic domain, were not required for association between Pbs2 and Hog1 (Fig. 1B, lanes 17 to 22). Thus, the Pbs2 kinase domain is not required for stable Hog1-Pbs2 association. In contrast, Pbs2Δ(203-648) cannot bind Hog1, indicating that Pbs2 residues 203 to 245 are also required for stable binding of Pbs2 to Hog1 (Fig. 1B, lanes 23 and 24). Consistent with these data, two additional partial deletions, namely, Pbs2Δ(102-230) and Pbs2Δ(210-230), are both incapable of binding Hog1, whereas a Pbs2Δ(5-107)Δ(246-648) double-deletion mutant that contains only the residues 108 to 245 could still stably bind Hog1 (Fig. 2A and 2B). From these data, we conclude that a segment encompassing residues 136 to 245 of Pbs2 contains a region that strongly interacts with Hog1, which we have named Hog1-binding domain 1 (HBD-1).

FIG. 2.

HBD-1 is required for optimal activation of Hog1 by osmostress. (A) A schematic diagram of additional Pbs2 deletion constructs. Their abilities to bind full-length GST-Hog1 in in vivo coprecipitation assays (shown in panel B) are indicated in the column at the right (+, binds GST-Hog1; −, does not bind GST-Hog1). (B) Results of in vivo coprecipitation assays of Hog1-Pbs2 binding. See the legend for Fig. 1B for details. (C) Results of time course analysis of Hog1 phosphorylation in a matching pair of strains comprised of an hog1Δ PBS2⁺ strain (QG137) and an hog1Δ PBS2Δ(102-230) strain (YM112) following osmostress stimulation. Both strains carry a HoG1 plasmid. Cells were collected at the indicated times after the addition of 0.4 M NaCl, and Hog1 activation (phosphorylation) was examined by immunoblotting the total lysate with an anti-phospho-p38 antibody which cross-reacts with phospho-Hog1 (P-Hog1) (37). Lanes 1 to 6 and 7 to 12 are directly comparable as they are from the same blot. The total amount of Hog1 protein in the lysates is also shown. (Pbs2 FL, full-length Pbs2.)

The Pbs2 HBD-1 domain is required for optimal activation of Hog1.

Osmostress activates Pbs2, which then activates the Hog1 MAPK by phosphorylating Thr-174 and Tyr-176 in the activation loop. We therefore determined if the Pbs2 HBD-1 domain might be important for the functional interaction of Pbs2 and Hog1 following osmotic stress. In wild-type cells, doubly phosphorylated Hog1 is detectable within 5 min of osmostress and returns to the baseline level by 30 min, as shown in Fig. 2C (upper panel). In PBS2Δ(102-230) mutant cells, Hog1 phosphorylation after the same osmostress is substantially weaker and of shorter duration (Fig. 2C, lower panel), even though the expression level of Pbs2Δ(102-230) is comparable to that of wild-type Pbs2. Thus, the loss of a stable Pbs2-Hog1 interaction mediated by HBD-1 results in a reduced Hog1 activation.

A Pbs2-binding domain in Hog1 that is distinct from the CD domain.

Next, we examined the binding site in Hog1 that is responsible for its stable interaction with Pbs2. A strong candidate for such a binding site is the CD domain located immediately to the C-terminal side of the Hog1 kinase domain (Fig. 3A and B). A MAPK CD domain is a cluster of negatively charged amino acids that functions as a binding site for multiple MAPK-interacting proteins, including MAPK activators (i.e., MAPKKs), MAPK inactivators (i.e., phosphatases), and downstream substrates, such as MAPK-activated protein kinases. According to an alignment of MAPK sequences, the Hog1 CD domain is composed of residues 302 to 316 (35). To examine the possible contribution of the CD domain of Hog1 to Pbs2 binding, we generated a Hog1 mutant (D307A/D310A, abbreviated as DADA) in which two critical acidic amino acids (Asp-307 and Asp-310) were replaced by alanine. As shown in Fig. 3C, Hog1-DADA failed to bind Pbs2, indicating that the Hog1 CD domain is indeed required for interaction between Hog1 and its activator Pbs2.

FIG. 3.

Pbs2-binding domains (CD and PBD-2) in Hog1. (A) A schematic diagram of the Hog1 MAPK. The locations of the kinase catalytic domain, the CD domain, and PBD-2 are shown. (B) The Hog1 amino acid sequence surrounding the CD domain. The C-terminal end of the kinase domain and the N-terminal end of the PBD-2 domain are also shown. Amino acid substitutions in the CD domain mutant D307A/D310A (DADA) are indicated. (C) Results of in vivo coprecipitation assay for Hog1-Pbs2 binding. p423GAL1-Pbs2-HA was cotransformed into a pbs2Δ host strain (TM261) together with a plasmid encoding either GST alone, full-length GST-Hog1, or GST-Hog1-DADA. Expression and detection of the tagged proteins were as described in the legend for Fig. 1B. (D) A schematic diagram of full-length Hog1 (Hog1 FL) and its various deletion derivatives. The kinase catalytic domain, the CD domain, and the PBD-2 domain are shown by different shadings. The abilities of the Hog1 constructs to bind full-length Pbs2 in in vivo coprecipitation assays (shown in panel E and also in Fig. 4A) are indicated in the column at the right (+, binds Pbs2; −, does not bind Pbs2). (E) Results of in vivo coprecipitation assays to determine the Pbs2 binding site in Hog1. A pbs2Δ host strain (TM261) was cotransformed with p423GAL1-Pbs2-HA and a plasmid encoding GST alone, full-length GST-Hog1 (GST-Hog1 FL), or its deletion derivatives. Expression and detection of tagged proteins were as described in the legend for Fig. 1B. IP, immunoprecipitation; IB, immunoblotting.

To examine if any other part of Hog1 can contribute to its association with Pbs2, we constructed various deletion derivatives of GST-Hog1 (Fig. 3D). These constructs were individually coexpressed with the full-length HA-tagged Pbs2 construct (Pbs2-HA) in a pbs2Δ strain. By the coprecipitation assay results, it is clear that the Hog1 kinase domain is unnecessary for stable binding to Pbs2 (Fig. 3E). Thus, Hog1Δ(1-271), in which most of the Hog1 kinase domain was deleted, still bound Pbs2 as efficiently as did the full-length Hog1 (Fig. 3E, lane 9). We further found that a Hog1 deletion mutant that lacks the most-C-terminal 77 amino acids [Hog1Δ(359-435)] could also bind strongly to Pbs2 (Fig. 3E, lane 4). In contrast, a Hog1 deletion mutant that lacks the C-terminal 99 amino acids [Hog1Δ(337-435)] but retains the entire CD domain (residues 302 to 316) completely failed to bind Pbs2 (Fig. 3E, lane 5). We therefore further examined the coprecipitation with Pbs2 of a fragment that completely lacks the CD domain, namely, Hog1(320-350). As shown in Fig. 4A, Hog1(320-350) efficiently bound to Pbs2. Thus, Hog1 appears to have two adjacent binding sites for Pbs2: the CD domain (residues 302 to 316) and a novel site that resides within residues 320 to 350. We have named the latter region the Pbs2-binding domain 2 (PBD-2), with the CD domain being PBD-1.

FIG. 4.

Hog1 PBD-2 binds Pbs2 HBD-1. (A) Results of in vivo coprecipitation assay. A yeast pbs2Δ strain (TM261) was transformed with p423GAL1-Pbs2-HA together with a plasmid encoding GST alone or GST-Hog1(320-350) under the control of the constitutive TEF2 promoter. Expression of the Pbs2-HA proteins was induced with 2% galactose for 4 h. GST-Hog1(320-350) or GST was precipitated from cell lysates, and coprecipitated Pbs2-HA was detected by immunoblotting, as described in the legend for Fig. 1B. The bottom panel shows the levels of GST and GST-Hog1(320-350) in the precipitate. (B) Results of in vitro pull-down assay. A pbs2Δ strain (TM261) was separately transformed with a plasmid encoding GST-Hog1(320-350) under the control of the constitutive TEF2 promoter, a plasmid that encodes Pbs2-HA, or the indicated deletion derivatives. The extract containing GST-Hog1(320-350) was mixed with a second extract containing the Pbs2-HA protein. GST-Hog1(320-350) was affinity purified from the mixture by using glutathione beads, and after extensive washing, the bound Pbs2-HA was detected by immunoblotting. The bottom panel shows the levels of GST-Hog1(320-350) bound to the beads. (C) Overexpression of the Hog1 PBD-2 domain is sufficient to inhibit activation of endogenous Hog1 following osmostress. The wild-type yeast strain TM100 was transformed with a plasmid encoding either GST alone or GST-Hog1(320-350) under the control of the constitutive TEF2 promoter. The cells were treated with 0.4 M NaCl for the indicated times before preparation of the cell lysates. Phosphorylated Hog1 (P-Hog1), total Hog1, and the expression levels of GST and GST-Hog1(320-350) were probed by immunoblotting. (D) A plasmid expressing GST-Pbs2-K389M under the control of the constitutive TEF1 promoter [pTEG1-Pbs2(K/M)] was cotransformed with pHog1-GFP or its DADA or ΔPBD-2 derivative into a pbs2Δ hog1Δ strain (YM105). GST-Pbs2 was precipitated from cell lysates, and coprecipitated Hog1-GFP proteins were detected by immunoblotting (top panel). The expression of the Hog1-GFP proteins in the total extract and the levels of GST-Pbs2 in the precipitates are also shown. Lanes 2 and 3 are duplicates. IP, immunoprecipitation; IB, immunoblotting.

The Hog1 PBD-2 docking site binds the Pbs2 HBD-1 docking site.

To determine the region in Pbs2 that binds the Hog1 PBD-2 site, a cell extract containing GST-tagged Hog1(320-350) and another cell lysate containing HA-tagged Pbs2 (or one of its deletion derivatives) were mixed in vitro. Following GST affinity purification, coprecipitated HA-Pbs2 was detected by immunoblotting (Fig. 4B). Fig. 4B, lane 1, containing a positive control, shows that Hog1(320-350) binds tightly to full-length Pbs2. The deletion of residues 200 to 281 or residues 280 to 353 from Pbs2 did not abolish its binding to Hog1(320-350) (lanes 2 and 3). In clear contrast, the deletion of residues 102 to 230 from Pbs2 completely prevented the binding of Pbs2 to Hog1 PBD-2 (lane 4). Thus, the binding target of Hog1 PBD-2 is located within residues 102 to 200 of Pbs2, which coincides with the region we have defined as HBD-1 (Fig. 2A). We next tested whether the overexpression of the PBD-2 domain alone [Hog1(320-350)] could compete with endogenous Hog1 for binding to Pbs2. This indeed seems to be the case, as the expression of GST-Hog1(320-350) significantly suppressed the osmostress activation of endogenous Hog1 (Fig. 4C).

These results, indicating that the PBD-2 domain of Hog1 binds Pbs2, might appear to be at odds with the earlier observation that the CD mutant (Hog1-DADA), which contains the intact PBD-2 site, does not significantly bind Pbs2. Indeed, the Hog1ΔPBD-2 mutant, namely, the Δ(320-350) protein, could bind Pbs2 nearly as well as the wild-type Hog1 protein (Fig. 4D). A possible resolution to this conflict is that the binding capacity of PBD-2 is relatively weak in the context of full-length Hog1 because of a steric hindrance by other domains of Hog1. When Pbs2 binds to Hog1 through the CD domain, this might induce a conformational change in Hog1, thereby exposing PBD-2 for interaction with Pbs2. If so, the association of Pbs2 with Hog1 via the CD domain might be reinforced by an additional interaction with PBD-2. The results of functional analyses shown below are consistent with this interpretation.

Both the CD and the PBD-2 domains contribute to Hog1 activation by Pbs2.

To study the roles of the Hog1 CD and PBD-2 domains in the osmotic activation of the HOG pathway, we constructed three HOG1 mutants that were defective in the CD site (DADA), the PBD-2 site [the ΔPBD-2 mutant has the Δ(320-350) mutation], or both (DADA ΔPBD-2). Then, we examined the osmostress-induced activation of the HOG pathway by three methods: phosphorylation of the Hog1 activation loop (6, 19), transcriptional activation of the HOG-specific 8xCRE-lacZ reporter gene (39), and the ability to grow on high-osmolarity medium (19).

In wild-type cells, Hog1 is rapidly activated (phosphorylated) upon osmostress (0.4 M NaCl) and is gradually dephosphorylated over the next 20 min (Fig. 5A, lanes 1 to 4). The CD mutant (Hog1-DADA) is also rapidly activated. But, unlike wild-type Hog1, it remains activated even after 20 min of osmostress (lanes 5 to 8). The PBD-2 mutant (Hog1ΔPBD-2) is more weakly activated, but it also remains phosphorylated after 20 min of osmostress (lanes 9 to 12). In clear contrast, the Hog1-DADA ΔPBD-2 double mutant was not activated at all (lanes 13 to 16). The results of HOG-specific reporter assays (Fig. 5B) were consistent with the results of the Hog1 phosphorylation assays. The 8xCRE-lacZ reporter is more strongly induced in HOG1-DADA mutants than in wild-type cells, reflecting the lack of down-regulation in the DADA mutant cells. The weaker induction of 8xCRE-lacZ in HOG1ΔPBD-2 mutants is also consistent with the weaker Hog1 phosphorylation. There is no detectable induction of the 8xCRE-lacZ reporter in the HOG1-DADA ΔPBD-2 double-mutant cells, reflecting the lack of Hog1 activation seen in the results shown in Fig. 5A. Finally, the cell growth of the mutants on high-osmolarity medium also confirmed the results of the phosphorylation and reporter assays. The HOG1-DADA ΔPBD-2 double-mutant cells were severely osmosensitive, whereas the HOG1-DADA and HOG1ΔPBD-2 single mutants were osmoresistant (Fig. 5C). Clearly, the Hog1-DADA ΔPBD-2 double-mutant protein is incapable of being activated by Pbs2.

FIG. 5.

Effects of the CD and PBD-2 mutations on Hog1 activation by osmostress. (A) Effects of HOG1 mutations on phosphorylation of the Hog1 MAPK upon osmostress stimulation. Yeast strain YM105 (hog1Δ pbs2Δ) carrying pRS414-PBS2 (a single-copy plasmid that expresses the wild-type PBS2 gene from its own promoter) was transformed with another single-copy plasmid that expresses the wild-type (WT) HOG1 gene (pRS416-HOG1) or the DADA, ΔPBD-2, or DADA/ΔPBD-2 mutant HOG1 genes. The cells were treated with 0.4 M NaCl for the indicated times before preparation of the cell lysates. Phosphorylated Hog1 (P-Hog1) and total Hog1 were probed by immunoblotting. (B) Effects of HOG1 mutations on the expression of the HOG-specific 8xCRE-lacZ reporter gene following osmostress. The same strains described for panel A were transformed with the 8xCRE-lacZ reporter plasmid pKY57 (39). The cells were treated with 0.4 M NaCl for 30 min before preparation of cell extracts for β-galactosidase assays. Error bars show standard deviations. +, present; −, absent. (C) Osmosensitivity of the HOG1 mutants. The same strains described for panel A were plated on YPD plates in the absence or presence of an osmostressor (1 M NaCl). Plates were photographed after 4 days of incubation at 30°C.

The CD and PBD-2 domains of Hog1 differentially bind Rck2 and Ptp2.

Because Hog1-DADA interacts only weakly with Pbs2, it might be expected that the phosphorylation of the Hog1-DADA protein during osmostress signaling would be weaker than the phosphorylation of the wild-type Hog1. The results shown in Fig. 5A, however, contradict this prediction. Hog1-DADA is phosphorylated nearly as strongly as wild-type Hog1 at 5 min and remains highly phosphorylated at 20 min, at which point the wild-type Hog1 is already deactivated. Hog1ΔPBD-2 also remains phosphorylated longer than wild-type Hog1. The lack of down-regulation of Hog1-DADA and Hog1ΔPBD-2 phosphorylation suggests that these mutants might also be compromised in their interaction with their specific protein phosphatases. Indeed, the CD domains of various MAPKs are known to bind specific protein phosphatases that inactivate the MAPKs (35). Thus, we investigated whether the CD and PBD-2 domains of Hog1 have any role in binding the major inactivator of Hog1, namely, the protein tyrosine phosphatase Ptp2 (21, 42, 44).

We first examined, as a control, the binding of another Hog1-interacting protein, the protein kinase Rck2. Rck2 is a homolog of the mammalian calcium/calmodulin-dependent kinases and is important for the down-regulation of protein synthesis during the early adaptation of yeast to external osmostress (3, 40). In coprecipitation assays, GST-Rck2 binds to wild-type Hog1 or to Hog1ΔPBD-2, but not to the CD mutant Hog1-DADA (Fig. 6A). The Hog1(320-350) fragment, which can bind Pbs2 efficiently, has no affinity for Rck2 at all (Fig. 6B, lane 3). Thus, for Rck2, the Hog1 CD domain seems to be the docking site.

FIG. 6.

The CD and PBD-2 domains of Hog1 differentially interact with Rck2 and Ptp2. (A) Plasmids expressing either GST-Rck2 (+) or control GST (−) under the control of the inducible GAL1 promoter (p426GAG1-Rck2 and p426GAG1, respectively) were cotransformed with a second plasmid encoding GFP-tagged full-length Hog1 (pHog1-GFP) or its DADA or ΔPBD-2 derivative into a hog1Δ strain (QG137). The expression of GST and GST-Rck2 was induced with galactose for 4 h. GST or GST-Rck2 was precipitated from cell lysates using glutathione-Sepharose beads as described in Materials and Methods. Coprecipitated Hog1-GFP proteins were detected by immunoblotting using an anti-GFP antibody (top panel). The middle panel indicates the expression of the Hog1-GFP proteins in the total extract. The bottom panel shows the levels of GST and GST-Rck2 in the precipitate. (B) The hog1Δ strain QG137 was cotransformed with a plasmid that encodes HA-tagged Rck2 (pGAL1-HA-Rck2) and a second plasmid encoding GST, GST-Hog1, or GST-Hog1(320-350) under the control of the constitutive TEF2 promoter [p426TEG1, p426TEG1-Hog1, or p426TEG1-Hog1(320-350), respectively]. The expression of HA-Rck2 was induced with galactose for 4 h. GST or GST-Hog1 fusion proteins were precipitated from cell lysates, and coprecipitated HA-Rck2 was detected by immunoblotting using an anti-HA antibody (top panel). (C) A plasmid expressing either GST-Ptp2-C/S (+) or control GST (−) under the control of the GAL1 promoter (pGAL1-GST-Ptp2-C/S and p426-GAG1, respectively) was cotransformed with pHog1-GFP or its DADA or ΔPBD-2 derivative into a ptp2Δ hog1Δ strain (SW110). The expression of GST and GST-Ptp2-C/S was induced with galactose for 4 h. GST or GST-Ptp2-C/S was precipitated from cell lysates, and coprecipitated Hog1-GFP proteins were detected as described for panel A. The bottom panel shows the levels of GST and GST-Ptp2-C/S in the precipitate. *, nonspecific band. (D) The ptp2Δ hog1Δ strain SW110 was cotransformed with pGAL1-HA-Ptp2 together with a second plasmid encoding GST alone, GST-Hog1, or GST-Hog1(320-350). The expression of HA-Ptp2 was induced with galactose for 4 h. GST or GST-Hog1 fusion proteins were precipitated from the cell lysates, and coprecipitated HA-Ptp2 was detected by immunoblotting using an anti-HA antibody (top panel). IP, immunoprecipitation; IB, immunoblotting.

Next, we examined the binding of Ptp2 to Hog1. It has been shown previously that only phosphorylated Hog1 can stably bind to Ptp2 (21, 42). However, because wild-type Ptp2 rapidly dephosphorylates bound substrate, its association with Hog1 is transient: it cannot form a stable complex with Hog1. In contrast, catalytically inactive Ptp2 mutants, such as Ptp2-C/S, can bind Hog1 more stably. The binding of wild-type Hog1 to Ptp2-C/S is shown in Fig. 6C (lanes 1 and 2). Although these cells were not osmotically stimulated, the basal activity of Pbs2, as well as the protection of phospho-Hog1 by bound Ptp2-C/S, allowed the detection of Hog1-Ptp2 binding. Unlike wild-type Hog1, however, the CD mutant Hog1-DADA could not bind Ptp2-C/S at all (Fig. 6C, lanes 3 and 4). This result is consistent with the findings of a previous report that Ptp2-C/S and Ptp3-C/S do not stably associate with another CD domain mutant, Hog1-D310N (20). Thus, the dephosphorylation of Hog1-DADA is very slow, as shown in Fig. 5A, presumably because of its inability to bind the Ptp2 (and presumably Ptp3) phosphatases.

A lack of binding, however, is not the reason for the slow dephosphorylation of Hog1ΔPBD-2, because it binds to Ptp2-C/S quite efficiently (Fig. 6C, lanes 5 and 6). Thus, like Rck2, Ptp2 appears to bind Hog1 mainly through the CD domain. We found, however, that the Hog1(320-350) fragment, which contains only the PBD-2 region, but not the CD domain, could bind either Ptp2 (Fig. 6D, lane 3) or Ptp2-C/S (data not shown). As expected, wild-type Ptp2 did not stably bind full-length Hog1 (Fig. 6D, lane 2). Thus, the PBD-2 region has an affinity for Ptp2, but this binding capacity seems to be masked in the context of full-length Hog1. Why, then, is the dephosphorylation of Hog1ΔPBD-2 so slow, if it can bind to Ptp2? We showed earlier that the CD and PBD-2 domains function synergistically in the activation of Hog1 by Pbs2 (Fig. 5). By analogy, we might expect that the two binding sites also synergistically interact with Ptp2. If so, the slower dephosphorylation of the Hog1-DADA and Hog1ΔPBD-2 mutants could be explained by their reduced affinity for Ptp2. Surprisingly, however, further analyses revealed that a more-complex interaction between Ptp2 and Hog1 is involved.

Hog1 PBD-2 mutants can bind Ptp2 in the absence of Hog1 phosphorylation.

Ptp2-C/S binds tightly either to wild-type Hog1 or to Hog1ΔPBD-2 (Fig. 7B, lanes 5 and 6). The current view is that Ptp2-C/S stably binds only to phosphorylated Hog1 and that, even in the absence of external osmostress, Pbs2 phosphorylates Hog1 at a low basal level. To confirm that Ptp2-C/S does not bind Hog1 in the complete absence of Tyr phosphorylation, we repeated the binding experiments in pbs2Δ mutant cells. As expected, the binding of Ptp2-C/S to wild-type Hog1 was much weaker in pbs2Δ than in PBS2⁺ cells (Fig. 7B, compare lanes 5 and 11). Interestingly, however, there was a significant binding of Ptp2-C/S to Hog1ΔPBD-2 even in pbs2Δ cells (Fig. 7B, lane 12), suggesting that Hog1ΔPBD-2 has a higher, not a lower, affinity to Ptp2-C/S than wild-type Hog1 does. Indeed, Hog1ΔPBD-2 can bind even wild-type Ptp2, either in PBS2⁺ cells or in pbs2Δ host cells (compare Fig. 7B, lanes 4 and 10, and Fig. 7C, lanes 6 and 12). These results indicate that Hog1ΔPBD-2 binds Ptp2 or Ptp2-C/S in a manner that is independent of phosphorylation.

FIG. 7.

Ptp2 binds the Hog1 PBD-2 domain in a phosphorylation-independent manner. (A) Amino acid sequence surrounding the CD and PBD-2 domains of Hog1. Alignment of the sequences of the ΔPBD-2 and 4A mutants with the wild-type (WT) Hog1 sequence is shown. Boldface indicates mutated amino acids. (B and C) Results of in vivo coprecipitation assays of binding between Ptp2 and Hog1. pGAL1-GST-Ptp2, pGAL1-GST-Ptp2-C/S, or a control vector for GST alone was transformed into either the SW110 (ptp2Δ hog1Δ) or the YM109 (ptp2Δ hog1Δ pbs2Δ) yeast strain. These strains were then transformed with a second plasmid, pHog1-GFP, pHog1ΔPBD-2-GFP, or pHog1-4A-GFP, as indicated in the figure. The expression of GST and GST-Ptp2 fusion proteins was induced with galactose for 4 h. GST-tagged proteins were precipitated from cell lysates using glutathione-Sepharose beads as described in Materials and Methods. Coprecipitated Hog1-GFP proteins were detected by immunoblotting using an anti-GFP antibody (top panels). The middle panels indicate the expression of the Hog1-GFP proteins in the total extract. The bottom panels show the levels of GST and GST-Ptp2 in the precipitate. *, nonspecific band. +, present; −, absent; IP, immunoprecipitation; IB, immunoblotting.

It was possible that the deletion mutation ΔPBD-2 distorted the structure of the Hog1 kinase domain, resulting in nonspecific binding of Ptp2 to Hog1ΔPBD-2. To eliminate this possibility, we made another mutant, Hog1-4A, in which four randomly chosen amino acids in PBD-2 (Asp-324, Leu-327, Trp-332, and Met-335) were mutated to alanine (Fig. 7A). The results shown in Fig. 7C show that Hog1-4A interacts with Ptp2 in a manner that is essentially identical to that of Hog1ΔPBD-2: namely, Hog1-4A can bind Ptp2 in the absence of any Hog1 phosphorylation by Pbs2 (Fig. 7C, lanes 10 and 12). Thus, both Hog1ΔPBD-2 and Hog1-4A stably bind to Ptp2 in a phosphorylation-independent manner.

Hog1 PBD-2 mutants are resistant to dephosphorylation by Ptp2.

We also found that both Hog1ΔPBD-2 and Hog1-4A show a slower dephosphorylation kinetics following osmostress than wild-type Hog1. After exposure to osmostress, the phosphorylation of wild-type Hog1 is transient: it peaks at 5 min and rapidly returns to the prestimulation level, by 30 min (Fig. 8A, lanes 1 to 6). In contrast, Hog1-4A remains phosphorylated for as long as 90 min after osmostress stimulation (Fig. 8A, lanes 7 to 12). There are two possible mechanisms to explain the slow dephosphorylation of Hog1-4A. The first potential mechanism is that Hog1-4A is defective in inducing phosphatase activities. Previously, we and others showed that activated Hog1 induces the expression and activity of the Ptp2 and Ptp3 phosphatases (17, 42). As a consequence, the dephosphorylation of the kinase-dead Hog1-K52N mutant, which can be phosphorylated by Pbs2 but has no kinase activity, is very slow in a hog1Δ host background; in a HOG1⁺ host, the dephosphorylation kinetics of Hog1-K52N is indistinguishable from that of wild-type Hog1 (42). Thus, the slow dephosphorylation of Hog1-4A might be similarly due to its inability to induce the expression and activity of its specific tyrosine phosphatases. The second potential mechanism is that Hog1-4A is more resistant to phosphatase attack than wild-type Hog1, even though this appears to contradict the finding that Hog1-4A binds Ptp2 better than wild-type Hog1 does (Fig. 7C).

FIG. 8.

The Hog1-4A mutant protein is resistant to phosphatase activity. (A) Effect of the 4A mutation on the kinetics of Hog1 phosphorylation following osmostress stimulation. The hog1Δ strain QG137 was transformed with a plasmid encoding either GST-Hog1 or GST-Hog1-4A under the control of the constitutive TEF2 promoter. The cells were then treated with 0.4 M NaCl for the indicated times before preparation of the cell lysates. Phosphorylated Hog1 (P-Hog1) was detected by immunoblotting using anti-phospho-p38 antibody (top panel). The total GST-Hog1 in the lysates was also probed by immunoblotting using anti-GST antibody (bottom panel). (B and C) The Hog1-4A mutation affects phosphorylation kinetics only in cis. In the experiment whose results are shown in panel B, a wild-type yeast strain, TM141, was transformed with a plasmid encoding GFP-tagged Hog1 (pHog1-GFP) with or without the 4A mutation, as indicated in the figure. In the experiment whose results are shown in panel C, an HOG1-4A mutant yeast strain (KY458) was transformed with pHog1-GFP with or without the 4A mutation. In both experiments, the cells were stimulated with 0.4 M NaCl for the indicated times, and phosphorylated Hog1 (P-Hog1) and phosphorylated Hog1-GFP (P-Hog1-GFP) were probed by immunoblotting. −, empty vector (pRS414). (D) Hog1-4A mutant protein is resistant to dephosphorylation by Ptp2 and Ptp3. A ptp2Δ strain (TM157) or a ptp2Δ ptp3Δ strain (QG144) was transformed with pHog1-4A-GFP. The cells were stimulated with 0.4 M NaCl for the indicated times. The phosphorylation of the endogenous Hog1-WT protein and the plasmid-encoded Hog1-4A-GFP protein was probed by immunoblotting (top panels). The expression levels of proteins in the lysates were probed by immunoblotting using anti-Hog1 antibody (bottom panels). WT, wild type; IB, immunoblotting.

To distinguish between the two possible mechanisms, we coexpressed tag-free Hog1 and GFP-tagged Hog1 constructs (with or without the 4A mutation) in the same cell. Because of the size differences, their phosphorylation status can be monitored simultaneously by immunoblotting. Figure 8B (lanes 5 to 8) shows that the phosphorylation/dephosphorylation kinetics of GFP-Hog1 is indistinguishable from that of tag-free Hog1; they are completely dephosphorylated by 30 min. In contrast, when both Hog1 and GFP-Hog1 contained the 4A mutation, they both had slow dephosphorylation kinetics (Fig. 8C, lanes 9 to 12). However, when the tag-free Hog1 was the wild type and the GFP-Hog1 had the 4A mutation, the tag-free Hog1 was rapidly dephosphorylated, whereas GFP-Hog1-4A was only slowly dephosphorylated (Fig. 8B, lanes 9 to 12). Conversely, when GFP-Hog1 was the wild type and the tag-free Hog1 had the 4A mutation, GFP-Hog1 was rapidly dephosphorylated, while Hog1-4A had slow dephosphorylation kinetics (Fig. 8C, lanes 5 to 8). In these cases, the induction of phosphatases has clearly occurred after the osmotic stimulation of the cell. Nevertheless, the 4A mutant protein (either GFP-Hog1-4A or Hog1-4A) remained phosphorylated long after the coexisting wild-type protein (Hog1 or GFP-Hog1) was dephosphorylated.

Previously, we and others have shown that Hog1 could be dephosphorylated, in the absence of Ptp2, by another tyrosine phosphatase, Ptp3 (17, 42). Thus, we examined the identity of the phosphatase(s) that was responsible for the differential dephosphorylation kinetics of wild-type Hog1 and Hog1-4A. In the absence of Ptp2 (namely, in ptp2Δ host cells), wild-type Hog1 could still be dephosphorylated as previously reported, while Hog1-4A was not dephosphorylated (Fig. 8D, left panel). In clear contrast, in the absence of both the Ptp2 and Ptp3 phosphatases, neither wild-type Hog1 nor Hog1-4A was dephosphorylated at all (Fig. 8D, right panel). These results clearly indicate that Hog1-4A is resistant to both Ptp2 and Ptp3.

We thus conclude that Hog1-4A and probably Hog1ΔPBD-2 are intrinsically resistant to attack by protein phosphatases, even though they can bind at least the Ptp2 phosphatase more avidly than wild-type Hog1. It might be that an interaction between the Hog1 PBD-2 domain and the phosphatases is required to induce a conformational change in the Hog1 activation loop so that it becomes a better substrate of Ptp2/Ptp3. The implications of these and earlier results will be considered in more detail in Discussion.

PBD-2 region in the mammalian p38 MAPK is functionally important.

To gain a further insight into the functional significance of the CD and PBD-2 docking sites, we compared the sequences of Hog1 MAPK homologs from various species to each other, as well as to non-Hog1-type MAPKs and non-MAPK serine/threonine kinases (see Fig. S1 in the supplemental material). From the sequence alignment, it can be seen that all MAPKs (both those homologous and those nonhomologous to Hog1) contain a CD domain with the characteristic D/E-X-X-D-E acidic motif. Non-MAPK Ser/Thr kinases, such as Ste20, Ste11, and Pbs2, are not homologous within the CD domain region. In contrast, a strong similarity between the PBD-2 domains is observed in Hog1-like MAPKs in species that have a Pbs2-like MAPKK, namely, budding yeast, filamentous fungi, and fission yeast. The sequence from Phe-322 through to His-344 is almost 100% identical among the six budding yeast Hog1 homologs that we compared. Residues extending from His-344, however, show strongly divergent sequences. The plant and animal Hog1 homologs have only marginal similarities in the PBD-2 region to the S. cerevisiae Hog1. The sequence analysis thus is consistent with the hypothesis that the PBD-2 domain of Hog1 is important for its functional interaction with Pbs2.

We then tested if the region of the mammalian p38 MAPK that corresponds to PBD-2 is functionally important. For this purpose, we made the truncation mutant p38α-ΔC, in which the C-terminal 34 amino acids were deleted (see Fig. S2A in the supplemental material). We also made a CD domain mutant of p38α by replacing the three critical acidic residues with Ala (p38α-3A). The activation of these p38α mutants by stresses (UV or osmostress) was examined in transfected COS-7 cells. As shown in Fig. S2B in the supplemental material, p38α-ΔC could not be activated at all by either stress, whereas the activation of the CD mutant was nearly identical to that of the wild type. Thus, the PBD-2 region of p38α is essential for its activation, even though its primary sequence is divergent from that of Hog1. It is yet to be determined whether the C terminus of p38 is involved in its docking to its activators, namely, the MKK3 and MKK6 MAPKKs.

DISCUSSION

In this report, we have defined a domain in the N-terminal noncatalytic region of the Pbs2 MAPKK that is involved in its specific binding to the osmoregulatory Hog1 MAPK. This docking domain, HBD-1, is required for the optimal activation of Hog1 by Pbs2 following osmotic stress. Previously, we showed that the N-terminal region of Pbs2 contains a docking site for the Ssk2 and Ssk22 MAPKKKs and another docking site for the Sho1 adaptor protein (residues 45 to 57 and 91 to 102, respectively) (18, 37). Because Pbs2Δ(5-107), lacking both of these docking sites, can still bind Hog1 as efficiently as full-length Pbs2, the binding of Pbs2 to Hog1 must occur independently of the interaction of Pbs2 with Ssk2/Ssk22 and Sho1. The catalytic domain of Pbs2 should also interact with Hog1, but such interaction appears to be transient and does not significantly contribute to a stable association between Hog1 and Pbs2.

We have also defined a novel docking site in Hog1 (termed PBD-2) that interacts with the Pbs2 HBD-1 domain and, together with the adjacent CD domain, is necessary for the activation of Hog1 by Pbs2. The Hog1 CD domain is required for interaction with the Hog1 activator (Pbs2), its inactivator (Ptp2), and its substrate (Rck2), whereas the Hog1 PBD-2 domain interacts only with Pbs2 and Ptp2. The primary binding site on Hog1 for both Pbs2 and Ptp2 appears to be the CD domain, because the Hog1-D307A/D310A (DADA) mutant that lacks two critical acidic residues cannot bind Pbs2 or Ptp2, even though the mutant contains the intact PBD-2 domain.

We expect that the Pbs2 HBD-1 domain interacts with both the Hog1 CD domain and the PBD-2 domain, since no other region of Pbs2 binds Hog1. The typical docking sites in MAPKKs that interact with the MAPK CD domains are short peptides with the consensus motif (R/K)₂-(X)_2-6-L/I-X-L/I (11). They are often found near the N-terminal end of MAPKKs. For example, the MAPKK Ste7 of the yeast mating pheromone pathway has two alternative docking sites near its N terminus: RRNLKGLNLNL (residues 9 to 19) and RRGIKKKLTL (residues 62 to 71) (1, 30). These Ste7 peptides bind to the Fus3 and Kss1 MAPKs with high affinities, but not at all to the Hog1 MAPK, suggesting a significant specificity difference among yeast MAPK CD domains (30). Indeed, no sequence that conforms to the (R/K)₂-(X)_2-6-L/I-X-L/I motif could be found in Pbs2. Furthermore, our attempts to further narrow down the binding site in Pbs2 that interacts with the Hog1 CD and/or PBD-2 domain have been unsuccessful. It is possible that, if a docking motif conforming to the consensus ever existed in the Pbs2 HBD-1 domain, it may have evolved beyond recognition because of restraints imposed by the dual interaction with the CD and the PBD-2 sites. It is interesting to note, therefore, that in the closely related halotolerant yeast Debaryomyces hansenii, two docking motifs (RRGMKLNL and KKPNFKLNL) are found in the Pbs2 homolog (33). Their locations, residues 178 to 185 and 241 to 249 in Dh, are consistent with the location of the HBD-1 domain in Sc, namely, residues 136 to 245. In the more-remotely related fission yeast Schizosaccharomyces pombe, residues 200 to 300 in the N-terminal noncatalytic domain of Wis1 (a homolog of Pbs2) are required for its binding to Spc1 (a homolog of Hog1) (24). Although their sequence similarity is limited, residues 200 to 300 of Wis1 might have a functional role analogous to that of the Pbs2 HBD-1 as defined in this work.

Superficially, the Hog1 PBD-2 site contributes very little to the binding of the full-length Hog1 to either Pbs2 or Ptp2, as the Hog1 CD mutant (DADA) has lost most of its capacity to bind Pbs2 and to Ptp2. However, the Hog1(320-350) fragment, which contains only the PBD-2 domain, binds stably to both Pbs2 and Ptp2. To explain this seemingly contradictory observation, we propose that PBD-2 in the full-length Hog1 molecule is not accessible for interaction unless a binding protein first attaches itself to the CD domain. Because the CD and PBD-2 domains are contiguous, it is likely that the binding of a protein to the CD site affects the accessibility of the PBD-2 site.

The induction of large conformational changes following binding to the CD domain is supported by structural data on other MAPKs. MAPKs are distinguishable from other eukaryotic Ser/Thr/Tyr protein kinase families by the presence of a segment of about 40 residues termed L16 which follows immediately after the most-C-terminal conserved feature, namely, subdomain XI (14). By sequence comparison, we can deduce that the Hog1 L16 segment consists of residues 307 to 349. Thus, both the CD domain (residues 302 to 316) and the PBD-2 domain (residues 320 to 350) are part of the L16 segment. The crystallographic structures of the extracellular signal-regulated kinase (ERK), Jun N-terminal protein kinase, and p38 subfamilies of mammalian MAPKs show that the L16 segment starts within the C-terminal lobe, climbs up along the linker region between the N- and C-terminal lobes on the other side of the ATP-binding cleft, and terminates in a helix (termed αL16 or α14) that abuts the characteristic five-strand beta sheet of the N-terminal lobe (2, 15, 41, 43, 45). Structural analyses have also shown that the binding of specific peptides to the CD domain induces large conformational changes in both the L16 segment and the activation loop, including the dual phosphorylation sites. For example, for ERK2, binding to its CD domain by a peptide derived from the hematopoietic protein tyrosine phosphatase HePTP or another peptide derived from the MAPKK MEK2 induces large conformational changes not only in the L16 segment but also in the activation loop, including the dual phosphorylation sites (46). Similar effects have also been observed for p38 and Jun N-terminal protein kinase MAPKs (8, 15).

Although the Hog1 PBD-2 domain binds both Pbs2 and Ptp2, its functional role appears to be different in each case. In the case of Pbs2, the Hog1 PBD-2 domain enhances the binding of Pbs2 to Hog1 via the CD domain. Thus, either a CD or a PBD-2 mutation, individually, does not completely prevent Hog1 activation by Pbs2, but the double mutation does. In fact, the CD mutant Hog1-DADA is more strongly activated than wild-type Hog1, even though the mutant protein interacts very weakly with Pbs2. The activation of the PBD-2 mutant, Hog1ΔPBD-2, is weaker than that of wild-type Hog1, but its activated state persists longer than that of wild-type Hog1. These apparently anomalous results are due to the superimposition of the effects of the mutations on the interaction of Hog1 with the activating kinase (Pbs2) and their effects on the inactivating phosphatase (Ptp2). Even if the Hog1 CD mutant is only weakly activated by Pbs2, it cannot be inactivated by phosphatases; the net result is an enhanced activation. This effect of the CD mutant is analogous to that of the sevenmaker gain-of-function mutants of Drosophila ERK and homologous mutations in mammalian ERK MAPKs which have significantly reduced sensitivity to MAPK phosphatases (5, 7, 10). The sevenmaker mutations (D319N in rat ERK2) are located at the residue corresponding to Hog1 D310 in the CD domain. When both CD and PBD-2 are mutated, Hog1 cannot interact with Pbs2 at all. As a consequence, Hog1-DADA ΔPBD-2 cannot be activated, and the hog1-DADA ΔPBD-2 double-mutant cells are severely osmosensitive.

Ptp2 is the protein tyrosine phosphatase that is primarily responsible for down-regulating the Hog1 kinase (17, 42). High-affinity binding of Ptp2 to Hog1 requires the phospho-tyrosine residue (phospho-Tyr-176) in the substrate, Hog1. Because the catalytically active (wild-type) Ptp2 enzyme rapidly dephosphorylates the bound substrate, it cannot stably bind to Hog1. The catalytically inactive Ptp2-C/S mutant protein, in contrast, binds phosphorylated Hog1 tightly because it cannot remove the phosphate from the substrate (42). Our analyses showed that the binding of Ptp2-C/S to Hog1 also depends on the CD domain, because the Hog1-DADA mutant protein cannot bind Ptp2-C/S at all under conditions in which wild-type Hog1 binds Ptp2-C/S strongly (Fig. 6C). The Hog1 CD domain must also be required for the transient binding of phosphorylated Hog1 to wild-type Ptp2, because the dephosphorylation of the Hog1 CD mutant protein is very slow (Fig. 5A).

It is likely that the CD domain is the primary binding site for Ptp2. However, Ptp2, as well as Ptp2-C/S, can bind the Hog1(320-350) fragment tightly. Because Hog1(320-350) does not contain the Ptp2 substrate residue (Tyr-176), it was expected that there would be no difference in its binding to Ptp2 and Ptp2-C/S. More important, the interaction between the Hog1 PBD-2 domain and Ptp2 had an unexpected effect on Hog1 dephosphorylation by Ptp2. Without a functional PBD-2 domain, Hog1 can still bind to Ptp2 (via the CD domain), but the dephosphorylation of Hog1 is severely retarded. Finally, we should note the possibility that another protein phosphatase, Ptp3, may also interact with Hog1 in a manner similar to that of Ptp2.

Taking these results together, we propose the following hypothesis for functional interaction among Pbs2, Hog1, and Ptp2. Initially, Hog1 binds Pbs2 via the CD domain (step I). A conformational change in the L16 segment of Hog1 exposes the PBD-2 site for a more-stable interaction with Pbs2 (step II). The phosphorylation of Hog1 then ensues (step III). Phosphorylated Hog1 binds Ptp2, again via the CD domain (step IV). A conformational change in the L16 segment allows for an additional interaction between phosphorylated Hog1 and Ptp2 (step V). Thus, phospho-Tyr-176 in phosphorylated Hog1 is productively aligned with the catalytic center (Cys666) of Ptp2. As the tyrosine residue is dephosphorylated, Ptp2 detaches itself from Hog1 (step VI). According to this model, the Ptp2-C/S mutant is trapped at step V, whereas wild-type Ptp2 is usually released from Hog1 in step VI. The Hog1-DADA mutant, defective in the CD site, binds neither Ptp2 nor Ptp2-C/S, as it cannot enter into step IV. In stark contrast, the Hog1ΔPBD-2 and Hog1-4A mutants, which are defective in the PBD-2 site, stably associate with Ptp2, possibly trapped at step IV. Bound Ptp2, however, cannot dephosphorylate these PBD-2 mutants (Fig. 8). Although the model requires further elaboration, it appears to capture the essential aspects of Pbs2-Hog1-Ptp2 interaction.