A conserved docking site in MEKs mediates high-affinity binding to MAP kinases and cooperates with a scaffold protein to enhance signal transmission.

The recognition of mitogen-activated protein kinases (MAPKs) by their upstream activators, MAPK/ERK kinases (MEKs), is crucial for the effective and accurate transmission of many signals. We demonstrated previously that the yeast MAPKs Kss1 and Fus3 bind with high affinity to the N terminus of the MEK Ste7, and proposed that a conserved motif in Ste7, the MAPK-docking site, mediates this interaction. Here we show that the corresponding sequences in human MEK1 and MEK2 are necessary and sufficient for the direct binding of the MAPKs ERK1 and ERK2. Mutations in MEK1, MEK2, or Ste7 that altered conserved residues in the docking site diminished binding of the cognate MAPKs. Furthermore, short peptides corresponding to the docking sites in these MEKs inhibited MEK1-mediated phosphorylation of ERK2 in vitro. In yeast cells, docking-defective alleles of Ste7 were modestly compromised in their ability to transmit the mating pheromone signal. This deficiency was dramatically enhanced when the ability of the Ste5 scaffold protein to associate with components of the MAPK cascade was also compromised. Thus, both the MEK-MAPK docking interaction and binding to the Ste5 scaffold make mutually reinforcing contributions to the efficiency of signaling by this MAPK cascade in vivo.

MAPK cascades contribute to the regulation of diverse responses, including hormone and growth factor action, cell differentiation, cell cycle progression, learning, inflammation, and stress responses, as well as tumorigenesis and other pathological processes (4). At the core of these cascades are a MAPK (also termed extracellular signal-regulated kinase, or ERK), and a MAPK/ERK kinase (MEK or MKK) that phosphorylates and thereby activates the MAPK. MEK activity can be regulated, in turn, via phosphorylation by various classes of upstream kinase (5). Activated MAPKs phosphorylate targets including transcription factors, other kinases, and other enzymes. Some MAPKs also regulate some of their targets by direct binding, rather than by (or in addition to) phosphorylation (6,7).
Both yeast and animal cells possess multiple parallel MAPK cascades. Saccharomyces cerevisiae contains four fully elaborated MAPK cascades, the mating pheromone response, filamentation invasion, cell integrity, and high osmolarity response pathways (8). These pathways, although distinct, share some components. Four mammalian MAPK pathways have been well characterized, the ERK1/2 module, the JNK module, the p38 module, and the ERK5 module (5). Other mammalian MEKs and MAPKs have been identified as well, but await assignment to a pathway. It is generally observed that MEKs in one pathway do not phosphorylate MAPKs in other pathways, but how this selectivity is achieved is not well understood.
MAPK cascades have been highly conserved throughout eukaryotic evolution. For example, human ERK1 and ERK2 are ~50% identical to the yeast mating/invasive growth MAPKs, Fus3 and Kss1. Likewise, the catalytic domains of human MEK1 and MEK2 are similar to that of the yeast mating/invasive growth MEK, Ste7. Indeed, molecular evolutionary analysis suggests that Fus3 and Kss1 are orthologs of ERK1 and ERK2, and that the yeast high osmolarity response MAPK, Hog1, is the ortholog of the mammalian JNK and p38 families (9). In fact, mammalian p38 can functionally replace Hog1 in S. cerevisiae (10), and ERK1 can substitute for Kss1 (11).
Given that MAPK cascades require the sequential action of at least three tiers of protein kinases, properly directed kinase-substrate interactions are necessary to achieve efficiency and fidelity in these signaling modules. Insights into kinase-substrate transactions have come from studies of the phosphorylation of small peptides derived from protein substrates (12,13), or selected from combinatorial libraries (14). Further understanding has come from crystal structures of such peptides bound to the catalytic clefts of protein kinases (15). However, peptides are often relatively poor substrates when compared with native proteins. Furthermore, phosphorylation site consensus sequences cannot fully account for the specificity observed in many kinase-substrate relationships. For example, all MAPKs phosphorylate substrates that have the core consensus, -(Ser/Thr)-Pro-. This sequence, by itself, cannot be the sole determinant of MAPK specificity because it is found in more than 75% of all proteins and because it is also at the core of phosphorylation sites recognized by cyclin-dependent kinases. Clearly, other factors must contribute to MAPK recognition of unique substrates. In the case of MEK recognition of MAPKs, altering the loop on MAPKs that is phosphorylated by MEKs does not result in the predicted alterations in specificity (16)(17)(18). Also, unlike most protein kinases, MEKs do not recognize peptide versions of their target site (19).
Additional contributions to protein kinase target selection are provided by "docking sites" on substrates that are separate from the phosphoacceptor residues (reviewed in Ref. 20). Moreover, some protein kinases use domains or subunits other than the catalytic domain to aid in substrate recognition (21)(22)(23). Another means to influence efficiency and fidelity of kinase-substrate transactions is via a "scaffolding" or "anchoring" protein that binds both kinase and substrate (24)(25)(26). The prototype MAPK-cascade scaffold is S. cerevisiae Ste5, which is essential for the mating pheromone response. Distinct regions of Ste5 bind the MEK Ste7 and the MAPKs Fus3/Kss1 (27). Ste5 also binds Ste11 (MEK kinase that phosphorylates and activates Ste7) (27)(28)(29) and Ste4 (β subunit of the membrane-bound Gβγ complex of the pheromone receptor-coupled G protein) (30)(31)(32). Hence, Ste5 is thought to function, first, by recruiting Ste11 to the plasma membrane, where it can be activated by the Ste20 protein kinase (which also binds Gβ) (33)(34)(35), and second, by co-localizing, sequestering, and properly organizing the component protein kinases of the mating MAPK cascade (36,37). In this second role, Ste5 is thought to enhance signal transmission from MEK kinase to MEK to MAPK, and to insulate the mating pheromone response pathway from other pathways that use some of the same components. Despite initial expectations, no metazoan homologues of Ste5 have been identified. However, analogous scaffolding functions have been ascribed to mammalian MP1 and KSR (which bind both MEK and ERK) (38,39), and to JIP1-JIP3 (which bind members of the JNK cascade) (40,41).
Current interest in docking and scaffolding stems not only from their influence on signaling specificity and insulation, but also from their role in shaping the kinetics, efficiency, and threshold properties of within pathway signal transmission (42)(43)(44). Given that similar functions have been attributed to kinase-substrate docking and to scaffolding, it seems possible that these mechanisms may act cooperatively within the cell. This has not been heretofore investigated, to our knowledge.
Here we present evidence that indicates that the MAPK-docking site of MEKs has been structurally and functionally conserved from yeast to humans. We also demonstrate that peptides corresponding to these MAPK-docking sites inhibit MEK-mediated ERK phosphorylation. Finally, we show that docking and scaffolding make mututally reinforcing, synergistic contributions to signal transmission in vivo.

Plasmids for in Vitro Transcription/Translation of Mammalian Genes
The following mammalian genes were used in this study: human MEK1 (MAP2K1; GenBank™ accession number NM_002755), human MEK2 (MAP2K2; L11285), human ERK1 (MAPK3; X60188), and rat ERK2 (M64300). To construct pGEM-MEK1, the MEK1 coding sequence was amplified by high-fidelity polymerase chain reaction (PCR) using Pfu DNA polymerase (Stratagene), primers LB111 (see Table I for the sequence of all oligonucleotides) and LB113, and template plasmid pBS-MEK1 (49). The resulting PCR product was digested with BamHI and SalI and inserted into the corresponding sites of pGEM4Z (Promega). pGEM-MEK1 Δ3-11 was constructed using the same procedure except that primer LB112 was substituted for LB111. To construct pGEM-MEK2, the MEK2 coding sequence was amplified by PCR using primers LB114 and LB116 and template plasmid pBS-MEK2 (49). The resulting PCR product was digested with EcoRI and XhoI and inserted into pGEM4Z that had been digested with EcoRI and SalI. pGEM-MEK2 Δ4-16 was constructed using the same procedure except that primer LB115 was substituted for LB114.
To construct pGEXLB, oligonucleotide LB101 annealed to LB102 was inserted into pGEX-4T-1 that had been digested with BamHI and EcoRI. This results in the replacement of an encoded Pro residue with a Gly-Gly-Gly-Gly-Gly-Ser-Gly sequence designed to facilitate the independent functioning of the GST and fusion moieties. The FUS3 coding sequence on an EcoRI-SalI fragment was excised from pGEM4Z-me-FUS3 (45) and inserted into the corresponding sites of pGEXLB, yielding pGEXLB-FUS3. The ERK2 coding sequence was amplified by PCR using as primers LB103 and LB104 and as the template plasmid pBS-ERK2 (45). The resulting PCR product was digested with BamHI and SalI and inserted into the corresponding sites of pGEXLB, yielding pGEX-GERK2. The plasmid encoding GST-ERK1 is described elsewhere (50).

Yeast Expression Plasmids for STE7 and Alleles
YCpT3-STE7 is a low copy (CEN), TRP1-containing vector, in which the STE7 open reading frame, modified by silent substitutions that introduce additional restriction sites, is expressed from the STE7 promoter. This vector, which was used as the starting point for construction of multiple alleles, was itself constructed in several steps. First, the single AatII site in plasmid YCpT-STE7 (45) was destroyed by digestion with AatII, treatment of the resulting linearized DNA fragment with the Klenow fragment of E. coli DNA polymerase I in the presence of all 4 deoxyribonucleotide triphosphates, and recircularization the resulting "filledin" product, generating YCpT2-STE7. Second, oligonucleotide LB88 annealed to LB89 was inserted into YCpT2-STE7 that had been digested with BamHI and MscI, yielding YCpT3-STE7. This resulted in the replacement of the sequence encoding residues 1-27 of Ste7 protein with a sequence encoding the same amino acids but containing silent substitutions that introduced a PstI site (spanning codons 7 and 8) and an AatII site (spanning codons 22-23). YCpT3-STE7 was able to fully complement the mating and invasive growth defects of a ste7Δ strain, and was used as the wild-type control for the experiments in this study. To construct YCpT3-STE7 9E10E , oligonucleotide LB105 annealed to LB106B was used to replace the corresponding PstI-AatII fragment of YCpT3-STE7. YCpT3-STE7 R10A , YCpT3-STE7 15A17A , and YCpT3-STE7 AAAA were constructed by the same procedure, except that oligonucleotides encoding the indicated substitutions were employed. The BamHI-PstI fragment of YCpT3-STE7 was replaced by oligonucleotide LB127 annealed to LB128, yielding YCpT3-STE7 Δ2-7 . The BamHI-AatII fragment of YCpT3-STE7 was replaced by oligonucleotide LB131 annealed to LB132, yielding YCpT3-STE7 Δ2-19 . The AatII-MscI fragment of YCpT3-STE7 was replaced by oligonucleotide LB133 annealed to LB134, yielding YCpT3-STE7 Δ20-26 . The BamHI-MscI fragment of YCpT3-STE7 was replaced by oligonucleotide LF4 annealed to LF5, yielding YCpT3-STE7 Δ2-22 .

Plasmids and Templates for in Vitro Transcription/Translation of STE7 and Alleles
The wild-type, AAAA, Δ2-7, Δ2-19, and Δ20-26 alleles of STE7 were subcloned from their YCpT3-based vectors into pGEM4Z as BamHI-HinDIII fragments. Fragments encoding the N-terminal third of these alleles (through residue 172 of wild-type STE7) were then amplified by PCR using primers SP6PL and STE7-X-172. The wild-type allele was produced as a fragment encoding the first 98 residues by PCR using primers STE7-1-X and STE7-X-98 (45). An SP6 promoter was added to this fragment in a second round of PCR using primers SP6EP and STE7-X-98, as described (45). The same procedure was used to produce N-terminal fragments of the Δ9-19, G14A, R10A, R10E, 9E10E, 15A17A, and 4E5E alleles, except that oligonucleotides encoding the indicated amino acid changes were used in place of STE7-1-X.

Transcription and Translation in Vitro
Transcription and translation reactions in vitro, the partial purification of translation products by ammonium sulfate precipitation, and the quantification of translation products were performed as described previously elsewhere (45). In general, polypeptides encoded by plasmid-borne alleles were produced using a coupled transcription/translation system (SPT3; Novagen); polypeptides encoded by PCR products were produced by sequential transcription and translation.

Binding Assays
GST fusion proteins were expressed and purified from bacteria and bound to glutathione-Sepharose (Amersham Pharmacia Biotech) as instructed by the manufacturer. GST fusion proteins bound to glutathione-Sepharose beads were quantified by comparison to known amounts of bovine serum albumin, following SDS-PAGE and staining with Coomassie Blue. Binding assays were performed as described (53). Binding reactions were analyzed by scintillation counting and SDS-PAGE. SDS-PAGE gels were first Coomassie stained (Gelcode blue, Pierce) and photographed, then treated with fluorography solution (Amplify; Amersham Pharmacia Biotech), dried down, and autoradiographed. For the experiments shown in Fig. 2, purified rat ERK2 was purchased from Calbiochem, anti-ERK2 antiserum from New England Biolabs, and purified human JNK1 (α 1 isoform, GenBank™ L26318) and anti-JNK1 antiserum from Santa Cruz Biotechnology. Binding constants were calculated as described (45), based on the known input concentrations and a determination of the amount complex formed, assuming a simple bimolecular binding reaction. A further assumption was that 100% of input molecules were active (i.e. capable of binding). We find K d values calculated in this manner to be useful for relative comparisons of the affinities of different protein-protein interactions measured using the same assay. We have noted, however, that the K d for (essentially) the same interaction, determined in two different assays (e.g. GST cosedimentation versus coimmunoprecipitation), can vary up to 20-fold.

Protein Kinase Assays
Kinase reactions (20 µl) contained 20 mM MOPS (pH 7.2), 10 mM MgCl 2 , 1 mM EGTA, 25 mM β-glycerophosphate, 1 mM sodium orthovanadate, 1 mM dithiothreitol, 1 µg (1.25 µM) of inactive mouse ERK2 (K52R mutation; New England Biolabs), 0.1 unit of active human MEK1 (Upstate Biotechnology), 50 µM ATP, 1 µCi of [γ-32 P]ATP, and the indicated concentration of peptide. ERK2 and peptide were preincubated in buffer for 10 min at 37 °C, then returned to ice prior to the addition of ATP and MEK1. Reactions were for 20 min at 30°C . Reactions were analyzed by SDS-PAGE and quantified on a PhosphorImager. The results were also verified qualitatively by immunostaining with antisera specific for the dually phosphorylated ERK isoform (New England Biolabs).

Bioassays for Pheromone Response
Mating assays were performed as described (28). Pheromone-imposed cell cycle arrest and recovery were measured in an agar diffusion (halo) bioassay as described (45). For determination of reporter gene expression, strains JCY107 or YLB105 were first transformed with the appropriate STE5 and/or STE7 expression plasmids. Plasmid YEpL-FUS1Z (containing the lacZ reporter gene driven by the FUS1 promoter) was introduced in a second transformation. Cells were grown in liquid culture, treated with 1 µM αF mating pheromone for 90 min, and harvested as described (45). For analysis of expression stimulated by the STE11-4 allele, plasmid YCpH-STE11-4 was introduced in a third transformation. Transformant colonies were harvested directly from plates after 3-4 days growth. Cell extracts were prepared and β-galactosidase activity quantified as described (45).

A Conserved Motif Near the N Termini of MEKs
Our previous study (45) demonstrated that the MAPKs Kss1 and Fus3 each bound to the MEK Ste7 in the absence of any additional yeast proteins. Other yeast MAPKs (Hog1 and Mpk1) did not bind to Ste7. The Ste7-MAPK complexes displayed a higher affinity and stability than would be expected from a prototypical enzyme-substrate interaction. Indeed, complex formation did not require active site-phosphoacceptor residue contacts, and apparently did not position the proteins in an enzyme-substrate orientation (45).
Relatively stable, high-affinity associations have been documented for other cognate MEK-MAPK pairs (54)(55)(56)(57). For the yeast Ste7-Fus3 (or Kss1) interaction, as well as for several mammalian MEK-MAPK interactions, the region of the MEK responsible for MAPK binding has been localized to the N-terminal, noncatalytic domain (45,(58)(59)(60)(61)(62)(63). In contrast to the catalytic domains, the noncatalytic domains of MEKs do not exhibit extensive sequence conservation across widely divergent species (Fig. 1A). However, within this domain of Ste7, we noted a short sequence motif (which we dubbed the "MAPK-docking site") that was conserved in seven other MEKs across a broad range of species and positioned at or very near the N termini of these (64). The more recent availability of sequence information for many additional MEKs further strengthens our original conclusion (Fig. 1B). Notable features of the MAPK-docking site motif are a cluster of at least two basic residues (Lys, Arg, rarely His) separated by a spacer of 2-6 residues from a hydrophobic-X-hydrophobic (or hydrophobic-X-hydrophobic-X-hydrophobic) sequence, where the hydrophobic residues are long-chain aliphatics (Leu, Ile, sometimes Val). Both in the spacer and in the sequence immediately C-terminal to the hydrophobic-X-hydrophobic element, there is a high propensity for the presence of Pro, Asn, and/or Gly, which are residues that are both turnforming and helix-breaking (65).

MAPK-docking Site of a MEK Is Sufficient for Binding of a Cognate MAPK
Our previous study indicated that residues 2-22 of Ste7 were necessary for the interaction of this MEK with its cognate MAPKs Kss1 and Fus3 (45). To determine whether this small region of Ste7 was sufficient for MAPK binding, the first 25 residues of Ste7 were fused to Schistosoma japonicum glutathione S-transferase (GST), and the resulting fusion protein was expressed in bacteria and purified by adsorption to glutathione-agarose (GSH beads). GST-Ste7 1-25 (or GST alone as a control) was then incubated with either Kss1 or Hog1 MAPKs that had been produced in radiolabeled form by in vitro translation. Bead-bound complexes were collected by sedimentation and analyzed by SDS-PAGE and autoradiography. Kss1 bound to GST-Ste7 1-25 , but not to GST; Hog1 did not bind detectably to either ( Fig. 2A). Fus3 also bound specifically to GST-Ste7 1-25 (data not shown). Hence, residues 1-25 of Ste7 are indeed sufficient for the specific binding of its cognate MAPKs.
To ascertain whether the putative MAPK-docking sites in human MEK1 and MEK2 were sufficient for binding to mammalian ERK1 and ERK2, the first 19 or 22 residues of MEK1 or MEK2, respectively, were also fused to GST. The resulting polypeptides were then expressed in bacteria, purified on GSH beads, and incubated with purified ERK2 or JNK1. Bead-bound complexes were sedimented and analyzed by SDS-PAGE followed by immunostaining with ERK2-or JNK1-specific antisera. Both GST-MEK1 1-19 and GST-MEK2 1-22 bound ERK2; this binding was specific because ERK2 did not bind to either GST alone or empty beads (Fig. 2B). In contrast, GST-MEK1 1-19 and GST-MEK2 1-22 did not bind JNK1 above the nonspecific background displayed by either GST alone or empty beads (Fig. 2B). ERK1 also bound specifically to GST-MEK1 1-19 and GST-MEK2 1-22 (data not shown). To achieve readily detectable ERK binding in these experiments, a higher input of the bead-bound GST-docking site fusions was required than was needed to observe specific Kss1 binding to GST-Ste7 1-25 (compare Coomassie-stained panels in Fig. 2, A and B), indicating that Kss1-Ste7 interaction is of higher affinity than the ERK2-MEK1 or ERK2-MEK2 interactions.
To delineate the contribution to the MEK-MAPK interaction of residues that are highly conserved in the MAPK-docking site motif (Fig. 1B), Lys 3 and Lys 4 in the basic cluster of MEK1 were both changed to glutamate, and this K3E/K4E double mutant was fused to GST, yielding GST-MEK1  . Similarly, a L12A/I14A double mutant of the hydrophobic-X-hydrophobic motif of MEK2 was prepared, yielding GST-MEK2 1-22ATA . Both mutant derivatives were clearly defective in MAPK binding (Fig. 2C). ERK2 binding to GST-MEK1 1-19EE was nearly abolished, and ERK2 binding to GST-MEK2 1-22ATA was substantially reduced. Furthermore, while GST-MEK1 1-19EE did not acquire the capacity to bind JNK1 (Fig. 2C, lane 4), GST-MEK2 1-22ATA exhibited trace binding to JNK1 (Fig. 2C, lane 6), suggesting that this mutation led to a decrease in binding specificity.

The MAPK-docking Site Mediates ERK1 and ERK2 Binding to MEK1 and MEK2
Having demonstrated that the MAPK-docking sites in MEK1 and MEK2 were sufficient for binding to ERK1 and ERK2, we asked if these docking sites were necessary for the binding of full-length MEK1 or MEK2 to ERK1 and ERK2. For this purpose, GST-ERK1 and GST-ERK2 were expressed in and purified from bacteria, and then incubated either with in vitro translated MEK1 or MEK2, or with MEK1 and MEK2 derivatives from which the MAPKdocking site had been deleted. MEK1 bound to GST-ERK1 and GST-ERK2 (but not to GST, as anticipated) whereas binding of the mutant lacking the docking site (MEK1 Δ3-11 ) was markedly reduced (Fig. 3A). Similarly, MEK2 bound to GST-ERK1 and GST-ERK2, but binding of MEK2 Δ4-16 was greatly diminished (Fig. 3B). We quantified formation of these complexes using a method described in detail elsewhere (Ref. 45 , see also "Experimental Procedures"). MEK2-ERK2 interaction had the highest apparent affinity (K d ≅ 9 µM); and MEK1-ERK1 interaction had a somewhat lower affinity (K d ≅ 29 µM). For all MEK-ERK pairings, deletion of the MAPK-docking site decreased binding affinity in the range 5-10-fold. Interestingly, it has been reported that the MP1 scaffold protein shows some specificity for MEK1-ERK1 over MEK2-ERK2 (38), and may thus compensate for the weaker intrinsic binding affinity of the MEK1-ERK1 pair.

Analysis of Conserved Residues Required for Docking
To delineate the sequence features in the MAPK-docking site of Ste7 that contribute to its very high affinity binding to Kss1, Fus3, ERK1, and ERK2, we introduced a number of substitution and deletion mutations into the MAPK-docking site of Ste7 (Fig. 4A). Radiolabeled full-length Ste7 bound with high affinity to GST-Fus3 (K d ≅ 0.1 µM), whereas a mutant lacking the entire MAPK-docking site (Ste7 Δ2-19 ), bound with reduced, but still readily detectable, affinity (K d ≅ 1.4 µM). This result is consistent with our previous observations suggesting that, in addition to MEK-MAPK binding mediated by the docking site, there is also a detectable interaction between the catalytic domain of Ste7 and the MAPKs, possibly mediated by active site-phosphoacceptor loop contacts (45). Therefore, to probe the specific contributions of the docking interaction, binding of the N-terminal noncatalytic domain of Ste7 (Ste7N) to GST-Fus3 was examined (Fig. 4B). Ste7N corresponded to residues 1-172 or residues 1-98, depending on the experiment (see "Experimental Procedures" for details). The results obtained were virtually indistinguishable when the longer or shorter construct (and mutant versions thereof) were compared. Like full-length Ste7, Ste7N bound with high affinity to GST-Fus3 (K d ≅ 0.2 µM), whereas deletion of the core of the MAPK-docking site (residues 9-19) greatly reduced the ability of Ste7N to bind to GST-Fus3 (Fig. 4B). A somewhat larger deletion (residues 2-19), that removed an upstream pair of basic residues in Ste7N (see Fig. 1B), eliminated essentially all of the residual binding to GST-Fus3. In contrast, a deletion (residues 20-26) that removed seven residues immediately C-terminal to the proposed MAPK-docking site caused only a slight reduction in binding affinity (Fig. 4B). This deletion removed a Pro, a Gly, and two Asn, which are residues frequently represented in the sequences in MEKs immediately Cterminal to the core docking site motif (see Fig. 1B). It is possible, however, that the next residue (Gly 27 ) partially compensated for the loss of the other turn forming, helix-breaking residues.
The roles of particular conserved residues in the MAPK-docking site were evaluated by the use of substitution mutations. Three progressively more radical substitutions of the basic doublet of the core docking site motif (R10A, R10E, and R9E/R10E) (Fig. 4A) perturbed the interaction of Ste7N with GST-Fus3 with increasing severity. Ste7N R10A exhibited a modest, but readily detectable, defect in binding (~50% of wild type); Ste7N R10E showed an even more dramatic loss of binding affinity (~25% of wild-type); and Ste7N R9E/R10E was as severely defective in binding to GST-Fus3 (~5% of wild-type) as a deletion of the entire docking site motif (Ste7N Δ9-19 ) (Fig. 4B). Substitution of Leu 15 and Leu 17 of the hydrophobic-X-hydrophobic element of the motif with Ala also substantially reduced formation of the GST-Fus3-Ste7N complex (Fig. 4B); substitution of these residues in conjunction with the basic doublet (Ste7N AAAA ) had an even more pronounced effect (Fig.  4, B and C).
To investigate sequence features of Ste7 that contribute to its high affinity binding of MAPKs (relative to MEK1 and MEK2), we looked for differences in the docking sites of these MEKs. Ste7 has a 7-residue extension (FQRKTLQ) between its initiator methionine and the basic pair in the core docking site motif, whereas MEK1 has only a single Pro and MEK2 has just two residues (Leu/Ala). Deletion of six residues in this N-terminal extension (Ste7N Δ2-7 ) greatly reduced binding to Fus3 (Fig. 4B), as did substitution of residues 2-8 with a single Pro (data not shown). We noted that this stretch of sequence, like the core docking site motif, contains a basic pair. Indeed, charge-reversal mutations at this site (Ste7N R4E/R5E ) resulted in a binding defect comparable to that observed for Ste7N Δ2-7 (Fig.  4B). However, Ste7N Δ2-7 still bound to GST-ERK2 with a higher affinity than did either MEK1 or MEK2 (data not shown), suggesting that additional variation among the MAPKdocking sites of Ste7, MEK1, and MEK2 also contribute to the observed differences in affinity.
To begin to explore the features of MAPKs that permit recognition by the MAPK-docking sites of MEKs, binding of the series of Ste7N mutants to GST-ERK2 was examined and compared with binding to GST-Fus3. The spectrum of relative binding of the different Ste7N mutants to GST-ERK2 was quite similar to that observed for GST-Fus3 (Fig. 4B). Mutants that bound relatively well to Fus3 (e.g. Ste7N G14A , Ste7N R10A , and Ste7N Δ20-26 ) also bound relatively well to ERK2, and mutants that bound poorly to Fus3 (e.g. Ste7N L15A/L17A , Ste7N Δ9-19 ) also bound poorly to ERK2. Moreover, the series of mutants in the basic cluster that showed a graded decrease in binding affinity to Fus3 (Ste7N R10A , Ste7N R10E , and Ste7N R9E/R10E ) showed the same pattern of binding to ERK2. These data indicate that the structural and chemical features of the region(s) of Fus3 and ERK2 that interact with the MAPK-docking site on MEKs must share substantial similarity.

Docking Site Peptides Inhibit MEK1 Phosphorylation of ERK2
To explore the role that MEK-MAPK docking plays in signal transmission from MEK to MAPK, we tested whether synthetic peptides corresponding to the MAPK-docking site sequence were able to act as competitors of MEK-dependent MAPK phosphorylation in vitro. Five different peptides were synthesized (Fig. 5A). Ste7 2-22 , MEK1 1-17 , and MEK2 1-20 correspond to the MAPK-docking sites of the respective MEKs. Scram7 contains the same residues as in Ste7 2-22 , but in a randomized ("scrambled") order. MEK2 1-20EEAA is identical to MEK2 1-20 , except that it carries mutations in four residues (R4E, R5E, L12A, and I14A) that are critical for MAPK binding (see Fig. 2C). Phosphorylation of catalytically inactive ERK2 by limiting amounts of MEK1 was measured in the absence and presence of increasing concentrations of peptide. All three MAPK-docking site peptides inhibited MEK1 phosphorylation of ERK2 and did so in a concentration-dependent manner (Fig. 5, B and C). Interestingly, these three peptides showed the same relative potency in inhibiting MEK1 phosphorylation of ERK2 (Ste7>MEK2>MEK1) as they did in binding to ERK2, consistent with the idea that the peptides inhibit by competing with MEK1 for ERK2 binding. The most potent peptide (Ste7 2-22 ) exhibited an IC 50 of ~10 µM. In contrast, the control peptides (Scram7 and MEK2 1-20EEAA ) had little or no effect on ERK2 phosphorylation, even at a 10fold higher concentration. To assess the specificity of this inhibition, the peptides were also tested for their ability to inhibit protein kinase A; none were able to do so (IC 50 ≫ 100 µM; data not shown).

Docking, by Itself, Plays a Minor Role in Transmission of the Mating Pheromone Signal
Analysis of the in vivo function of MEK-MAPK docking in mammalian cells is complicated by the fact that the expected phenotype of docking-defective mutants is recessive, and no cell lines nullizygous for both MEK1 and MEK2 are currently available. A further complicating factor is that MAPK cascade components participate in multiple, potentially redundant interactions. We chose, therefore, to examine the phenotype of MAPK-docking defective mutants of Ste7, the MEK required for yeast mating pheromone response. Key features of this differentiation-like process include pheromone-imposed cell-cycle arrest, pheromone-induced gene transcription, and a series of morphological and physiological changes culminating in the fusion of two haploid cells of opposite mating type to form a diploid (reviewed in Ref. 66).
To determine whether MEK-MAPK docking plays a positive role in transmission of the mating pheromone signal, several MAPK-docking defective alleles of STE7 were individually expressed under control of the endogenous STE7 promoter on a low-copy number (CEN) plasmid in a MATa ste7Δ strain. Under these conditions, all of the mutant Ste7 derivatives exhibited only a modest defect in mating pheromone response, as assessed by several types of assays. To obtain a quantitative measure of signal throughput, we monitored expression of a pheromone-inducible reporter gene (FUS1-lacZ), which is expressed at a low, but detectable, level in the absence of pheromone stimulation, but is induced about 100-fold after addition of a saturating concentration of pheromone. Dockingdefective mutants of Ste7 supported basal expression at 60% of the wild-type level. Following pheromone stimulation, docking-defective mutants of Ste7 supported induced expression at 85% of the level of expression observed for induced wild-type cells (data not shown). The docking-defective Ste7 mutants also exhibited only a slight defect in pheromone-imposed cell-cycle arrest (see below), and no significant defect in mating proficiency (see below).
We previously reported that a Ste7 Δ2-22 mutant had a moderate defect in pheromone response (45), in contrast to the rather minor defect we observed in this present study for Ste7 Δ2-19 and other docking site-defective mutants expressed in an otherwise wild-type background. Sequencing of the complete open reading frame of the Ste7 Δ2-22 construct used in our previous study (45) revealed a cloning artifact that resulted in truncation of the last three residues of the STE7 coding sequence (data not shown). We therefore reconstructed Ste7 Δ2-22 ; this new construct behaved identically to Ste7 Δ2-19 in all assays tested (data not shown).

Docking and Scaffolding Have Overlapping Functions
It seemed possible that the ability of the Ste5 scaffold protein to bind both Ste7 and Fus3 might mask any defect in signal transmission caused by ablation of the MEK-MAPK interaction mediated by the docking site. To address whether scaffolding by Ste5 might serve a partially redundant function with direct MEK-MAPK docking, it was necessary to devise an experimental strategy that would separate the scaffold function of Ste5 from its crucial role in Ste11 activation. Cytoplasmic proteins (like Ste5) that bind both a membranebound protein (such as Ste4-Ste18 = Gβγ) and another cytoplasmic protein (such as Ste11 = MEKK), and thereby recruit the latter protein to the membrane, are often referred to as adapters (25). To separate the adapter function of Ste5 from its subsequent role as a scaffold for the MAPK cascade components, we made use, first, of Ste5 point mutants with specific defects in Ste7 binding (28). It has been shown that Ste5 D746G greatly reduces, but does not completely prevent, Ste7 binding to Ste5, whereas Ste5 V763A/S861P eliminates detectable Ste7 binding to Ste5. Both Ste5 D746G and Ste5 V763A/S861P , however, exhibit undiminished binding to Ste4 and Ste11 (and to Fus3), consistent with the idea that they are specifically defective in the scaffold function of Ste5 (28).
To test whether there is functional overlap between scaffolding and docking, MAPKdocking defective alleles of STE7, or scaffolding-defective alleles of STE5, or both, were expressed from their own promoters on low-copy plasmids in a MATa ste5Δ ste7Δ doubledelete strain. These cells were then assayed for mating pheromone response. First, as a measure of the long-term effects of pheromone action, pheromone-imposed cell-cycle arrest and recovery from arrest were monitored using the standard halo bioassay (Fig. 6A). To obtain the optimal window for detecting a genetic interaction in this assay, the Ste5 D746G mutant was used because, like the Ste7 MAPK-docking site mutants, it displays only a modest signaling defect when expressed in place of Ste5 in an otherwise wild-type background. Indeed, when Ste5 D746G was coexpressed with wild-type Ste7, only a minor reduction in pheromone-imposed G 1 arrest was observed, as indicated by the roughened edges and slightly smaller diameter of the haloes (Fig. 6A). A comparable modest defect was observed in cells expressing wild-type Ste5 in combination with the docking-defective mutant Ste7 Δ2-19 . Strikingly, however, cells expressing both Ste5 D746G and Ste7 Δ2-19 displayed smaller and much more turbid haloes (Fig. 6A), indicating a marked reduction in the initial strength of pheromone-imposed arrest and/or a substantially accelerated rate of recovery (or both). Other Ste7 docking-defective mutants behaved comparably; in contrast, a control (Ste7 Δ20-26 ) behaved like wild-type Ste7 (data not shown). Thus, the scaffolding function of Ste5 acts synergistically in concert with the docking function of Ste7 (and vice versa) to achieve efficient and sustained pheromone-imposed G 1 arrest.
To determine whether docking and scaffolding likewise have mutually reinforcing functions in determining overall mating proficiency, quantitative mating assays were performed. Because mating efficiency can be measured over several orders of magnitude using this extremely sensitive assay, the Ste5 V763A/S861P mutant, which exhibits an essentially complete defect in Ste7 binding, was utilized. As demonstrated previously for ste5 and ste7 single mutants (67), a ste5Δ ste7Δ double mutant is completely sterile, and remained sterile (frequency of diploid formation <10 −7 ) when transformed with either a plasmid expressing STE5 alone (making the resulting strain functionally ste7Δ) or a plasmid expressing STE7 alone (making the resulting strain functionally ste5Δ) (data not shown). When transformed with both plasmids, however, these functionally wild-type cells mated with high efficiency, as judged by either a plate mating assay (Fig. 6B) or a quantitative mating assay (~50% diploid formation). Relative to this reconstructed wild-type situation, cells coexpressing Ste5 V763A/S861P and wild-type Ste7 mated with a reproducibly reduced efficiency, as assessed either on plates (Fig. 6B) or using the quantitative method (~12% of wild-type; Fig.  6C). In contrast, there was no statistically significant difference in mating efficiency between the reconstructed wild-type control and cells coexpressing wild-type Ste5 and any of the docking site-defective Ste7 mutants (Fig. 6C). However, all of the docking sitedefective Ste7 mutants showed a strong effect when each was coexpressed with Ste5 V763A/S861P : mating efficiency was typically reduced a further 10-20-fold over the impairment caused by Ste5 V763A/S861P alone, as judged either on plates (Fig. 6B) or using the quantitative test (Fig. 6C). Only Ste7 Δ20-26 , a mutant that has no MAPK-binding defect (Fig. 4B), behaved like wild-type Ste7 when combined with Ste5 V763A/S861P (Fig. 6B).
A way to bypass the adapter function of Ste5 for Ste11 activation is by using a constitutively hyperactive Ste11 variant (30,51). The Ste11-4 mutant contains a substitution (T596I) in the catalytic domain of Ste11 that substantially increases pheromone-independent transcription of FUS1 and other pheromone-responsive genes (51). Deletion of Ste4 (loss of Gβγ) has no effect on the degree of FUS1 expression elicited by Ste11-4; however, deletion of Ste5 reduces Ste11-4-dependent FUS1 expression severalfold (Ref. 51, see also below). The fact that loss of Ste5 reduces signaling by constitutively active Ste11 has been interpreted to reflect the role of the scaffold function of Ste5 in potentiating MEKK (Ste11) phosphorylation of MEK (Ste7), and/or MEK phosphorylation of MAPK (30).
A plasmid expressing Ste11-4 was transformed into cells containing wild type Ste7, or MAPK docking-site mutants of Ste7, and either containing or lacking Ste5. Expression of a FUS1-lacZ reporter gene was monitored in a quantitative assay. When both wild-type Ste5 and wild-type Ste7 were present, the presence of Ste11-4 resulted in a high level of reporter gene expression (Fig. 6D). Reporter gene expression was abolished in the absence of Ste7, and was reduced by 3-fold in the absence of Ste5, consistent with previous results (51). Ste7 mutants with alterations of the MAPK-docking site displayed a reduced ability to transmit the Ste11-4 signal in the presence of Ste5; reporter gene expression was about 30% of that observed for wild-type Ste7. However, these mutants exhibited a much more substantial defect, relative to wild-type Ste7, in the absence of Ste5, an additional 5-20-fold reduction in reporter gene expression (Fig. 6D and data not shown). In contrast, the Ste7 Δ20-26 mutant, which has no MAPK-docking defect in vitro, behaved just like wild-type Ste7 in either the presence or absence of Ste5. Moreover, we found that cells containing Ste5 V763A/S861P behaved identically to cells lacking Ste5 entirely (Fig. 6E), supporting the contention that Ste5 V763A/S861P is ineffective as a scaffold. Consistent with the results seen in cells entirely lacking Ste5, cells expressing both Ste5 V763A/S861P and Ste7 Δ2-19 displayed a markedly synergistic defect in transmitting the Ste11-4-dependent signal (Fig. 6E). Hence, as in the previous two experiments, the phenotype of the docking defect was enhanced by the scaffolding defect. Thus, as judged by three separate experimental approaches and using three different biological end points (cell-cycle arrest, mating, and transcriptional induction), mutants of Ste7 defective in MAPK docking displayed a substantial reduction in pheromone response only when the scaffold function of Ste5 was compromised.

Conservation of the MEK-MAPK Docking Interaction
This study investigated stable, high affinity complex formation between mammalian MEK1 and MEK2 and their MAPK targets ERK1 and ERK2, as well as between the yeast MEK Ste7 and its targets Fus3 and Kss1. For all of these interactions, a short N-terminal region (the MAPK-docking site) in these MEKs was found to be both necessary and sufficient for the formation of stable MEK-MAPK complexes. The MAPK-docking site is conserved between yeast Ste7 and human MEK1 and MEK2, and putative docking sites are also found in MEKs from organisms representative of many different phyla and even across kingdoms (Fig. 1B). Mutation or deletion of conserved residues in the docking site substantially reduced complex formation between cognate yeast and mammalian MEK-MAPK pairs. These same mutations significantly reduced the ability of Ste7 to transmit the mating pheromone signal in vivo, as revealed when the contribution of the Ste5 scaffold protein was compromised or eliminated. In addition, peptides corresponding to the MAPK-docking sites (but not mutant versions thereof) inhibited MEK1-mediated phosphorylation of ERK2 in vitro. Moreover, the Ste7-docking site was able to bind to and inhibit mammalian ERKs, and yeast Fus3 and mammalian ERK2 displayed the same relative binding affinities to a series of Ste7-docking site mutants. We conclude that MEK-MAPK docking has been structurally and functionally conserved from yeast to humans.

Function of MEK-MAPK Docking
MEKs phosphorylate their target MAPKs on a Thr and a Tyr located on a surface loop (the "activation loop") situated between conserved protein kinase subdomains VII and VIII. The MEK-MAPK docking interaction is distinct from this active site/target peptide interaction.
Our previous study demonstrated that the activation loop of Kss1 was not required for Ste7 binding, and that the C-terminal catalytic domain of Ste7 was also dispensable for highaffinity Ste7-MAPK association (45). Thus, neither the active site of the MEK, nor the target loop on the MAPK, are required for the docking interaction. As such, the docking interaction and the active site-target peptide interaction could constitute a double selection for fidelity in MEK-MAPK transactions, analogous to that proposed for some tyrosine kinase-substrate interactions (Ref. 23 and see Fig. 7A). Furthermore, the docking interaction represents an independent target site for drug development. In this regard, our finding that MAPK-docking site peptides can inhibit MEK1 phosphorylation of ERK2 in vitro is notable as a "proof of concept," because demonstrating that peptides serve as effective inhibitors often indicates the possibility for designing more pharmacologically useful, small-molecule mimetics (68,69). Surprisingly, we found that a peptide based upon the docking site of a yeast MEK (Ste7) was better at binding and inhibiting the interaction of a mammalian MEK with its target MAPK (ERK2) than were peptides from the cognate human MEKs. Thus, comparative studies of evolutionarily conserved protein-protein interactions of this sort may provide additional insights for the development of more effective drug leads.
The peptide inhibition experiments presented here support and extend previous observations that had suggested that the MEK-MAPK docking interaction increases the efficiency of MAPK activation. One line of evidence came from studies of the lethal factor (LF) of Bacillus anthracis (anthrax). LF is a protease that cleaves the MAPK-docking site sequence in both MEK1 and MEK2 between the doublet of basic residues and the hydrophobic-Xhydrophobic element (70,71), and hence would be predicted to substantially reduce the MEK-MAPK docking interaction, based on our results. Indeed, LF treatment inhibits MEK activation of MAPK in an in vitro kinase assay, and also inhibits ras-stimulated MAPK phosphorylation in NIH-3T3 cells (70). A positive role for the MEK N terminus in MAPK activation was also indicated by analysis of N-terminal truncation and substitution mutants in MEK1 (63,72,73). In many cases, however, the reported effect of these alterations on MAPK activation was not dramatic, as is discussed further below. An additional function for high-affinity MEK-MAPK binding has been demonstrated by Nishida and colleagues (58,74), who found that an N-terminal region of MEK1 that contains the MAPK-docking site, in combination with an immediately adjacent nuclear export signal (residues 33-44), functions to retain unphosphorylated ERK in the cytoplasm.

MAPK-docking Sites in Other Proteins
After our description of the nature of the MAPK-docking site motif in MEKs (64), we and others noted sequence similarity between MAPK-docking sites in MEKs and putative or identified MAPK-docking sites in MAPK substrates, phosphatases, and scaffolds 2 (20,73,75). Indeed, mutagenesis studies of MAPK-docking sites found in MEKs (this study), MAPK phosphatases (76), and transcription factors (77) have all demonstrated a requirement for a cluster of basic residues and a nearby hydrophobic-X-hydrophobic element. This situation suggests that these proteins could compete for MAPK binding, and that any individual complex may be dynamic in the cell. It further suggests that a peptide or appropriately engineered drug that disrupts the MEK-MAPK docking interaction might also block the interaction of MAPKs with downstream substrates and other partners. On the other hand, MAPKs do not recognize all their high-affinity targets in the same way. For example, Kss1 binds with high-affinity not only to Ste7 (45), but also to its target transcription factor, Ste12 (52). Mutations in Kss1 that impair binding to Ste12 have no effect on docking to Ste7 (6), demonstrating that the Ste7-and Ste12-binding sites on Kss1 are distinct. In addition, it has been demonstrated that at least one other type of independent ERK-docking site motif (FXFP) exists (75), which presumably associates with a structural element in MAPKs that is distinct from that mediating binding to the MAPK-docking site motif in MEKs. Thus, it should be possible to develop drugs that specifically block interaction of a kinase with a subset of its activators or targets. Such agents might provide more specific and less toxic means for therapeutic intervention than kinase inhibitors that target the active site. MAPK inhibitors could be efficacious in treating several pathological conditions, including inflammation, cancer, and myocardial injury (68,78).
Although the MAPK-docking site of MEKs is found near the N terminus of these enzymes, similar sequences in MAPK substrates are often found elsewhere within these polypeptides. However, our results support the idea that the class of MAPK-docking site that exists in MEKs can function modularly, because it is able to bind its cognate MAPK when the site is positioned at its native location at the N terminus (45), or at the C terminus, as we have shown here (e.g. GST-Ste7 1-25 ), or even in the middle of a polypeptide, as we have shown elsewhere (e.g. GST-Ste7 1-172 (6,45)).
The complementary site(s) on MAPKs that interact with the MAPK-docking motif of MEKs and related sequences are not completely clear. There is good evidence for the involvement of an acidic patch that is the site of the sevenmaker mutation in the Drosophila MAPK rolled (73,79). This acidic patch probably interacts with the basic cluster of the MAPKdocking site (73). Other studies, however, have implicated the N terminus of a MAPK (63), or a surface loop lying below the catalytic pocket (80). The MEK-MAPK enzyme-substrate interaction may also involve multiple regions of MAP kinases (81). Whatever region(s) of the MAPKs interact with the MAPK-docking site in MEKs, our results suggest that its properties have been well conserved, as both yeast and mammalian MAPKs exhibited the same spectrum of binding affinities when tested against various docking site mutants.

Variation in the Strength of MEK-MAPK Docking
One aspect of MEK-MAPK docking that has not been well conserved between yeast and humans is the strength of this interaction. We first characterized the Ste7-Kss1 and Ste7-Fus3 interaction by means of a co-immunoprecipitation assay using in vitro translated proteins (45). Using this assay, Ste7-ERK2 and MEK1-ERK interactions could not be detected (45). However, those experiments were performed using dilute concentrations of the interacting partners, such that associations with K d > 0.5 µM could not be detected. Indeed, as we have estimated here, the K d for interaction of full-length Ste7 with GST-ERK2 is about 1 µM, whereas binding of Ste7 to GST-Fus3 is about 10-fold stronger. Of course, the assay method used also influences these values. For example, previously we calculated the K d for Ste7-Fus3 interaction to be about 5 nM, when we used radiolabeled native (untagged) molecules, free in solution, which were then immunoprecipitated (45); however, using the cosedimentation assay described here, radiolabeled Ste7 associated with bead-bound GST-Fus3 with a K d of about 100 nM. This difference suggests that the structure of Fus3 in the GST fusion may be suboptimal for Ste7 binding (compared with native Fus3) and/or that in vitro translation of Fus3 yields a higher fraction of properly folded molecules (compared with expression and purification of the GST-Fus3 chimera from bacteria). Nonetheless, as assessed by either assay method, Ste7-Fus3 interaction was ~100-fold stronger than that of the tightest mammalian MEK-ERK complex (MEK2-ERK2). This situation is likely counterbalanced by the fact that MEK1, MEK2, ERK1, and ERK2 are present in mammalian cells at much higher concentrations than the corresponding levels of Ste7, Fus3, and Kss1 in yeast cells (82). We showed here that these species-specific differences in binding affinity seem to be due largely to changes in the MEKs, rather than in the MAPKs. In fact, the MAPK-docking site of Ste7 bound to mammalian ERK1 and ERK2 with a significantly higher affinity than did the docking sites of human MEK1 and MEK2. Presumably, therefore, differences in the precise nature of, and spacing between, the core elements in the MAPK-docking motif, and differences in the surrounding residues, can fine tune the affinity of these sites to adjust the strength of MEK-MAPK association so that it is appropriate to its physiological setting. Ste7-Fus3 complexes appear to represent a particularly high-affinity version of the MEK-MAPK docking interaction.

Functional Overlap between Docking and Scaffolding in Signal Transmission
Interaction between the active site of a protein kinase and the target phosphoacceptor residues on its substrate (Site 1, Fig. 7A) can be enhanced either by docking between other regions of the kinase and substrate (Site 2, Fig. 7A) or by the binding of both to a scaffold protein (Fig. 7B). This logic suggested that scaffolding and docking might have similar, mutually reinforcing roles in achieving efficient signal transmission (Fig. 7C). Indeed, by combining scaffolding-defective alleles of STE5 with docking-defective alleles of STE7, we have demonstrated here that scaffolding and docking do in fact have overlapping functions in vivo. We used three different means to specifically compromise the scaffold function of Ste5, and to isolate the scaffold function of Ste5 from its role as an adapter at the Ste11 (MEKK) activation step of the mating pheromone response pathway. First, we used a Ste5 point mutant with a reduced, but measurable, binding affinity for Ste7. Strains expressing this mutant alone, or MAPK-docking site mutants of Ste7 alone, showed only modest defects in pheromone response, but the double mutants exhibited a strikingly synergistic defect. Second, we used a strain expressing a double-point mutant of Ste5 that is unable to bind Ste7 and, hence, completely crippled in its scaffolding function, as we demonstrated. This mutant had a significant signaling defect, that was dramatically enhanced when combined with MAPK-docking site mutants of Ste7. Third, the scaffolding function of Ste5 was ablated by deleting the STE5 gene, and the requirement for Ste5 as an adapter in Ste11 activation was bypassed by using a constitutively active allele of Ste11. Once again, the diminished signaling caused by the absence of scaffolding or docking was substantially enhanced by the combination of both defects. Therefore, we conclude that docking and scaffolding share mutually reinforcing roles in promoting signal transmission.
The minor signaling defects we observed in yeast expressing MAPK-docking site mutants of Ste7 in an otherwise wild-type background are similar to the very modest effects of MEK1 N-terminal mutations on ERK2 activation seen in cell-based assays. For example, Tanoue et al. (73) observed less than 50% reduction in ERK2 activity in such circumstances. Hence, in mammalian cells, as in yeast, the binding of MEK and MAPK to a scaffold, such as MP1, KSR, or JIP, may compensate for loss of direct MEK-MAPK docking.

Implications of Cooperative Docking and Scaffolding
The functional overlap we observed between scaffolding and docking suggests that these two means of bringing kinase and substrate together may share common properties. First, based on our results in yeast, both clearly play a positive role in signal transmission. In this capacity, each helps to facilitate signaling, but neither is absolutely essential for achieving at least some signal throughput. Second, both may participate in insulating a given pathway from parallel pathways that share similar or identical components. Thus, the MAPK-docking site of Ste7 bound to Fus3 (and Kss1), but not Hog1, and the MAPK-docking sites in MEK1 and MEK2 bind ERK1 and ERK2, but not JNK1. Loss of Ste5 drastically reduces pheromone response, but has no effect on any other yeast MAPK pathway (8); similarly, JIP-family scaffolds bind JNKs, but not ERK2 or p38 (40). Third, both scaffolding and docking proteins can dictate subcellular localization of their associated components (25,74,83), although in some cases (e.g. Ste5) this localization function may be independent of docking or scaffolding per se (30,37). Finally, in some cases, docking and/or scaffolding could hinder signal amplification (42). Our results indicate that, in the yeast mating MAPK cascade, the net effect of docking and scaffolding is to augment signal propagation.
Similarly, Scott and Zuker (84) have proposed that the InaD scaffold protein limits signal amplification in Drosophila photoreceptor cells, yet acts overall as a positive effector of phototransduction.
It is possible to conceptualize docking and scaffolding such that one may be considered a special case of the other (compare Fig. 7, A and B). One popular model suggests that MAPK-cascade scaffolds function by making the dual phosphorylation of MAPKs by MEKs processive rather than distributive (42,43). Such a situation would have the effect of making the stimulus-response properties of the MAPK cascade less switch-like (82). Like a scaffold, it is conceivable that the MEK-MAPK docking interaction also increases the likelihood of efficient dual phosphorylation. Yet, our evidence suggests that dual phosphorylation cannot occur without prior dissociation of the high-affinity Ste7-MAPK complex, indicating nonprocessivity (45). Hence, docking and scaffolding may function to increase the local effective concentrations of kinase and substrate (85), or to hold the proteins in an enzymesubstrate orientation long enough for a relatively slow catalytic step to occur (86). If so, docking and scaffolding may not be incompatible with switch-like responses.
A strong argument could have been made that docking and scaffolding function to ensure signaling specificity and to insulate against adventitious activation, at the expense of signal flux. That is, selectivity gained by stable complex formation might need to be "paid for" by kinetics limited by the dissociation of the stable complexes and by a reduced capacity for signal amplification. It is remarkable then that most of the extant experimental evidence indicates that docking and scaffolding potentiate signal transmission. Whether docking and scaffolding also function to increase signaling specificity in vivo remains a largely untested hypothesis, although there is some evidence that Ste5 may play such a role (87). No mechanistic model of scaffolding and/or docking has received both theoretical and experimental support. Appreciation of their overlapping roles may facilitate the development of such a model.

FIG. 1. A MAPK-docking site near the N terminus of MEKs
A, schematic representation of S. cerevisiae Ste7, and human MEK1 and MEK2, showing the positions of the MAPK-docking site (solid bar) and the protein kinase catalytic core (hatched region). The percent amino acid identity between Ste7 and MEK1 and between MEK1 and MEK2 is indicated for the N-terminal and catalytic domains. Relatedness between Ste7 and MEK2 (<20 and 36% identity, respectively, in the N-terminal and catalytic domains) is similar to that between Ste7 and MEK1. Length of each protein is indicated on the right. B, alignment of the N termini of representatives of the MEK/MKK/ MAPKK family of dual-specificity protein kinases. All sequences start from the initiator methionine. Positions of the doublet of basic residues and the nonpolar residues of the

FIG. 2. The MAPK-docking site is sufficient for MAPK binding
A, Kss1, but not Hog1, binds to the MAPK-docking site of Ste7. Radiolabeled ( 35 S) Kss1 and Hog1 proteins were prepared by in vitro translation, partially purified by ammonium sulfate precipitation, and portions (5% of the amount added in the binding reactions; input) were resolved on a 10% SDS-polyacrylamide gel (lanes 1 and 2). Samples (1 pmol) of the same proteins were incubated with 1 or 10 µg of GST (lanes 3-6) or GST-Ste7 1-25 (lanes 7-10), bound to glutathione-Sepharose (GSH) beads, and the resulting bead-bound protein complexes were isolated by sedimentation and resolved by 10% SDS-PAGE on the same gel. The gel was analyzed by staining with Coomassie Blue for visualization of the bound GST fusion protein (lower panel), and by autoradiography for visualization of the bound

FIG. 3. High-affinity interaction between full-length MEK and ERK requires the MAPKdocking site
A, MEK1, but not MEK1 Δ3-11 , binds to ERK1 and ERK2. Radiolabeled ( 35 S) MEK1 and MEK1 Δ3-11 proteins (~1 pmol) were incubated with 10 µg of purified GST, GST-ERK1, or GST-ERK2 that were pre-bound to GSH beads, and bead-bound protein complexes were isolated and analyzed as in Fig. 2A. Inputs (5% of the amount added in the binding reactions) are shown in lanes 1 and 2. B, MEK2, but not MEK2 Δ4-16 , binds to ERK1 and ERK2. Experimental design as in A, except that the radiolabeled proteins were MEK2 and MEK2 Δ4-16 , as indicated.

FIG. 5. Inhibition of MEK1-dependent phosphorylation of ERK2 by MAPK-docking site peptides
A, peptides used. Sequences of the synthetic peptides prepared for this study are shown; substitution mutations are indicated in bold and underlined. The corresponding residues of the full-length proteins are given on the right. B and C, effect of peptides on phosphorylation of ERK2 by MEK1. Purified, full-length, catalytically inactive ERK2 (1.25 µM) was incubated with 1 unit of purified active MEK1 and [γ-32 P]ATP for 20 min, in the absence or presence of the specified concentrations of the indicated peptides. ERK2 phosphorylation was quantified by SDS-PAGE followed by analysis of relative incorporation using a PhosphorImager™ (Molecular Dynamics, Inc.). B, autoradiogram of a representative experiment. C, results plotted as percent phosphorylation relative to that observed in the absence of any added peptide. Data are the average of at least three experiments.

FIG. 7. Diagrammatic representation of interactions between MEKs, MAPKs and scaffold proteins
A, the ability of a MEK to interact with its cognate MAPK both by an active site-target residue interaction (1) and by a docking interaction (2) comprises, in principle, a "double selection" that has the potential to promote the efficiency and fidelity of MAPK activation. To aid conceptualization, these two interactions are shown as occurring simultaneously. Alternatively, they may occur in kinetically distinct steps (45,85). The region of the MAPK that is phosphorylated by the MEK is represented by a chain of five circles. B, view of scaffolding as special case of docking (or visa versa). Compare with A, the scaffold protein is indicated by an S. C, overlapping roles of docking and scaffolding in promoting MAPK cascade signaling. MAPK activation is promoted by direct MEK-MAPK binding via the active site (1) and the docking site (2), and by binding of both MEK and MAPK to a scaffold (S) protein (3). In support of the scheme shown, it is known that the catalytic domain of Ste7 binds to Ste5 (27,29), whereas the N terminus of Ste7 binds to Fus3 (this study and Refs. 29 and 45). It is likely, therefore, that Ste7 can bind to both Ste5 and Fus3 simultaneously. We have shown here that mutations in the MAPK-docking site of MEK (2), combined with mutations in the MEK-binding site (3) of the scaffold Ste5, or with deletion of the scaffold entirely, leads to a synergistic reduction in MAPK activation in the yeast mating pheromone response pathway.

Oligonucleotides used in this study
Where appropriate, restriction sites used for subcloning are underlined. Where appropriate, the first codon of the MEK1, MEK2, ERK2, or STE7 coding sequence is shown in bold.