Characterization of an ERK-binding Domain in Microphthalmia-associated Transcription Factor and Differential Inhibition of ERK2-mediated Substrate Phosphorylation*

Efficient and specific signaling by mitogen-activated protein kinases (MAPKs) is enhanced by docking sites found on many MAPK substrates and regulators. Here we show that the MAPKs ERK1 and ERK2 form a stable complex (Kd ∼ 6 μm) with their substrate the microphthalmia-associated transcription factor (MITF). Complex formation requires a domain of MITF of ∼100 residues that is nearby, but C-terminal to, the MAPK phosphorylation site at Ser73. MITF derivatives lacking this ERK-binding domain do not bind ERK2 and are phosphorylated less efficiently by ERK2. The ERK-binding domain of MITF bears no obvious resemblance to previously characterized MAPK docking motifs; in particular, it does not contain a consensus D-site. Consistent with this, ERK2-MITF binding does not require the integrity of the CD/sevenmaker region of ERK2. Furthermore, D-site peptides, which are able to potently inhibit ERK2-mediated phosphorylation of the Elk-1 transcription factor (IC50 = 3 μm), are relatively poor inhibitors of ERK2-mediated phosphorylation of MITF, exhibiting >15-fold selectivity for inhibition of Elk-1 versus MITF. These observations demonstrate substrate-selective kinase inhibition: the possibility that small molecules that target docking interactions may be used to selectively inhibit the phosphorylation of a subset of the substrates of a kinase.

Signal transduction networks that regulate gene expression and metabolism are crucial for cell growth, differentiation, and survival. Accordingly, aspects of the pathology of many diseases result from defects in signaling pathways or the transcriptional responses they regulate (1).
Many extracellular signals are transmitted to downstream targets by an evolutionarily conserved protein kinase cascade designated the mitogen-activated protein kinase (MAPK) 2 cascade. The MAPK (also known as extracellular-signal-regulated kinase, or ERK) cascade is a three-kinase module that transmits signals from cell-surface receptors to downstream targets through sequential phosphorylations (2). The MAPK is phosphorylated, and thereby activated, by an upstream MAPK/ERK kinase (MEK), which has previously been activated by a MEK kinase, such as the proto-oncoprotein Raf. To further propagate the signal, activated MAPKs phosphorylate other kinases such as RSK1 (3), which then regulate downstream effectors. In addition, MAPKs directly phosphorylate many key effectors, including the ubiquitous transcription factors Elk-1, Ets-1, and ATF-2 (4 -6).
Tissue-specific transcription factors are also targets of MAPK phosphorylation and provide cell-type specificity to MAPK-regulated responses (7,8). The microphthalmia-associated transcription factor (MITF) is a tissue-specific MAPK substrate found in melanocytes. MITF is a master transcriptional regulator for melanocytes, coordinating the expression of genes involved in stem cell maintenance, differentiation, melanogenesis, and cell survival (9 -11). MITF haploinsufficiency leads to Waardenburg syndrome type IIa, which is characterized by deafness and pigmentation defects caused by melanocyte loss in the inner ear and the skin (12). In addition, dominant-negative mutations in MITF are associated with another auditory-pigmentary disorder, Teitz syndrome (13,14) (reviewed in Ref. 15). MITF is also a potential target in the treatment of skin cancer, because it may impart a selective advantage to melanoma cells at low expression levels (16,17), while having an antiproliferative effect at higher expression levels (18).
The ability of MAPKs to effectively and specifically recognize their substrates is enhanced by their capacity to bind to those substrates with relatively high affinity. Often, this interaction occurs via MAPK-docking sites on substrate proteins (24,25). These docking sites are small regions within substrates, removed from the site of phosphorylation, that bind with relatively high affinity to MAPKs, in some cases showing specificity for particular MAPKs.
Another docking motif for MAPKs (consensus FXFP) has been named the "DEF motif" (37,38). This motif binds to a hydrophobic pocket (distinct from the one that the D-site binds to) that is located below the MAPK active site. This pocket is occluded in unphosphorylated ERK2 (35); thus, the DEF motif preferentially binds to the dually phosphorylated, activated isoform of ERK2 (35).
Protein kinases are important therapeutic targets (39 -42). Most existing kinase inhibitors bind to the active site, and thus inhibit the ability of a kinase to phosphorylate all of its substrates. The concept of kinase-substrate docking raises new possibilities for drug discovery (43)(44)(45)(46). First, molecules that target docking interactions can also be effective inhibitors, as exemplified by the ability of D-site peptides to inhibit MAPK-dependent reactions in vitro (47)(48)(49)(50)(51) and in vivo (52,53). Second, if MAPKs and other protein kinases use different parts of their surfaces to dock to different substrates, then it might be possible to selectively inhibit a subset of interactions, while leaving others relatively unimpaired. This could be advantageous in limiting toxicity and other side effects.
Here we report our characterization of an interaction between ERK1 and ERK2 and their substrate MITF and demonstrate that D-site peptides are selective for the inhibition of D-site-containing substrates, and do not effectively inhibit ERK2 phosphorylation of MITF in vitro.
Plasmids for Transfection into Mammalian Cell Lines-To construct pcDNA-MITF-V5, the full-length MITF coding sequence was amplified by high fidelity PCR (using the cDNA Image clone as template), digested with EcoRI and ApaI and ligated into the corresponding sites of the pcDNA3.1-V5 vector (Invitrogen). Similarly, pcDNA-MITF-(98 -198)-V5 was constructed by amplifying the appropriate region of MITF, and ligating the resulting PCR product into the EcoRI and ApaI sites of pcDNA3.1-V5. Plasmid pcDNA-MITF-⌬EBD-V5 (⌬98 -190) was constructed by site directed mutagenesis of pcDNA3.1-V5-MITF using primers that consisted of the DNA sequence of the regions that flank the DNA resulting in residues 98 -190. The construction of pcDNA3.1-FLAG-ERK2 has been described previously (56).
Binding Assays-In vitro transcription and translation were performed using a coupled reaction (SP3, Novagen). The translation products were partially purified by ammonium sulfate precipitation as described previously (48). GST fusion proteins were expressed in bacteria, purified, and quantified as previously described (48,57). The methods used in the co-sedimentation experiments, and the calculation of dissociation constants, have also been described previously (26,48).
Proteins and Antisera-GST-Elk-1-(307-428) and active mouse ERK2 were purchased from New England Biolabs, Inc. The anti-ERK antibody was from Santa Cruz Biotechnology. Mouse monoclonal anti-V5 antibody was obtained from Invitrogen. AlexaFluor-488-conjugated anti-mouse antibody was purchased from Molecular Probes, Inc.
Co-sedimentation of ERK from HEK293 Cell Lysates-HEK293 cells (obtained from ATCC) were cultured at 37°C in a humid atmosphere containing 5% CO 2 in air. Cells (5 ϫ 10 5 per well) were seeded into 6-well dishes and grown to confluency in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin (100 units/ml), streptomycin (100 g/ml), and 2 mM L-glutamine. Twenty hours prior to harvesting, cells were transferred to media containing 0.5% serum. For epidermal growth factor stimulations, cells were incubated with 100 ng/ml epidermal growth factor (in phosphate-buffered saline) for 5 min prior to harvesting. Cells were then scraped into 0.5 ml of lysis buffer (50 mM HEPES, pH 7.6, 150 mM NaCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1% Triton X-100, 10% glycerol) containing protease and phosphatase inhibitor cocktails (Sigma) (58). Lysates were centrifuged at 20,000 ϫ g for 5 min to remove any insoluble material, and the resultant supernatants were pre-cleared with 20 l of blocked 50% reduced glutathione beads for 30 min at 4°C. Lysates were then incubated with the specified amounts of bead-bound GST fusion protein at 4°C for 1 h. Beads were then washed twice with 0.5 ml of ice-cold cell wash buffer (20 mM HEPES, pH 7.6, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol) and resuspended in SDS-PAGE loading buffer. Samples were analyzed by Western blotting with anti-total-MAPK or anti-phospho-MAPK antibodies (Cell Signaling). Blots were then incubated with Alexa-Fluor-680/800 nm secondary antibody conjugates (Molecular Probes) and were analyzed using the Odyssey Infra-Red Imaging System and software (LI-COR BioSciences) according to the manufacturer's instructions.
Immunofluorescence Microscopy-The protocol for immunostaining has been described previously (56).

MITF Binds ERK2 Specifically and with High
Affinity-To assess the nature of the interaction between human ERK2 and its target transcription factor MITF, rat ERK2 was fused at its N terminus to Schistosoma japonicum glutathione S-transferase (GST), and the resulting fusion protein was expressed in bacteria and purified by adsorption to glutathione-Sepharose beads. ERK2 prepared in this manner is obtained in its unphosphorylated, unactivated state. GST-ERK2 (or GST alone as a negative control) was then incubated with full-length human MITF that had been produced in radiolabeled form by in vitro translation ( Fig. 1). As a positive control, GST-ERK2 beads were incubated with radiolabeled human MEK2; we have previously shown that ERK2-MEK2 binding can be detected using this assay (48). Bead-bound complexes were collected by sedimentation, washed extensively, and analyzed by SDS-PAGE and autoradiography. As shown in Fig. 1, full-length MITF bound efficiently to ERK2 (K d ϳ 6 M, see TABLE ONE); this binding was specific, because precipitation of MITF protein did not occur when GST was used instead of the GST-ERK2 fusion protein. MEK2 also bound specifically to GST-ERK2, as previously reported (48). Additionally, fulllength MITF also bound to a GST-ERK1, although with somewhat reduced affinity compared with the GST-ERK2 (data not shown; see also  (Fig. 2A). To localize the ERK-binding domain of MITF, polypeptide fragments of MITF were produced and tested for binding to GST-ERK2 (Fig. 2B). As shown in Fig. 2C, MITF 1-288 and MITF  bound to GST-ERK2 comparably to full-length MITF. In contrast, MITF 199 -419 and MITF 199 -288 showed very weak binding to ERK2, and MITF 289 -419 did not bind ERK2 at all.
To further delineate the minimal ERK-binding region, additional MITF fragments were produced and tested and are summarized in Fig.  2B and TABLE ONE. MITF 1-180 did not bind to ERK2, indicating that the C-terminal border of the ERK-binding domain lies between residues 180 -198. Analysis of the N-terminally truncated fragments revealed several interesting observations. First, MITF  and MITF 70 -419 bound to ERK2 better than full-length MITF, with K d values of ϳ2 M (TABLE ONE). This suggests that the first 34 residues contain some FIGURE 1. MITF binds ERK2 with relatively high affinity. A, rat ERK2, fused to GST, was tested for binding to 35 S-labeled MITF or MEK2. B, 35 S-labeled proteins were prepared by in vitro translation and partially purified by ammonium sulfate precipitation, and portions (10% of that used in the binding reaction, lane 1) were resolved on a 10% SDS-PAGE gel. Samples (ϳ1 pmol) of the same proteins were incubated with 2.5 g of GST or GST-ERK2 and bound to glutathione Sepharose beads, and the resulting bead-bound 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, and by autoradiography for visualization of the bound radiolabeled protein. C, three independent experiments such as that shown in B were quantified on a PhosphorImager. The axis shows the percentage of the total input radiolabeled protein that was precipitated (i.e. co-sedimented) with the GST fusion. Standard error bars are shown.  (Fig. 3A). Furthermore, an MITF derivative lacking residues 98 -198 displayed markedly reduced binding to ERK2, whereas a derivative lacking the first 198 residues did not bind to ERK2 at all (Fig. 3, B and C).
Typical MAPK docking sites can be constituted within short peptides (31,47,49). To determine if the ERK2-binding portion of MITF could be further localized to a short, contiguous, "peptide-like" region, a series of overlapping 15-mer peptides, covering the region between MITF residues 61 and 205, were synthesized and anchored to a cellulose membrane (62). This array was then incubated with radiolabeled ERK2. In contrast to MEK1 2-16 and MEK2 2-16 D-site peptides included, as controls, on the same array, none of the MITF-derived peptides bound appreciably to ERK2 (data not shown). Hence, there is no short contiguous stretch of amino acids within MITF residues 61-210 that is capable of binding to ERK2 with reasonable affinity and stability. This result, combined with the localization results described above, support the notion that the ERK-binding domain (EBD) of MITF is indeed perhaps best conceptualized as a globular domain contained within residues 98 -198.  (23,61), the basic helix-loop-helix leucine zipper domain that mediates DNA binding (bHLH-LZ), and identified phosphorylation sites (shown as circles) for ERK2, GSK3␤, p38, and RSK1 (21,22,54,80). B and C, 35 S-labeled MITF derivatives were tested for binding to GST (10 g) or GST-ERK2 (2 or 10 g). Other experimental details were as in Fig. 1. In B, 35 S-labeled MITF  and MITF 98 -198 were tested for binding to 40 pmol of GST or GST-ERK2. Other experimental details were as in Fig. 1. B, 35 S-labeled MITF, MITF ⌬98 -198 , and MITF ⌬1-198 (also known as MITF 199 -419 ) were tested for binding to 40 pmol of GST, GST-ERK1, or GST-ERK2. Other experimental details were as in Fig. 1. C, three independent experiments such as that shown in B were quantified on a PhosphorImager. Standard error bars are shown. D, the location of the minimal ERK-binding domain (residues 98 -198) is shown on a schematic representation of MITF. Our data also indicate that the region around the ERK2 phosphorylation site at Ser 73 also contributes to the affinity of the ERK-MITF interaction. Also shown is a region (residues 1-35) that appears to partially inhibit ERK-MITF binding.
MITF and ERK2 Co-immunoprecipitate from Human Cells-To assess the ability of MITF and ERK2 to associate in cells, human kidney epithelial cells (HEK293) were transfected with plasmid constructs expressing ERK2 from a strong promoter, and co-transfected with a plasmid expressing full-length MITF tagged at its C terminus with the V5 epitope. Cells were lysed 48 h post-transfection, and MITF was immunoprecipitated using an anti-V5 antibody. As is shown in Fig. 4, ERK2 efficiently co-precipitated with full-length MITF under these circumstances (Fig. 4B, lane 2). However, ERK2 did not co-immunoprecipitate with MITF-⌬EBD, an otherwise identical MITF derivative lacking the ERK-binding domain (lane 4). As expected, ERK2 was not immunoprecipitated by the anti-V5 antibody alone (lane 6).
The lack of interaction between ERK2 and MITF-⌬EBD was not due to mislocalization of MITF-⌬EBD, because both wild-type MITF and MITF-⌬EBD localized to the nucleus (Fig. 4C). This is consistent with the proposal that MITF residues 210 -222 contain its primary nuclear localization signal (63).
To determine if the ERK-binding domain of MITF was sufficient to associate with ERK2 in cells, a plasmid expressing V5-tagged MITF 98 -198 was co-transfected into HEK293 cells with the vector overexpressing ERK2, cells extracts were prepared, and MITF 98 -198 was immunoprecipitated. As shown in Fig. 4D, ERK2 specifically co-precipitated with MITF 98 -198 . In summary, as determined by co-immunoprecipitation, ERK2 and MITF stably associated in human cells, and the ERK-binding domain of MITF was necessary and sufficient for this interaction.
GST-MITF Fusion Proteins-To further assess the ERK-MITF interaction, we attempted to express several GST-MITF fusion proteins. We were unable to obtain soluble GST-MITF derivatives containing MITF residues 1-34; however, fusion proteins lacking these residues were soluble (data not shown). It is interesting to note that residues 1-34 of MITF also inhibited binding to ERK2 (see above), although this inhibition must have been due to a different mechanism (i.e. not because of reduced solubility), because all the in vitro-translated proteins were comparably soluble. Soluble fusion proteins that could be purified in high yield were obtained by fusing residues 35-198 of MITF to GST, yielding GST-MITF  ; this protein contains both the MAPK-phosphorylation site and the ERK-binding domain of MITF. As a control, GST-MITF  , which contains the MAPK phosphorylation site but lacks the ERK-binding domain, was also created. These proteins were then purified from bacteria and used in several assays, described below.
N Terminus of MITF Selectively Binds to ERK2-To explore the specificity of the MITF-MAPK interaction, three proteins that were not expected to bind to MITF were produced by in vitro translation and tested for binding to GST-MITF  . These proteins were the human MAPK/ERK kinases MEK1 and MKK6 and the bacterial enzyme ␤-galactosidase. Also included in this analysis were ERK2 and the human p38␣ MAPK. As is shown in Fig. 5, none of the in vitro-translated proteins bound to either GST or GST-MITF  , and neither MEK1, MKK6, nor ␤-galactosidase bound appreciably to GST-MITF  , as expected. Furthermore, when compared with ERK2, p38 bound to GST-MITF  relatively poorly (Fig. 5). Thus, the N terminus of MITF binds to ERK2 (and ERK1) with specificity.
Fus3 and Kss1, two yeast MAPK proteins, also bound with relatively high affinity to GST-MITF  but not to GST-MITF  or GST (data not shown). These two proteins are presumed orthologs of ERK1/2 (64) and have been shown to have binding specificities similar to ERK1/2 (31,48).
Both Phosphorylated and Unphosphorylated ERK Bind to MITF-To examine the binding of phosphorylated ERK1 and ERK2 to MITF, the following experiment was performed (Fig. 6A): human HEK293 cells were serum-starved for 16 -20 h to shut down flux though the Ras/ MAPK pathway. Half of the cells were then stimulated for 5 min with epidermal growth factor to activate ERK1 and ERK2. Cell lysates were prepared, co-sedimented with GST-MITF fusion proteins, and analyzed by immunoblotting with an anti-total-ERK antibody or anti-phospho-ERK antibodies. A fusion of GST to the N terminus of human MEK2 was also included in this experiment as a positive control; this protein has been previously shown to bind with high affinity to ERK2 (47). As shown in Fig. 6B (top left panel), unphosphorylated ERK2 bound to GST-MEK2 1-64 and GST-MITF  , but not to GST or GST-MITF  ; unphosphorylated ERK1 was difficult to visualize in

. ERK2 and MITF co-immunoprecipitate from HEK293 cells, and the ERKbinding domain of MITF is necessary and sufficient for this interaction.
A, V5-epitope-tagged MITF derivatives were expressed in HEK293 cells and immunoprecipitated; co-precipitation of ERK2 was assessed. B, HEK293 cells were co-transfected with plasmid DNA encoding either V5-tagged full-length MITF, MITF-⌬EBD (a derivative lacking residues 98 -198), or empty vector, and plasmid encoding ERK2. Cells were lysed 48 h post-transfection, and MITF was immunoprecipitated using an anti-V5 antibody. Immunoprecipitated complexes (IP) along with 10% of the input lysate (10%) were analyzed by Western blotting using antibodies against the V5 tag and ERK2. Both MITF and MITF-⌬EBD run as doublets. The asterisk indicates a cross-reacting band present in the lysate. C, HeLa cells were transfected with plasmid DNA encoding either V5-tagged fulllength MITF, MITF-⌬EBD, or empty vector. Cells were analyzed by immunofluorescence microscopy using an anti-V5 antibody. Cells were also counterstained with 4Ј,6-diamidino-2-phenylindole to visualize nuclei. Scale bar ϭ 20 m. D, HEK293 cells were cotransfected with plasmid DNA encoding V5-tagged MITF 98 -198 or empty vector, and plasmid DNA encoding ERK2. MITF 98 -198 was immunoprecipitated and complexes were analyzed as above.
this experiment. Furthermore (lower right panel), both active ERK1 and active ERK2 also bound to GST-MEK2 1-64 and GST-MITF  , but not to GST or GST-MITF  . Hence the ERK-binding domain of MITF can bind to both phosphorylated and unphosphorylated ERK1 and ERK2 in cell extracts.
Mutations in the CD/7m Domain of ERK2 Have No Effect on MITF Binding-Examination of residues 70 -198 of MITF did not reveal the presence of any known MAPK docking site motifs. In particular, there was neither a clear consensus D-site ((K/R) 1-3 -X 1-5 -X), nor a consensus DEF-site (FXFP). Using the program Scansite (65), one weak match to the D-site consensus 182 KRELTACIFP 191 , was found. Alteration of this putative D-site by site-directed mutagenesis, however, did not diminish MITF-ERK binding (data not shown), indicating that this region is not a functional D-site mediating the ERK-MITF interaction. Scansite also identified a weak match to the DEF-site in the same region, 188 CIFP 191 . Interestingly, residues 187-192 (ACIFPT) are encoded by an exon that is spliced out of one of the two common isoforms of MITF (66,67). Deletion of residues 187-192 (mimicking the alternative splice, and both removing the putative DEF-site and truncating the putative D-site) did not affect the binding of MITF  to GST-ERK2 (data not shown). Thus, neither of the putative MAPK docking sites in the ERKbinding domain of MITF are required for MITF-ERK binding.
D-site-mediated interactions (as well as those mediated by the related MAPK-docking site found in MAPKAPKs) require the integrity of an acidic patch on the surface of the MAPKs designated the "conserved docking" (CD) region (32). The CD region is also called the sevenmaker region, after the allele of the Drosophilla MAPK rolled that first indicated this region's importance in MAPK signaling (68). The fact that the ERK-binding domain of MITF did not contain a consensus D-site, nor more than two acidic residues in a row, suggested that it did not bind to the CD/sevenmaker region. To directly test this suggestion, the ability of ERK2 derivatives with mutations in the CD region to bind to MITF was assessed. Two ERK2 mutants were tested. The ERK2-7m1 mutant (D319N) is equivalent to the sevenmaker mutation in rolled (D334N). The ERK2-7m2 double mutant (D316A,D319A) is equivalent to mutations in yeast Kss1 and Fus3 that caused a severe defect in binding to Ste7 MEK , as well as to the D-site-containing substrates Dig1 and Dig2 (31). In vitro-translated and radiolabeled ERK2, ERK2-7m1, and ERK2-7m2 were incubated with GST-MITF  , GST-MITF 35-98 , or GST. In addition, GST fusions to two MEK proteins, yeast Ste7 (GST-Ste7 1-172 ) and human MEK2 (GST-MEK2 1-64 ), were also included. As shown in Fig. 7, wild-type ERK2 bound to GST-Ste7 and GST-MEK2, as previously reported (47,48). In contrast, the CD/sevenmaker region mutants of ERK2 were defective in binding to GST-Ste7 and GST-MEK2; this was expected, based upon previously reported findings with similar MAPK mutants (31, 32, 69). Notably, both the   . B, 35 S-labeled ERK2, p38, MEK1, MKK6, or ␤-galactosidase were prepared by in vitro translation and partially purified by ammonium sulfate precipitation, and portions (5% of that used in the binding reaction) were resolved on a 10% SDS-polyacrylamide gel (lane 1). Samples (ϳ1 pmol) of the same proteins were incubated with ϳ200 g of GST, GST-MITF  , or GST-MITF 35-98 bound to glutathione-Sepharose beads, and the resulting bead bound 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 (a representative sample shown in the lowest panel), and by autoradiography for visualization of the bound radiolabeled protein (upper five panels). C, results of the above experiment, quantified by PhosphorImager, normalized to the binding of ERK2 to GST-MITF 35-198. Error bars show standard deviations (n ϭ 2). . Phosphorylated and unphosphorylated ERK2 from HEK293 cell lysates bind to MITF. A, lysates prepared from a human embryonal kidney cell line were tested for binding to GST-MITF and controls. B, lysates from non-stimulated or epidermal growth factor-stimulated HEK293 cells were incubated with 1 nmol (30 -50 g) of beadbound GST, GST-MITF  , or GST-MITF  , or GST-MEK2 1-64 . Co-sedimentation assays were performed as described under "Experimental Procedures." The amounts of co-sedimented total and phosphorylated ERK in each sample were assessed by Western blotting with antibodies against total ERK and dually phosphorylated ERK (phospho ERK). Also shown are lysate samples corresponding to 5% of the total protein input (lanes 1 and 6). Equal sedimentation of the GST fusions was verified by staining with Coomassie Blue (not shown).
CD/sevenmaker-region mutants of ERK2 bound to MITF  just as well as did wild-type ERK2. Finally, none of the ERK2 derivatives bound to MITF  or GST. Thus, the ERK2 CD/sevenmaker mutations abolished binding to D-site containing MEKs, but did not affect binding to MITF. From this it can be concluded that the integrity of the CD/sevenmaker region is not required for the ERK2-MITF interaction.
The MITF-ERK2 Interaction Potentiates ERK2 Phosphorylation of MITF-The docking of MAPKs to substrates containing D-sites or FXFP sites has been shown to facilitate the MAPK-dependent phosphorylation of those substrates (29,37,70). To ascertain whether the ERKbinding domain of MITF potentiated the phosphorylation of MITF by ERK2, in vitro kinase assays were performed using activated ERK2 as the enzyme, and GST-MITF  or GST-MITF 35-98 as the substrate. As shown in Fig. 8, GST-MITF  , which lacks the ERK-binding domain, was phosphorylated 2-to 4-fold less efficiently than GST-MITF  , which contains the ERK-binding domain. MITF  S73A, in which the phosphoacceptor serine 73 was changed to an alanine, was not phosphorylated at all, as expected (Fig. 8, B and C).
D-site Peptides Differentially Inhibit ERK2 Phosphorylation of Elk-1 versus MITF-D-site-based peptides have been shown to inhibit the phosphorylation of D-site-containing substrate proteins by MAPKs (47,49,71). Most of this inhibition is presumably caused by the peptide binding to the MAPK and competitively inhibiting the docking of the MAPK to its substrate (49). The results described above indicated that the ERK-binding domain MITF does not contain a D-site, and that the MITF-ERK2 interaction does not involve the MAPK CD/sevenmaker region, one of the patches on the MAPK surface that D-sites bind to. Thus we hypothesized that the ERK2-mediated phosphorylation of MITF might not be strongly inhibited by D-sites peptides (Fig. 9A). To test this idea, in vitro kinase assays were performed using activated FIGURE 7. The MITF-ERK2 interaction does not require the CD/sevenmaker region of ERK2. A, wild-type ERK2, or CD/sevenmaker mutants thereof (7m), were tested for binding to GST-MITF and controls. B, 35 S-labeled ERK2, ERK2-7m1, or ERK2-7m2 (ϳ1 pmol) were tested for binding to GST, GST-MITF  , GST-MITF  , GST-Ste7 1-172 , or GST-MEK2 1-64 (lanes 2-6, respectively). The amount of each GST fusion protein was 50 g, except for GST-Ste7 1-172 , for which only 1 g was used, due to its high binding affinity for MAPKs (48). Other experimental details were as in Fig. 5. C, results of the above experiment, quantified by PhosphorImager, normalized to the binding of wild-type ERK2 to the indicated GST fusion. Error bars show standard deviations (n ϭ 2). ERK2 as the enzyme and GST-Elk1 307-428 , GST-MITF  , or GST-MITF 35-98 as the substrate. To certain reactions, the Ste7 D-site peptide was added to a final concentration of 5-50 M. This peptide (sequence FQRKTLQRRNLKGLNLNLHPD) was chosen because it displayed the most potent inhibitory activity of several that were tested previously (48,49). A peptide consisting of a scrambled version of this sequence (scram7, sequence RLQPNLQDKLNHFRTNLKGLR) was added to certain reactions as a control. The results are shown in Fig. 9 (B and C).
The Elk-1 transcription factor contains one of the first characterized D-sites for ERK2 (29). Accordingly, ERK2-mediated phosphorylation of GST-Elk was potently inhibited by the native Ste7 D-site (IC 50 ϳ 3 M), but not by the scrambled peptide, as we have previously reported (49). In contrast, the phosphorylation of GST-MITF  , which does not contain a D-site, was inhibited much less efficiently by the Ste7 peptide site (IC 50 ϳ 50 M). Furthermore, this weak inhibition was also seen with MITF  , which lacks the ERK-binding domain (EBD); thus, it was apparently not caused by the peptide blocking the productive interac-tion of the MITF EBD with ERK2, and may have instead been caused by a peptide-induced conformational change in ERK2 (30,34).
The selectivity of the Ste7 D-site for inhibiting ERK2-mediated phosphorylation of Elk-1 versus MITF was estimated by dividing the IC 50 for its inhibition of ERK2 phosphorylation of MITF (50 M) by the IC 50 for its inhibition of ERK2 phosphorylation of Elk-1 (3 M). This calculation indicated a selectivity of Ͼ15-fold. Thus, peptides that block D-sitemediated docking interactions can be used to selectively inhibit the phosphorylation of a D-site containing substrate (Elk-1), while not substantially impairing the phosphorylation of a different substrate that does not contain a D-site (MITF).

DISCUSSION
MAPKs, and many other kinases, bind to their substrates and regulators using high affinity docking interactions as part of a bipartite recognition strategy (docking site and phosphorylation site) that enhances signal transmission and may enable a "double selection" for fidelity (24,25,72). In this study, we have demonstrated a stable, high affinity interaction between the MAPKs ERK1 and ERK2 and their transcription factor target MITF. Both phosphorylated and unphosphorylated ERK1 and ERK2 were found to bind to MITF with an affinity in the low micromolar range, comparable to the affinities seen for other MAPK-interacting proteins (31,47,73) (Figs. 1 and 6). The high affinity ERK-binding region of MITF was localized to a domain of ϳ100 amino acids lying C-terminal to the ERK2 phosphorylation site at Ser 73 (Figs. 2 and 3). This ERK-binding domain was necessary and sufficient for the stable association of ERK2 and MITF in human cells (Fig. 4). The ERK-binding domain exhibited specificity for MAPK binding, because it did not bind to irrelevant proteins such as MEK1, MKK6, and ␤-galactosidase (Fig.  5). The presence of the ERK-binding domain increased the efficiency of Ser 73 phosphorylation (Fig. 8). Thus we conclude that, as it does with other many substrates, ERK2 uses a bipartite recognition strategy to facilitate the specific and efficient phosphorylation of MITF.
The ERK-binding domain is devoid of known MAPK-docking sites such at the D-site or the FXFP site. Consistent with the lack of a D-site or a related docking site containing a basic submotif, we found that ERK-MITF binding did not require the integrity of the MAPK CD/sevenmaker region, a patch of acidic residues required for MAPK binding to multiple substrates and regulators (31,32) (Fig. 7). These results suggested that MITF might bind to a part of the MAPK surface different from those regions that are utilized for binding to D-sites (30,(32)(33)(34)(35). Consistent with this idea, we showed that a D-site-based peptide that potently inhibited the ERK2-mediated phosphorylation of a D-site-containing substrate (Elk-1) only weakly inhibited the phosphorylation of MITF (Fig. 9). These results provide a proof-of-principal for the possibility of small-molecule MAPK inhibitors that would block the phosphorylation of some substrates while allowing phosphorylation of others (Fig. 9A).
Diversity of MAPK-binding Motifs-There are other examples of proteins that bind to MAPKs using domains that do not contain D-sites or FXFP sites. These include yeast Ste12, which binds preferentially to unphosphorylated Kss1 MAPK (74); mammalian PEA-15, which binds to ERK1/2 via its DED domain (75); and Ets-1, which binds via its pointed domain (76). The ERK-binding domain of MITF is neither a pointed nor a DED domain and does not have any recognizable homology to Ste12; thus, it would appear to constitute a novel MAPK-binding domain. Of course, it is possible that MITF and these other proteins share overlapping binding surfaces on the MAPK. Interestingly, the ERK-binding domain contains within it the strongest of the two transcriptional activation domains of MITF (23). FIGURE 9. Differential inhibition of Elk-1 versus MITF phosphorylation by a D-site peptide. A, the ability of a D-site peptide to inhibit ERK2 phosphorylation of Elk-1 was compared with its ability to inhibit ERK2 phosphorylation of MITF. Purified GST-MITF , GST-MITF  , or GST-Elk-1 307-428 (1 M each) was incubated with 10 units (ϳ1 nM) purified active ERK2 and [␥-32 P]ATP for 20 min in the absence or presence of the specified concentration of the Ste7 D-site peptide or Scram7, a scrambled control peptide. B, results are plotted as percent phosphorylation relative to that observed in the absence of peptide. MITF or Elk-1 phosphorylation was analyzed by SDS-PAGE and quantified on a PhosphorImager. Data are the average of two to four experiments, with triplicate data points in each experiment. C, an autoradiogram of a representative experiment. The fastest migrating Elk-1 band, which represents an Elk-1 phosphorylation site that appears to be unaffected by peptide inhibition (49), was omitted from the quantification.
Progress toward More Specific Kinase Inhibitors-There is increasing interest the possibility of developing drug candidates that modulate kinase function in a selective mechanism-based manner, rather than by competing for ATP binding (43,45,77). Clearly it might be useful to develop "substrate-selective kinase inhibitors": agents that could inhibit certain phosphorylation events while leaving others unperturbed. Some progress along these lines has been made recently. A peptide inhibitor of cyclin-dependent kinase-mediated Rb phosphorylation has been developed that does not affect cyclin-dependent kinase-mediated histone phosphorylation (78). CMPD1 is a small molecule inhibitor of p38␣ MAPK that exhibits substrate-selective inhibition of MK2 versus ATF-2 (79). In addition, the finding that the FXFP-site binds to a different part of the MAPK surface than the D-site (35) suggests that it should be possible to specifically inhibit FXFP interactions.
Here, we used a model peptide based on the D-site class of MAPK docking sites and showed that this peptide substantially inhibited ERK2mediated phosphorylation of Elk-1, but did not dramatically affect the phosphorylation of MITF. In fact, the selectivity of the D-site peptide for inhibiting Elk-1 versus MITF phosphorylation was over 15-fold. The simplest explanation of these results is that the peptide bound to ERK2 and blocked the ERK2-Elk-1 docking interaction by competing with the D-site on Elk-1 but did not block the ERK2-MITF interaction. Thus, the D-site peptide acted as a substrate-selective ERK2 inhibitor.
A potential caveat to this interpretation lies in the fact that the D-site peptide did inhibit MITF phosphorylation somewhat, with an IC 50 of ϳ50 M. It seems unlikely that this effect was due to the peptide competing with the ERK-binding domain of MITF, because the peptide also inhibited the phosphorylation of an MITF fragment that lacked the ERK-binding domain, and did so with an IC 50 of ϳ50 M. One possible explanation for this inhibition is that the binding of the D-site peptide may have caused a conformational change in ERK2 that reduced its catalytic activity. Recent structural studies of MAPK-D-site complexes are consistent with this possibility (30,34).