The Docking Interaction of Caspase-9 with ERK2 Provides a Mechanism for the Selective Inhibitory Phosphorylation of Caspase-9 at Threonine 125*

Caspase-9 plays a critical role in the initiation of apoptosis by the mitochondrial pathway. Activation of caspase-9 is inhibited by phosphorylation at Thr125 by ERK1/2 MAPKs in response to growth factors. Here, we show that phosphorylation of this site is specific for these classical MAPKs and is not strongly induced when JNK and p38α/β MAPKs are activated by anisomycin. By deletion and mutagenic analysis, we identify domains in caspase-9 and ERK2 that mediate their interaction. Binding of ERK2 to caspase-9 and subsequent phosphorylation of caspase-9 requires a basic docking domain (D domain) in the N-terminal prodomain of the caspase. Mutational analysis of ERK2 reveals a 157TTCD160 motif required for recognition of caspase-9 that acts independently of the putative common docking domain. Molecular modeling supports the conclusion that Arg10 in the D domain of caspase-9 interacts with Asp160 in the TTCD motif of ERK2. Differences in the TTCD motif in other MAPK family members could account for the selective recognition of caspase-9 by ERK1/2. This selectivity may be important for the antiapoptotic role of classical MAPKs in contrast to the proapoptotic roles of stress-activated MAPKs.

Apoptosis is an evolutionarily conserved form of cell death that plays a fundamental role in development and tissue homeostasis in multicellular organisms (1). Caspases, a family of cysteine aspartyl proteases, play a central role in the implementation of apoptosis (2). Caspases are synthesized as inactive zymogens (procaspases) that are activated by oligomerization and/or proteolytic processing to form active enzymes in response to apoptotic stimuli (2). In the intrinsic or mitochondrial pathway of apoptosis induced by a wide variety of stress or damage signals, cytochrome c is released from the mitochondria into the cytosol, where it binds to Apaf-1 (apoptotic protease-activating factor 1) and induces its oligomerization (3,4).
This leads to the recruitment of procaspase-9, promoting its dimerization, activation, and autocatalytic processing (5,6). Caspase-9 then initiates a proteolytic cascade, cleaving and activating downstream effector caspases, such as caspase-3 and caspase-7. These effector caspases further promote cytochrome c release in a positive feedback mechanism and directly catalyze the cleavage of key structural and regulatory proteins that results in the morphological and biochemical changes associated with apoptotic cell death (7).
Signal transduction pathways activated by extracellular and intracellular stimuli can impinge on the intrinsic apoptotic pathway and control cell fate. Abnormal or constitutive activation of cell signaling pathways may lead to tumorigenesis through inappropriate suppression of cell death (8,9). One important control point of apoptosis targeted by cell signaling pathways downstream of cytochrome c release is at the level of the apoptosome (10). Several signal transduction pathways modulate apoptosome function through phosphorylation of caspase-9 or Apaf-1 (11)(12)(13)(14). Notably, we have shown previously that caspase-9 is directly inhibited through phosphorylation by ERK1/2 mitogen-activated protein kinases (MAPKs) 4 at Thr 125 in response to epidermal growth factor (EGF) or a phorbol ester, TPA (12). ERK1/2 are classical MAPKs that are activated mainly by growth factors or mitogenic stimuli. They form part of the larger MAPK family that also includes the c-Jun N-terminal kinases (JNKs) and p38 kinases, which are activated by a variety of cellular stresses as well as extracellular stimuli (15). Individual MAPKs have different biological functions that are determined not only by their selective activation in response to different stimuli but also by their distinct substrates (16). All MAPKs preferentially catalyze the phosphorylation of substrates containing the minimal consensus sequence Ser/Thr-Pro, which is recognized by the active site of the kinase (17). Additional interactions are mediated by docking motifs in substrates, often located distant from the site of phosphorylation, which provide high affinity and selective recognition by particular MAPKs (18,19).
In general, activation of ERK MAPKs is associated with suppression of apoptosis, whereas JNK and p38 MAPKs are involved in the induction of apoptosis (20). Although these roles may depend on cell type and stimulus and are likely to be integrated with the control of cell growth and the cell division cycle, which may even have antagonistic effects on cell survival, this suggests that distinct MAPKs are likely to have different effects on proteins involved in the control of apoptosis, such as caspase-9. Here, we have investigated the recognition of caspase-9 by MAPKs. We show that Thr 125 in caspase-9 is selectively phosphorylated by ERK1/2 but not JNK or p38␣/␤. We identify a basic docking motif (D domain) in the prodomain of caspase-9 that regulates the binding and phosphorylation of caspase-9 by ERK1/2. We also demonstrate a requirement for a TTCD motif in ERK2, which is not conserved in p38 or JNK, for mediating efficient phosphorylation of caspase-9. Molecular modeling indicates that there is a direct interaction between the D domain of caspase-9 and the TTCD motif in ERK2. These findings uncover the molecular basis for the selective phosphorylation and regulation of caspase-9 by ERK1/2, providing a mechanism by which ERK1/2 may selectively suppress apoptosis in contrast to JNK and p38␣/␤ MAPKs.

MATERIALS AND METHODS
Plasmid Constructs and Site-directed Mutagenesis-Human caspase-9 was amplified from U2OS cells and cloned into pcDNA3 (Invitrogen). To replace serine residues in caspase-9 with alanine residues, we used the QuikChange TM site-directed mutagenesis kit (Stratagene, Cedar Creek, TX) as indicated by the manufacturers. The mutations were verified by sequence analysis. For expression of recombinant proteins, cDNAs were subcloned into pGEX-4T-1 (Amersham Biosciences). Expression of recombinant proteins was induced in Escherichia coli BLR (DE3) at 30°C for 2 h by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside. Glutathione S-transferase (GST)-tagged proteins were affinity-purified with glutathione-Sepharose 4B (Amersham Biosciences) before elution in buffer (10 mM Hepes-KOH at pH 7.5, 150 mM NaCl, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, and 1 g ml Ϫ1 each of aprotinin, leupeptin, and pepstatin A) containing 100 mM glutathione. Glutathione was removed by filtration through a PD-10 column (Amersham Biosciences). Constructs for HAtagged rat ERK (HA-ERK2 wild type, HA-D319N-ERK2) and GST-RSK-D2 were provided by Dr. J. Blenis (Harvard Medical School, Boston, MA). Additional HA-ERK2 constructs were generated using the QuikChange TM site-directed mutagenesis kit (Stratagene, Cedar Creek, TX).
Phosphorylation of Recombinant GST-Caspase-9-0.5 g of GST-caspase-9, inactivated by mutation of the catalytic cysteine (C287A), including proteins with additional residues changed in the N-terminal region, was incubated for 15 min at 30°C with active His-ERK2 in kinase buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , and 1 mM dithiothreitol) containing 100 M [␥-32 P]ATP (specific activity, 6 ϫ 10 6 cpm nmol Ϫ1 ). The reaction was terminated by the addition of reducing SDS-PAGE loading buffer. Samples were subjected to SDS-PAGE followed by autoradiography.
Peptide Competition Assay-Recombinant GST-caspase-9 (0.5 g) was incubated with purified active His-ERK2 and [␥-32 P]ATP for 15 min at 30°C in 15 l of kinase buffer with the indicated concentrations of wild type peptide (ADRRLLR-RCRLRLV), the R10A peptide (ADRRLLARCRLRLV), the R10A/R11A peptide (ADRRLLAACRLRLV), and an unrelated control peptide R-DIS (CPKSKKVKVSHRSHST). Caspase-9 phosphorylation was analyzed by SDS-PAGE and autoradiography. Gels were analyzed by staining with Coomassie Blue for visualization of caspase-9 to verify equal amounts in the reaction. For quantification of phosphorylation, protein bands were excised and analyzed by scintillation counting. Experiments were conducted three times, and a representative data set is shown.
GST Pull-down Assay-0.5 g of GST-caspase-9 C287A, including proteins with additional residues changed in the N-terminal region, was incubated in HeLa cytosolic S100 extract with 1 M okadaic acid for 60 min. GST proteins were recovered on glutathione-Sepharose beads (Amersham Biosciences) by incubating for 1 h at 4°C with rotation. Beads were pelleted by centrifugation and washed three times in buffer (20 mM Tris-HCl, pH 7.5, 150 mM, NaCl, 10% glycerol, 1% Triton X-100, 2 mM EDTA, 1 mM Na 3 VO 4 , 50 mM NaF, and 5 mM ␤-glycerophosphate) before boiling in reducing SDS-PAGE sample buffer. Proteins were separated by SDS-PAGE and immunoblotted with antibodies to ERK1/2, phospho-Thr 125 caspase-9 or GST as indicated.
Inducible Expression of Caspase-9 in Cells-cDNAs of human caspase-9 (wild type or alanine-mutated at arginine 10) were subcloned into the pIND expression plasmid before cotransfection of recombinant pIND and pVgRXR into U2OS cells, in which we have previously characterized the regulation of caspase-9 by ERK1/2 (12). Stably transfected cells were selected and maintained in DMEM supplemented with 10% fetal bovine serum and 100 g ml Ϫ1 penicillin/streptomycin together with 500 g ml Ϫ1 zeocin and 800 ng ml Ϫ1 G418 sulfate. Transcriptional expression of exogenous caspase-9 was induced with the indicated concentrations of ponasterone A (AXXORA, Nottingham, UK).
ERK2 Kinase Assay-Transfections of HEK293 cells were performed using 10 g of CsCl-purified HA-ERK2 plasmid DNA and 60 l of Superfect according to the manufacturer's instructions (Qiagen). After 3 h, culture medium was replaced with medium containing 10% serum. 20 h after transfection, cells were serum-starved overnight before being stimulated with 50 ng ml Ϫ1 EGF for 5 min. Cells were lysed in lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 2 mM EDTA, 1 mM Na 3 VO 4 , 50 mM NaF, 5 mM ␤-glycerophosphate, 1 g ml Ϫ1 leupeptin, 1 g ml Ϫ1 aprotinin, 1 g ml Ϫ1 pepstatin A, 1 mM phenylmethylsulfonyl fluoride, 0.1% ␤-mercaptoethanol, and 1 M okadaic acid) and HA-ERK2 immunoprecipitated with 40 l of anti-HA agarose beads (Sigma) for 1 h at 4°C with rotation. Beads were then pelleted by centrifugation and washed three times in lysis buffer. The washed beads were mixed with substrates in a kinase buffer (50 mM Tris-HCl at pH 7.5, 10 mM MgCl 2 , and 1 mM dithiothreitol) containing 100 M [␥-32 P]ATP (specific activity 6 ϫ 10 6 cpm nmol Ϫ1 ) and incubated for 15 min at 30°C. The reaction was terminated by the addition of reducing SDS-PAGE loading buffer. Samples were then subjected to SDS-PAGE followed by autoradiography. For quantification of phosphorylation, protein bands were excised and analyzed by scintillation counting. To ensure that differences in phosphorylation of caspase-9 and RSK by mutagenesis of ERK2 were not due to differences in kinase activity caused by mutation, quantification of phosphorylation was calculated taking into account any differences in activity toward myelin basic protein (MBP).
Molecular Modeling-Two Protein Data Bank structures, the caspase-9 prodomain (21,22) and phosphorylated and activated rat ERK2 (23), were run through ClusPRO (24), a fully automatic docking software program that first docked the proteins based only on the shape complementarity (no electrostatics) and retained the top 20,000 structures. The complexes generated from the docking step were then rapidly screened, and the desolvation and electrostatic energies were calculated. The top 1500 electrostatic conformations and the top 500 desolvation conformations were kept for further analysis. For each of the 2000 conformations retained in the filtering stage, the best 10 were determined using a clustering algorithm that scores them based on free energy landscape calculations. PyMOL software (DeLano Scientific) was used to render models.

RESULTS
Caspase-9 Is Phosphorylated at Thr 125 in Cells by ERK1/2 but Not by JNK or p38 MAPKs-We have previously established that ERK1/2 directly phosphorylate and inhibit caspase-9 at a single major site, Thr 125 (12). We were interested to determine if other members of the MAPK family might also phosphorylate caspase-9 in cells. We transfected HEK293 cells with caspase-9 and stimulated them for 60 min with anisomycin, a protein synthesis inhibitor that activates the JNK pathway (25) and all isoforms of p38 MAPK (26). Although anisomycin stimulated the phosphorylation of both JNK at sites associated with its activation and its downstream target c-Jun, showing that the JNK pathway was indeed activated, no increase in phosphorylation of caspase-9 at Thr 125 was detected (Fig. 1A). By contrast, stimulation by the phorbol ester TPA resulted in ERK1/2 activation and subsequent phosphorylation of caspase-9 at Thr 125 , as expected. TPA-stimulated phosphorylation of ERK1/2 and caspase-9 were blocked by co-incubation with the MEK1/2 FIGURE 1. Caspase-9 is phosphorylated in cells at Thr 125 by ERK1/2 but not by JNK or p38 MAPKs. A, HEK 293 cells were transfected with caspase-9 (C287A). After 3 h, culture medium was replaced with medium containing 10% serum. 20 h after transfection, cells were treated with 2 M PD184352 for 5 min prior to the addition of 1 M TPA for 15 min or 10 g/ml anisomycin for 60 min. Cell lysates were immunoblotted with antibodies against phospho-Thr 125 caspase-9, caspase-9, phospho-JNK, JNK, phospho-c-Jun, phospho-ERK1/2, or ERK1/2. B, HEK 293 cells were transfected with caspase-9 (C287A). After 3 h, culture medium was replaced with medium containing 10% serum. 20 h after transfection, cells were treated with 10 M SB 203580 or 2 M PD184352 for 5 min prior to the addition of 1 M TPA for 15 min or 10 g/ml anisomycin for 60 min. Cell lysates were immunoblotted with antibodies against phospho-Thr 125 caspase-9, caspase-9, phospho-p38, p38, phospho-MAPKAPK-2 (Thr 222 ), phospho-ERK1/2, or ERK1/2. inhibitor PD184352, which had no effect on the JNK pathway and specifically blocks the activation of ERK1/2 at this concentration (27).
Anisomycin also activated the p38/SAP2K pathway in transfected HEK293 cells, resulting in phosphorylation of both p38 and the downstream substrate MAPKAP kinase-2. As expected, phosphorylation of MAPKAP kinase-2 was completely inhibited by SB203580, an inhibitor of p38␣ and p38␤. SB203580 also partially inhibited phosphorylation of p38 itself, as previously described for the ␤-isoform (28,29). However, no significant phosphorylation of caspase-9 at Thr 125 occurred when p38 was stimulated (Fig. 1B). TPA-stimulated phosphorylation of caspase-9 at Thr 125 was completely inhibited by PD184352 but not by SB203580. Together these results show that caspase-9 is selectively phosphorylated in cells by ERK1/2 but not by JNK or p38␣/␤ MAPKs.
A Putative ERK Docking Site in the N-terminal Domain of Caspase-9-We wished to elucidate the mechanism by which ERK1/2 specifically recognize and phosphorylate caspase-9. A clue was uncovered with the observation that phosphorylation of caspase-9 by recombinant ERK2 was substantially reduced upon removal of the first 79 amino acids of the prodomain region of caspase-9 ( Fig. 2A). We therefore examined the sequence of caspase-9 in this region for potential ERK1/2 docking motifs that might promote recognition of the substrate. We identified amino acids 6 -16 of caspase-9 as a putative MAPK docking motif known as a D domain or kinase interaction motif (KIM), which is characterized as a cluster of basic residues often situated N-terminal to a hydrophobic motif containing Leu, Ile, or Val separated by one amino acid, denoted A -X-B (where is a small hydrophobic amino acid) (30,31). Related sequences are found in a number of classical MAPK-interacting proteins, including substrates, regulators, and scaffolding proteins (Fig. 2B). This motif lies within the prodomain of caspase-9, which also contains a CARD region that is involved in interactions with Apaf-1 (Fig. 2C).
Requirement of the D Domain of Caspase-9 for Binding and Phosphorylation by Endogenous ERK1/2-To determine if this motif forms part of an ERK1/2 docking site (D domain), we mutated to alanine three residues in the basic cluster (arginines 7, 10, and 11) or two hydrophobic residues (leucines 14 and 16) that might form the A -X-B motif and tested the effect on phosphorylation by purified ERK2 of recombinant caspase-9 expressed as fusions with GST (Fig. 3A). In an in vitro kinase assay under conditions where phosphorylation was linear with respect to time, wild type caspase-9 was phosphorylated by ERK2, and this phosphorylation was dependent on Thr 125 , the site of phosphorylation. Phosphorylation of the triple arginine mutant (R7A/R10A/R11A) at Thr 125 was strongly reduced A, phosphorylation of GST-caspase-9 (WT) or N-terminally truncated GST-caspase-9 lacking the first 79 amino acids (⌬79) by active ERK2 and [␥-32 P]ATP. Caspase-9 phosphorylation was analyzed by SDS-PAGE and autoradiography (top). B, sequence alignment of known D domains in selected MAPK substrates aligned with the putative docking site in caspase-9 (residues 6 -16). Basic residues are highlighted in boldface type, and the hydrophobic A -X-B motif is underlined. C, schematic of caspase-9 with the putative docking domain located in the N-terminal prodomain, which also contains a CARD region (underlined, boldface type). The sites of proteolytic processing are indicated by small arrows. (WT) or GST-caspase-9 with mutation of basic residues (arginines 7, 10, and 11; AAA) or a hydrophobic motif (leucines 14 and 16; 14/16A) in the N-terminal region to alanine. Recombinant GST-caspase-9 proteins were incubated with active ERK2 and [␥-32 P]ATP and then analyzed by SDS-PAGE and autoradiography. B, recombinant GST-caspase-9 (WT) or mutated GST-caspase-9 lacking the site of phosphorylation (T125A) or mutated in the docking domain (7/10/ 11A) were incubated in HeLa cytosolic extract with OA for 60 min. GST proteins were recovered on glutathione-Sepharose beads and immunoblotted with antibodies to ERK1/2, phospho-Thr 125 caspase-9, or GST, as indicated. Experiments were conducted at least three times, and representative data from one experiment are shown. FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 compared with the wild type protein, whereas the double leucine mutation (L14A/L16A) only slightly reduced phosphorylation. This indicates that the one or more basic amino acids within this region of caspase-9 are important for recognition by ERK2.

Docking Interaction between Caspase-9 and ERK2
To examine whether the D domain of caspase-9 is required for the binding and phosphorylation by endogenous ERK1/2 under nearly physiological conditions, we tested the phosphorylation of recombinant caspase-9 in cell extracts that reproduce the regulation of caspase-9 in vitro (12). Previous studies have shown that incubation of such extracts with okadaic acid (OA), an inhibitor of PP1 and PP2A phosphatases, induces activation of endogenous ERK1/2 and phosphorylation of caspase-9 at Thr 125 (12). As expected, we found that GSTcaspase-9 (WT) recovered from cell extracts was phosphorylated on Thr 125 in response to treatment with OA (Fig. 3B). Coprecipitation of ERK1/2 with caspase-9 was also stimulated by OA. Interestingly, mutation of the Thr 125 phosphorylation site to the nonphosphorylatable amino acid alanine resulted in an increase in ERK1/2 binding. These results suggest that the interaction of ERK1/2 with caspase-9 is likely to be promoted by the phosphorylation and activation of ERK1/2, whereas subsequent phosphorylation of the substrate destabilizes the interaction. Phosphorylation of caspase-9 at Thr 125 was strongly reduced by mutation of basic residues of the D domain to Ala (R7A/R10A/R11A), and this decrease in phosphorylation correlated with loss of ERK1/2 binding to caspase-9 containing Thr 125 and a substantial decrease in binding to the T125A mutant, demonstrating that the basic residues in the D domain play an important role in the binding and phosphorylation of caspase-9 by ERK1/2 in cell extracts.
To investigate the sequence requirements of the D domain in more detail, residues were individually mutated to Ala, and recombinant proteins were purified for use as substrates (Fig.  4). Mutation of Arg-10 had a strong inhibitory effect, with Leu 9 and Arg 11 substantially reducing phosphorylation, indicating that these residues are particularly important for the recognition of caspase-9 by ERK2. Mutation of the hydrophobic residues Leu 14 and Leu 16 also partially inhibited caspase-9 phosphorylation but not as strongly as mutations of the basic residues Arg 10 and Arg 11 .
Peptides Mimicking the Docking Site Specifically Inhibit Phosphorylation of Caspase-9 by ERK2-As an additional test of the putative docking site in caspase-9, we tested the ability of peptides mimicking the motif to interfere with the phosphorylation of caspase-9 by ERK2. Three peptides were derived from the sequence of amino acid residues 4 -17 in caspase-9. WT peptide contained Arg 10 and Arg 11 , apparently key residues in the docking domain of caspase-9, whereas one or both of these sites were mutated to Ala in peptides R10A and R10A/R11A, respectively. An unrelated peptide (R-DIS) was used as an additional control (Fig. 5A). Increasing concentrations of WT peptide were found to inhibit phosphorylation of caspase-9 by ERK2, whereas peptide R10A inhibited to a lesser extent (Fig. 5B). Peptide R10A/R11A produced very poor inhibition of phosphorylation, confirming the importance of these residues in the interaction of caspase-9 with ERK2. Together with mutational analysis of caspase-9, these results confirm the role of the D domain in caspase-9 for its recognition and phosphorylation by ERK2 in vitro.
The D Domain of Caspase-9 Is Essential for Mediating Phosphorylation by ERK1/2 in Cells-We extended this work on the role of docking interactions for regulating caspase-9 phosphorylation by ERK1/2 to studies on intact cells. In initial experiments, we failed to observe any necessity for the docking site for the phosphorylation of caspase-9 at Thr 125 in cell lines transiently expressing caspase-9 (data not shown). However, since docking domains are thought to work by increasing the efficiency and specificity of phosphorylation by kinases, we were concerned that strong overexpression of caspase-9 and therefore a greatly increased substrate concentration may have swamped this level of regulation. We therefore opted for an inducible system that permitted expression of wild type caspase-9 and docking site mutants in cells at levels resembling that of the endogenous protein. Selective antibiotics were used to isolate human U2OS cell lines stably expressing the construct for either WT caspase-9 or docking site mutant caspase-9 (R10A). Expression of exogenous proteins was stimulated with increasing concentrations of the inducing agent ponasterone A (Pon A) (Fig. 6A). The exogenous caspase-9 proteins were HA-tagged to discriminate from endogenous caspase-9 and to allow their specific retrieval by immunoprecipitation. A concentration of 2.5 M ponasterone A for 24 h was chosen, since this induced caspase-9 to levels comparable with those of the endogenous protein in both wild type and docking site mutant cell lines (Fig. 6B).
To test if the docking site was required for caspase-9 phosphorylation by ERK1/2 in cells, expression of wild type and mutant caspase-9 was induced for 24 h followed by overnight serum starvation. Cells were then stimulated for 5 min with EGF before lysates were harvested and proteins were recovered by HA-immunoprecipitation. Fig. 5B shows equal induction of wild type and docking site mutant caspase-9 prior to immunoprecipitation and shows that immunoprecipitation of HA-caspase-9 was efficient in both lysates. We found that An example of a typical kinase assay is shown. A graphic representation of phosphorylation levels is also shown. The intensity of phosphorylation was standardized to that of WT (100%). The amino acid residues that were mutated into alanine are indicated (single-letter codes). Experiments were conducted at least three times, and representative data from one experiment are shown.
immunoprecipitated WT caspase-9 was phosphorylated on Thr 125 in response to treatment with EGF, correlating with activation of ERK1/2. In contrast, there was minimal phosphorylation of caspase-9 mutated at the docking site, despite ERK1/2 being similarly activated. Furthermore, the MEK1/2 inhibitor UO126 inhibited phosphorylation of wild type caspase-9 in response to EGF, showing again that phosphorylation was dependent on the ERK1/2 pathway (Fig. 6C). Thus, the D domain of caspase-9 is required for its phosphorylation by ERK1/2 in cells in response to a physiological stimulus.

ERK2 Recognition of Caspase-9 Does Not Require the CD Domain or Docking Groove Residues but Involves a TTCD Motif in ERK2-
Having identified a docking domain in caspase-9 as necessary for phosphorylation by ERK1/2, we wished to uncover the complementary regions of the kinase that participate in the interaction. This might reveal the basis for the selectivity of caspase-9 phosphorylation by ERK1/2. Several regions have been identified in MAPKs that play roles in docking interactions with other proteins (see Fig. 10 for a structural model of ERK2 indicating potential docking regions). Mutational studies with ERK2 have defined an acidic patch on the surface-exposed L16 loop of the kinase opposite to its catalytic cleft which acts as a MAPK conserved docking motif (CD site) common to substrates, activators, and inactivators (32).
To assess the relative importance of the CD domain in ERK1/2 docking to caspase-9, we made mutations in ERK2 at Asp 316 and Asp 319 , previously identified as key residues mediating CD interactions (Fig. 7A). Cells were transfected with the wild type and mutant ERK2 constructs for 24 h and then serum-starved before subsequent stimulation of the ERK1/2 pathway by EGF. Immunoprecipitation was then performed, followed by in vitro kinase assays using recombinant caspase-9 or p90 ribosomal S6 kinase (RSK) as substrates, with immunoprecipitated ERK as the source of active kinase. As shown in Fig. 6B, both single and double mutations in the CD domain of ERK2 did not inhibit its ability to phosphorylate caspase-9. In contrast, these mutations almost entirely abolished RSK phosphorylation by ERK2, confirming previously published results (33). This shows that, unlike RSK, these residues in the CD domain are not important for ERK2 docking to caspase-9.
Chang et al. (34) have identified a separate hydrophobic docking groove in p38␣ in which residues Ile 116 and Gln 120 are essential for recognition of a peptide containing a docking domain (Fig. 7A). Conservation of a docking groove similar to that of p38 has been proposed to exist in other MAPK family members (34). To examine the possibility that this type of interaction plays a role in the recognition of caspase-9 by ERK2, we GST-caspase-9 (0.5 g) was incubated with purified active His-ERK2 and [␥-32 P]ATP for 15 min at 30°C with the indicated concentrations of WT, R10A, R10A/R11A, or R-DIS peptides. Caspase-9 phosphorylation was analyzed by SDS-PAGE and autoradiography (top). Gels were analyzed by staining with Coomassie Blue for visualization of caspase-9 to verify equal amounts in the reaction (bottom). Results are plotted as percentage phosphorylation relative to that observed in the absence of any added peptide. Experiments were conducted at least three times, and representative data from one experiment are shown. FEBRUARY 15, 2008 • VOLUME 283 • NUMBER 7 mutated residues within the proposed docking groove of ERK2. Mutation of Leu 113 and Gln 117 failed to significantly inhibit caspase-9 phosphorylation by ERK. Indeed, we found that mutation of Gln 117 greatly increased phosphorylation. These mutations had little effect on ERK2 phosphorylation of RSK (Fig.  7B). Thus, key residues within the proposed hydrophobic docking groove are not required for ERK2 docking to either caspase-9 or RSK, which argues against this mechanism of interaction being important for the recognition of these substrates.

Docking Interaction between Caspase-9 and ERK2
Last, we tested whether residues within an adjacent acidic patch of ERK2 might play a role in interaction with caspase-9. The sequence Thr 157 -Thr 158 -Cys 159 -Asp 160 ( 157 TTCD 160 ) located within this acidic patch is different from that found in p38␣ (Glu 160 -Asp 161 -Cys 162 -Glu 163 ), which has been referred to as the ED site (35), whereas other important residues are largely conserved. In JNK1, the corresponding residues are Ser 161 -Asp 162 -Cys 163 -Thr 164 . This region may therefore play a role in the ability of different MAPKs to distinguish between substrates containing D domains (31,35). We found that mutation of both Thr 157 and Thr 158 to Ala (T157A/T158A) greatly reduced phosphorylation of caspase-9 by ERK2 and had a lesser effect on the phosphorylation of RSK in vitro. Similarly, mutation of Thr 157 and Thr 158 to a more "p38-like" sequence (T157E/T158D) also dramatically decreased phosphorylation of caspase-9 by ERK2 (Fig. 8). This was in contrast to mutation to a more "JNK-like" sequence (T157S/T158D), which had no inhibitory effect on either the phosphorylation of caspase-9 or RSK.
Mutation of Asp 160 in ERK2 to Ala or Asn also substantially reduced phosphorylation of caspase-9 (Fig. 9). Similarly, mutation to Glu, as found in p38␣ at this position, or Thr, as in JNK2, greatly inhibited phosphorylation of caspase-9, showing the critical importance of Asp 160 in ERK2 for its ability to recognize and phosphorylate caspase-9. By contrast, although mutation of Asp 160 to Asn or Thr greatly decreased ERK2 phosphorylation of RSK, mutation to alanine or glutamate had little effect, so there are likely to be differences between the interaction of this region of ERK2 with caspase-9 and RSK. Together, these data show that the TTCD motif within an acidic patch on ERK2 can account for its ability, in contrast to p38␣ and JNK, to recognize and phosphorylate caspase-9 at Thr 125 . Comeau et al. (24) as a nonbiased predictor of complex formation between phosphorylated and active ERK2 (23) and the prodomain of caspase-9 (21,22). The best fit model, derived from shape complementarity and then desolvation and electrostatic energies, showed an interaction involving the D domain of caspase-9 with ERK2, consistent with our identification of this domain in caspase-9 by mutational and deletion analysis (Fig. 10). The total solventaccessible surface area buried between the two domains is 1489 Å, a value that is comparable with those observed for the formation of stable protein-protein interfaces (e.g. an antibodyantigen complex). Hydrogen bonds involving residues Arg 36 with Glu 312 , Gln 57 with Ser 120 , Arg 65 with Gln 313 , Asp 42 with Gln 313 , Asp 61 with Tyr 314 , and Glu 63 with Tyr 126 are established between caspase-9 and ERK2, respectively. Most notably, a hydrogen bond between Arg 13 with Thr 157 and a salt bridge between Arg 10 and Asp 160 of ERK2 and caspase-9, respectively, are formed, in strong agreement with our identification of these residues as being critical for the interaction between ERK2 and caspase-9. Together with the mutational analysis, this strongly suggests that the TTCD motif of ERK2 interacts directly with the D domain of caspase-9, and this interaction is critical for the recognition of caspase-9 as a substrate by ERK2. Additional runs of ClusPro performed using structures of ERK2 where Asp 160 was "in silico" mutated to either Ala, Thr, or Asn resulted in the abolishment of the binding mode previously observed, suggesting that Asp 160 is crucial for this modeled interaction.

DISCUSSION
We have previously established caspase-9 as a target of ERK1/2, with phosphorylation at a single major site, Thr 125 , inhibiting caspase-9 activation (12). Here, we have revealed a docking mech-  A). B, Western blot analysis of endogenous levels of wild type HA-caspase-9 (WT) or mutated HA-caspase-9 (10A) induced in stably transfected U2OS cell lines by 2.5 M ponasterone A. Samples are shown prior to (pre) or after (post) immunoprecipitation of HA-caspase-9. C, phosphorylation of caspase-9 by ERK1/2 stimulated by EGF is inhibited by mutation of Arg 10 . WT or mutant R10A (10A) HA-caspase-9 expression was induced in U2OS stable cell lines by 2.5 M ponasterone A for 24 h before serum starvation overnight. Cells were then stimulated for 5 min with 50 ng ml Ϫ1 EGF before HA-caspase-9 was precipitated from cell lysates with an HA-antibody/ agarose conjugate. Immunoprecipitates (IP) were immunoblotted with antibodies to phospho-Thr 125 caspase-9 or caspase-9 (top). Cell lysates were also immunoblotted with antibodies to phospho-ERK1/2 and ERK1/2 (bottom). Experiments were conducted at least three times and representative data from one experiment is shown.
anism for the interaction of ERK1/2 with caspase-9. This interaction is related to previously identified interactions between MAPKs and their substrates/regulators but shows distinctive features. We propose that our findings uncover the molecular basis for selective phosphorylation of caspase-9 by ERK1/2 but not other MAPKs, such as p38␣ and JNK.
The sequence of the docking site that we have identified in caspase-9 that enables interaction with ERK1/2 resembles previously characterized D domains in other MAPK substrates, defined as a cluster of basic amino acids followed by a hydrophobic motif ( A -X-B , where is a small hydrophobic amino acid) (30). Several residues in the N-terminal prodomain region of caspase-9 are important for ERK1/2 docking and phosphorylation of caspase-9, with Arg 10 being particularly significant. Arg 10 , Arg 11 , and other basic residues are exposed on the surface of this domain in caspase-9 and form a highly positively charged patch that is available for interaction with other proteins (21,22). However, Leu 14 and Leu 16 , which would form the hydrophobic A -X-B motif for the D domain in caspase-9, appear to be less important for phosphorylation by ERK1/2. Although it has been suggested that hydrophobic residues of the A -X-B motif in MAPK substrates are more critical for docking than other residues in the docking domain (30,36,37), in some MAPK substrates, such as RSK, this feature is not conserved (38). Therefore, although many MAPK substrates share related D domain docking motifs, in any given substrate, the specificity determinants for docking may differ, for instance in the number or spacing of basic residues or the importance of a hydrophobic A -X-B motif for docking interactions.
In some MAPK substrates, additional motifs have been recognized that play a role in interaction with the kinase. Originally described as a specific docking domain for ERK1/2, recent studies have shown that the DEF (docking site for ERK) motif (39) may also allow docking of p38␣ (40). With regard to caspase-9, we found that mutation of potential DEF sites consisting of Phe residues in the C-terminal half of caspase-9 did not reduce the capacity of ERK2 to phosphorylate caspase-9 (data not shown). Given the observation that mutation of the D domain strongly reduces phosphorylation by ERK1/2 both in vitro and in cells, our studies indicate that the D domain is the major docking site in caspase-9 that directs phosphorylation at Thr 125 by ERK. Nevertheless, weak binding of ERK1/2 to the caspase-9 R7A/R10A/R11A mutant, particularly when combined with the T125A mutation, does suggest an additional mode of interaction that is independent of the D domain. Residual phosphorylation of D domain mutants in extracts and cells could be due to a weak interaction with ERK1/2 or might represent a relatively minor additional kinase activity that targets Thr 125 independently of the D domain.
We have also investigated the regions of ERK2 that are involved in the interaction with caspase-9. We found no evidence that key acidic residues within the CD domain (32) of ERK2 are required; although mutation of Asp 316 and Asp 319 did FIGURE 8. Requirement for the TTCD motif of ERK2 for efficient phosphorylation of caspase-9. The indicated 157 TTCD 160 motif HA-ERK2 mutants were expressed in HEK 293 cells, immunoprecipitated and then used in kinase reactions using either GST-caspase-9, GST-RSK, or MBP (each 0.5 g) as a substrate and [␥-32 P]ATP. The 32 P-labeled bands detected by autoradiography that correspond to GST-caspase-9, GST-RSK, and MBP, identified by protein staining, are shown. Levels of HA-ERK2 (ERK) and phosphorylated HA-ERK2 (pERK) in the immunoprecipitates are shown by immunoblotting. Quantification of phosphorylation of GST-caspase-9 and GST-RSK, shown below, was standardized against activity toward MBP. Experiments were conducted at least three times, and representative data from one experiment are shown. FIGURE 9. Mutation of Asp 160 in ERK2 inhibits phosphorylation of caspase-9. HA-ERK2 mutants of Asp 160 were expressed in HEK 293 cells, immunoprecipitated, and then used in kinase reactions using either GSTcaspase-9, GST-RSK, or MBP (each 0.5 g) as a substrate and [␥-32 P]ATP. The 32 P-labeled bands detected by autoradiography that correspond to GSTcaspase-9, GST-RSK, and MBP, identified by protein staining, are shown. Levels of HA-ERK2 (ERK) and phosphorylated HA-ERK2 (pERK) in the immunoprecipitates are shown by immunoblotting. Quantification of phosphorylation of GST-caspase-9 and GST-RSK, shown below, was standardized against activity toward MBP. Experiments were conducted at least three times, and representative data from one experiment are shown.
not affect phosphorylation of caspase-9, these mutations inhibited phosphorylation of RSK, showing a clear difference in the recognition of these two substrates by ERK2. We also found no evidence that the hydrophobic docking groove identified in p38␣ and proposed to be present in other MAPKs by Chang et al. (34) is involved in the interaction of ERK1/2 with caspase-9 or indeed with RSK.
Our study did, however, reveal that the sequence 157 TTCD 160 in ERK2 governs docking to caspase-9. We refer to this site in ERK2 as the TTCD motif. Interestingly, conversion of Thr 157 and Thr 158 to Glu and Asp, respectively, as found at the corresponding positions in p38␣ ("ED site") (35), significantly reduced the ability of ERK2 to phosphorylate caspase-9, whereas mutation to Ser and Asp, found in JNK1, did not. This suggests that Thr 157 in the 157 TTCD 160 motif of ERK2 is particularly important for the interaction with caspase-9. Although a conservative change in this position to Ser is tolerated, an acidic residue (Glu) is not. Notably, a dramatic reduction in phosphorylation of caspase-9 by ERK2 was observed when Asp 160 was mutated. By contrast, recognition of RSK by ERKs tolerated mutation of Asp 160 to alanine or glutamate, although not asparagine or threonine. This residue has also been shown to be important for the interaction of ERK2 with the tyrosine phosphatases PTP-SL and STEP (41) and its regulator MAPK phosphatase 3 (42), whereas in Drosophila ERK2 D160N is a gain-of-function mutation (43). Thus, it appears that Asp 160 is a critical residue for the docking interactions of ERK2 with certain upstream regulators and some but not all substrates.
Recently, Liu et al. (31) determined the crystal structure of ERK2 bound to a D domain peptide derived from MKP3 (KIM MKP3 ) and showed that an acidic patch and a hydrophobic groove in ERK2 engaged the basic and A -X-B residues, respectively, in the peptide. The acidic patch included CD domain residues Asp 316 and Asp 319 , together with Asp 160 , whereas the hydrophobic groove included Leu 113 . Our molecular modeling indicates the likely basis for the similarities and differences between the interaction of KIM MKP3 and the prodomain of caspase-9 with ERK2. Similar to KIM MKP3 , the interaction between the prodomain of caspase-9 and ERK2 is likely to involve a salt bridge formed between Asp 160 of ERK2 and Arg 10 within the D domain of caspase-9, which would account for the essential role of these residues in the interaction of the proteins. However, the CD domain residues Asp 316 and Asp 319 of ERK2 do not form close contacts with the prodomain of caspase-9, being closest to the neutral residues of the N-terminal of caspase-9. This may account for the lack of requirement of these residues in ERK2 for the recognition of caspase-9 as a substrate. Furthermore, the docking groove in ERK2 also does not have close contacts with the proposed hydrophobic A -X-B motif (Leu 14 -Arg 15 -Leu 16 ) or indeed other hydrophobic residues in the caspase-9 prodomain. This may explain the lack of requirement for Leu 113 and Gln 117 of ERK2 and the only minor role of Leu 14 and Leu 16 of caspase-9 in the recognition of caspase-9 by ERK2. Overall, the prodomain of caspase-9 is predicted to form a smaller area of interaction with ERK2 than MKP3 that is highly dependent on the interaction of Asp 160 with Arg 10 (Fig. 10). Firm conformation of this interaction will, however, require solution of the crystal structure of the ERK2caspase-9 complex.
Importantly, mutation of Asp 160 in ERK2 to residues found in either p38␣ or JNK strongly inhibited its ability to phospho- FIGURE 10. Molecular interaction between ERK2 and caspase-9. The structure of phosphorylated active ERK2 (23) is shown on the left with previously described regions involved in docking interactions highlighted. The hydrophobic docking groove is shown in orange, the TTCD motif in red, and the CD domain in green. The structure of the caspase-9 prodomain (21,22) is shown on the left, with the D domain highlighted in blue. In the center, the structure of the docked complex predicted by ClusPRO is shown, with the salt bridge formed between Asp 160 of ERK2 and Arg 10 of caspase-9 expanded. rylate caspase-9. Thus, the TTCD motif can account for the ability of ERK1/2, but not p38 and JNK, to phosphorylate and inhibit caspase-9 activation. This specificity and recognition of a critical component of the intrinsic apoptotic pathway may be significant for the ability of the classical MAPKs to suppress the apoptosis, whereas the stress-activated p38␣ and JNK pathways often promote apoptosis (20).
Consistent with our previous studies (12), our results also show that the interaction of ERK1/2 with caspase-9 in cell extracts is enhanced when ERK1/2 is phosphorylated and activated by OA treatment (Fig. 4). This suggests a specific interaction of caspase-9 with the activated form of ERK1/2, although a role for another phosphorylation event stimulated by OA cannot be excluded. Conversely, the interaction of caspase-9 with ERK1/2 is greatly increased by mutation of the phosphorylated residue in caspase-9, Thr 125 , to nonphosphorylatable alanine. This indicates that once caspase-9 is phosphorylated by ERK1/2, the affinity of the interaction is decreased. These results suggest a model (Fig. 11) for the dynamic interaction of ERK1/2 with caspase-9, in which active ERK1/2 interacts through its TTCD motif with the D domain on caspase-9. In addition to this tethering interaction, there are likely to be other interactions that bring the phosphorylation site in caspase-9 in close proximity to the active site of ERK2. Once ERK1/2 catalyzes the phosphorylation of Thr 125 , it appears to be released, presumably through a conformational change in caspase-9 that disrupts the interactions between the proteins. Thus, the catalytic efficiency of ERK2 toward caspase-9 would be enhanced.
Interference with the docking interaction by small molecules could potentially prevent the inhibition of caspase-9 activation by phosphorylation at Thr 125 and therefore promote apoptosis, for instance in cancer cells in which classical MAPKs are constitutively activated. The differences in the interaction of ERK2 with caspase-9 and other substrates suggest that specificity in the modulation of this anti-apoptotic function might be achieved while maintaining other essential functions of the kinase. However, the basic region near to the N terminus of caspase-9, including Arg 10 and Arg 11 , is also involved in the interaction with acidic residues in the CARD of Apaf-1 (21,22). Thus, promotion of apoptosis by preventing the interaction of ERK1/2 with caspase-9 would need to be selective for interaction of the kinase and not simultaneously prevent the binding of Apaf-1, which is required for caspase-9 activation. Indeed, it is possible that ERK1/2 and Apaf-1 can each antagonize the other's antiapoptotic and proapoptotic functions, respectively, through competition for binding to caspase-9.