TAB1-Induced Autoactivation of p38α Mitogen-Activated Protein Kinase Is Crucially Dependent on Threonine 185

ABSTRACT p38α mitogen-activated protein kinase is essential to cellular homeostasis. Two principal mechanisms to activate p38α exist. The first relies on dedicated dual-specificity kinases such as mitogen-activated protein kinase kinase (MAP2K) 3 (MKK3) or 6 (MKK6), which activate p38α by phosphorylating Thr180 and Tyr182 within the activation segment. The second is by autophosphorylation of Thr180 and Tyr182 in cis, mediated by p38α binding the scaffold protein TAB1. The second mechanism occurs during myocardial ischemia, where it aggravates myocardial infarction. Based on the crystal structure of the p38α-TAB1 complex we replaced threonine 185 of p38α with glycine (T185G) to prevent an intramolecular hydrogen bond with Asp150 from being formed. This mutation did not interfere with TAB1 binding to p38α. However, it disrupted the consequent long-range effect of this binding event on the distal activation segment, releasing the constraint on Thr180 that oriented its hydroxyl for phosphotransfer. Based on assays performed in vitro and in vivo, the autoactivation of p38α(T185G) was disabled, while its ability to be activated by upstream MAP2Ks and to phosphorylate downstream substrates remained intact. Furthermore, myocardial cells expressing p38α(T185G) were resistant to injury. These findings reveal a mechanism to selectively disable p38α autoactivation and its consequences, which may ultimately circumvent the toxicity associated with strategies that inhibit p38α kinase activity under all circumstances, such as with ATP-competitive inhibitors.

target" effect since it occurs with diverse scaffolds binding to different locations in or around the ATP-binding pocket. The occurrence of toxicity is not surprising considering the ubiquitous nature of p38␣ expression and its involvement in signaling networks that maintain normal cell function. For example, in cardiomyocytes p38␣ itself plays a beneficial role in promoting protection in preconditioning but is harmful in myocardial ischemia (13,14). Targeting TAB1 would be futile, as its knockout causes embryonic lethality and it plays an important role in the TAK1 signaling pathway (15,16). As a result, selectively targeting the TAB1-dependent autoactivation of p38␣ may provide an opportunity to avoid the toxicity seen with complete inhibition of p38␣ kinase activity, while still deriving therapeutic benefit.
To target the interaction between p38␣ and TAB1, we need to understand the structural detail of the p38␣-TAB1 complex at an atomic level. Therefore, we previously solved the crystal structure of p38␣ and the interacting region of TAB1, which showed that long-range structural changes within p38␣ result in an increased affinity for ATP (6). The N-terminal and C-terminal lobes of p38␣ move toward each other, and this movement positions the activation loop containing Thr180 and Tyr182 toward the catalytic site, which facilitates autophosphorylation. TAB1 binding also causes residues Tyr182 to Thr185 to extend into a short ␣-helical segment, accompanied by hydrogen bond formation between the side chains of Thr185 in the activation segment and Asp150 of the HRD motif (Fig. 1A). This hydrogen bond holds the activation loop in its new position where Thr180 and Tyr182 lie close to the catalytic site, likely promoting the autophosphorylation reaction. In keeping with these observations, Thr185 and Asp150 are conserved across p38 isoforms and species (Fig. 1B).
In this study, we hypothesized that preventing the formation of the hydrogen bond between Thr185 and Asp150 will interfere with the autophosphorylation of Thr180 and Tyr182, since the required and unique orientation of the activation segment will become energetically unfavorable. To investigate our hypothesis, we mutated Thr185, since Asp150 lies within the HRD motif, which is involved in substrate recognition and forms part of a central spline conserved among all protein kinases (17).

RESULTS
TAB1 still binds p38␣(T185G). p38␣ autoactivation is caused by TAB1 interacting with p38␣; i.e., without the p38␣-TAB1 interaction, the autophosphorylation of p38␣'s T-G-Y motif is not promoted. Before testing our hypothesis that the mutation in p38␣(T185G) impairs TAB1-induced p38␣ autoactivation, we needed to confirm that the p38␣(T185G) mutant kinase is able to bind TAB1. Using isothermal titration calorimetry (ITC), we found that p38␣(T185G) binds to TAB1 with an affinity similar to that of wild-type p38␣ (wtp38␣) ( Table 1). In our biophysical and crystallographic experiments, we have used a peptide spanning residues 384 to 412 of TAB1 [TAB1(384 -412)], and we have previously shown that no other residue beyond this region is involved in the interaction with p38␣ and that the peptide can recapitulate the behavior of full-length TAB1 in vivo (6). p38␣(T185G) prevents TAB1-induced autophosphorylation without interfering with catalytic activity or preventing phosphorylation by upstream MAP2K. To test our hypothesis, we carried out an in vitro kinase (IVK) assay using recombinant wtp38␣ or p38␣(T185G) in the presence of TAB1(384 -412). TAB1 caused autoactivation of wtp38␣, whereas it was not able to cause autoactivation of p38␣(T185G) ( Fig. 2A). This result suggests that the hydrogen bond that forms between threonine 185 and aspartic acid 150 is essential in the process of p38␣ autoactivation caused by TAB1. To further examine this mechanism, we ectopically expressed p38␣(T185G) in mammalian HEK293 cells together with TAB1. At 24 h after overexpression there was a clear increase in the phospho-p38␣ (T-G-Y) signal in cells transfected with wtp38␣ and TAB1; however, no such increment was observed in cells transfected with p38␣(T185G) and TAB1 (Fig. 3A). These results in HEK293 cells recapitulate the result from the in vitro kinase assay and provide concrete evidence to support our hypothesis that the hydrogen bond formed between threonine 185 and aspartic acid 150 is a prerequisite for TAB1-induced p38␣ autoactivation.
Having obtained results in support of our hypothesis, we next examined whether hydrogen bond formation had a similar role in p38␣'s classical activation pathway. To investigate this, we carried out an in vitro kinase assay with p38␣ and the dualspecificity kinase MAP2K6, which is an upstream activator of p38. In an IVK reaction, the constitutively active MAP2K6dd was able to activate p38␣(T185G) in a manner similar FIG 1 (A) Superimposition of the crystal structures of free wtp38␣ (blue) and wtp38␣ (gray) in complex with TAB1 (green). In the free structure the side chain oxygens of Thr185 and Asp150 are separated by 5.6 Å, which prevents the formation of a hydrogen bond, and the activation loop is facing away from the catalytic site, a position not compatible with the autophosphorylation reaction. Upon TAB1 binding, Tyr182 to Thr185 form an alpha-helical segment that is stabilized by the hydrogen bond formed between Thr185 and Asp150 with side chain oxygens now 2.5 Å apart, typical of hydrogen bond formation. Consequently, the activation loop moves toward the catalytic site, bringing the T-G-Y motif in proximity to the ␥-phosphate position of ATP, within its binding site, a position compatible with the autophosphorylation reaction. (B) Asp150 and Thr185 are conserved across p38 isoforms, species, and other mitogen-activated protein kinases. to that for wtp38␣ (Fig. 2C). We obtained the same result when we transfected HEK293 cells with the p38␣(T185G) and MAP2K3 (Fig. 3A). MAP2K3 and MAP2K6 equally activated both wtp38␣ and p38␣(T185G), suggesting that the classical activation pathway is not affected by the hydrogen bond between threonine 185 and aspartic acid 150. Next, we examined whether the catalytic activity of p38␣(T185G) was affected. We carried out an in vitro kinase assay with active p38␣(T185G) or active wtp38␣ and activating transcription factor 2 (ATF2) or the scaffold protein TAB1. ATF2 is a classic substrate of p38␣, and TAB1 is both an activator of p38␣ through autophosphorylation and a substrate of p38␣ (16). The results from these IVK assays showed that p38␣ (T185G) phosphorylates ATF2 and TAB1 in a manner similar to that for wtp38␣ ( Fig. 2B and D). These results suggest that the hydrogen bond between threonine 185 and aspartic acid 150 does not impact p38␣'s kinase activity toward its downstream substrates. They also reinforce the conclusions of the ITC experiment described in Table  1; namely, p38␣(T185G) has an affinity toward TAB1 that is similar to that of wtp38␣. p38␣(T185G)-TAB1 complex crystal structure. We then solved the X-ray structure of the p38␣(T185G)-TAB1 complex (PDB code 5O90) ( Table 2) and compared it with that of the wtp38␣-TAB1 complex (PDB code 4LOO) (Fig. 4). As expected, the crystal structure revealed many features that are shared between these complexes. The mutation of p38␣ does not affect the TAB1 interaction, and in both structures TAB1 binds in a bipartite manner on the kinase C lobe and induces conformational changes that propagate through p38␣: the N-and C-terminal lobes of p38␣ move toward each other, causing significant closure around the ATP-binding pocket. In the wt complex as part of this rearrangement, threonine 180 of the T-G-Y motif flips orientation so that it points into the active site, and residues Tyr182 to Thr185 of the activation loop form a short helical structure stabilized by a hydrogen bond between threonine 185 and aspartic acid 150; the remainder of the activation loop is also well ordered and clearly visible in the crystal structure. In contrast to this, in the T185G structure the majority of the activation loop is disordered and not visible (residues 173 to 181); the first turn of the helix from Tyr182 to Gly185 is still present as tyrosine 182 forms a crystal contact stabilizing this arrangement, but the helix has relaxed away from aspartic acid 150 and there is a substantial increase in the B factors of the helix residues (Fig. 4A).
Moreover, the T185G mutation has had a profound impact on both the absolute B factor and B-factor distribution. As shown in Fig. 4A, the center of the T185G mutant complex close to the hinge region is still reasonably well ordered, but there is a strong correlation between distance from the hinge and increased average residue B factor, as opposed to the case for the wt structure. Consistent with these data, we propose that TAB1 in both structures causes the N and the C lobes of p38␣ to move toward each other, but the key hydrogen bond formed by threonine 185 and aspartic acid 150 is then required to lock the two domains together in the closed state. In the mutant structure, the absence of the hydrogen bond causes significant flexibility between the N-and C-terminal domains in the crystal structure, thus explaining the observed B-factor distribution; in addition, the T-G-Y motif is no longer oriented toward the ATP-binding site, thus impeding p38␣ autoactivation. and TAB1 or MKK3. TAB1-mediated activation of p38␣ is impaired in the mutant, whereas no difference is detected between the wt and the T185G mutant with MKK3-mediated activation. (B) HEK293 cells transfected with wtp38␣ or p38␣(T185G) exposed for 10 min to a buffer simulating ischemia. p38␣ activation is SB203580 sensitive, confirming autophosphorylation. Arrows indicate ectopic p38␣, which is hemagglutinin tagged and heavier then endogenous p38␣. (C) Quantification of phopho-p38␣ normalized against total p38␣ in HEK293 cells exposed to simulated ischemia (n ϭ 3). *, P Ͻ 0.05 versus wtp38␣ control; #, P Ͻ 0.05 versus wtp38␣ ischemia.
Role of T185 in TAB1-Induced Autoactivation of p38␣ Molecular and Cellular Biology p38␣(T185G) inhibits TAB1-induced cell damage in HEK293 and H9C2 cells. To further validate the selective inhibition of p38␣, we transfected HEK293 cells with either wtp38␣ or p38␣(T185G) and subjected them to simulated ischemia, an experimental setup that tries to recapitulate the low-oxygen and low-pH conditions experienced by cells during ischemia, and we relied on endogenous TAB1 to cause p38␣ activation. After 10 min of incubation in ischemic buffer, we detected increased phospho-p38␣ signal in cells transfected with wtp38␣ (Fig. 3B). This was inhibited by the ATPcompetitive p38␣ inhibitor SB203580 (10 M), confirming the mode of p38␣ activation to be autoactivation. In cells transfected with p38␣(T185G), we observed a diminished phospho-p38 signal with ischemia, confirming impaired autoactivation.
We next created a stable H9C2 myoblast cell line expressing either wtp38␣ or p38␣(T185G). The cells were incubated with TAB1(384 -412) peptide fused with a cell-permeative peptide derived from the HIV1 transactivator of transcription (TAT) overnight at 20 M. The next day, the activity of lactate dehydrogenase (LDH), a marker of cell injury, was measured in the cell medium. The cells expressing wtp38␣ incubated with the TAB1 peptide produced a substantial increase in LDH activity compared to cells incubated with the scrambled form of TAB1 peptide. In cells expressing p38␣(T185G), the TAB1 peptide incubation resulted in less LDH release than in the cells expressing wtp38␣ (Fig. 5A). The amount of LDH release correlated with the level of p38␣ activation (Fig. 5B). The result suggests that TAB1 causes cell injury via p38␣ activation, and the T185G mutation significantly diminishes the autophosphorylation of p38␣ and release of LDH.

DISCUSSION
In this study, we have revealed a key structural change that occurs in p38␣ during TAB1-mediated p38␣ autoactivation. Previously we found that TAB1 binds to p38␣ in a bipartite manner, which causes a structural rearrangement within p38␣ that leads to its autoactivation (6). Here, we have built on this observation to show that the formation of a hydrogen bond between Thr185 and Asp150 is an important event in the autoactivation process. Using recombinant protein in IVK assays and exogenous protein expression in mammalian cells, we have shown that preventing the formation of this hydrogen bond significantly reduces p38␣ autoactivation and subsequent cellular damage. We have also solved the crystal structure of the p38␣(T185G)-TAB1 complex to show that the hydrogen bond between Thr185 and Asp150 is crucial in initiating the autophosphorylation reaction, providing the energy to orient the T-G-Y motif toward the kinase catalytic site. In the mutant structure, as a result of the absence of the hydrogen bond between Gly185 and Asp150, the C-terminal end of the activation loop (Tyr182 to Thr185) has a higher temperature factor than in the wild type, the activation loop (Leu171 to Val181) is highly mobile to a point that it is not visible in the X-ray structure, and the T-G-Y motif is no longer locked next to the kinase catalytic site. Every attempt to create a clinically effective small molecule to inhibit p38␣ has been unsuccessful. Despite many clinical trials with several drugs, none has passed the phase III stage, with losmapimod being the latest drug to fail (18). The major issue has been toxicity, such as liver damage, skins rashes, and gastrointestinal disorders, which is not surprising considering the ubiquitous nature of p38␣ expression and its nodal position in cell signaling (19,20). Most p38␣ inhibitors are type I ATP-competitive inhibitors, which compete with ATP and inhibit p38's catalytic activity under all circumstances (21). Due to the high structural similarity of the ATP-binding sites, they nonselectively block alpha and beta isoforms of p38 and in some cases also the gamma and delta isoforms, depending upon their achieved concentration (22). This results in blanket inhibition of p38 under all circumstances, including during beneficial homeostatic function, thereby leading to harmful side effects. An alternative strategy is needed whereby we avoid blanket inhibition of p38 and aim to achieve circumstance-selective inhibition. This makes the TAB1-mediated p38␣ activation pathway an attractive target, since it may be a gateway to ischemia-selective inhibition. The data we have collected have unveiled a critical structural change in p38␣ during ischemia-triggered autoactivation induced via TAB1. Our findings reveal a mechanism that can selectively block p38␣ activation caused by TAB1 without affecting the classical activation pathway or other isoforms. This mechanism proves the concept that it is possible to selectively disable p38␣ catalytic activity and potentially achieve circumstance-selective inhibition of p38␣ during myocardial ischemia.

MATERIALS AND METHODS
Cloning and expression of p38␣/p38␣(T185G) in mammalian cells. p38␣(T185G) was cloned in two steps. First, the overlapping N-terminal and C-terminal regions were produced with complementary internal primers containing the desired mutations. In the second step, the PCR products from the first reaction were combined to form the template for a second PCR with external primers to produce a full-length p38␣ harboring the desired mutation. wtp38␣, p38␣(T185G), MKK3b, and TAB1 were cloned into pCDNA3 containing the cytomegalovirus (CMV) immediate early promoter and simian virus 40 (SV40) polyadenylation sequence for high-level expression in mammalian cells. wtp38␣, p38␣(T185G), and MKK3b had an N-terminal hemagglutinin (HA) tag, whereas TAB1 was CYP tagged. Gateway cloning technology was used to clone p38␣/p38␣(T185G) with a V5 tag into pCDNA6.2 that had the blasticidin resistance gene.
Transfection. HEK293 cells were seeded on 6-well plates in full growth medium (Dulbecco's modified Eagle's medium containing 584 mg/liter L-glutamine, 10% [vol/vol] fetal bovine serum, 1% [vol/vol] penicillin-streptomycin) at 37°C and transfected upon reaching 60% confluence. The transfection was done using Turbofect reagent. One hundred microliters of Opti-MEM solution was added to an Eppendorf tube, followed by 1 g of DNA sample. Two microliters of Turbofect reagent was added to the tube, mixed gently, and left for 30 min for complex formation. An appropriate volume of Opti-MEM was added into the tube to make 1 ml total volume. The tube was inverted gently for complete mixture of the transfection reagents, which were then pipetted onto the cells in 6-well plates in a dropwise manner. The cells were incubated overnight at 37°C, after which the transfection medium was replaced with serum-free medium (SFM). In the experiments with inhibitors, drugs were incubated with SFM for the appropriate amount of time as needed. For Western blot analysis, the SFM was removed, cells were washed with phosphate-buffered saline (PBS), and 200 l of 2ϫ SDS sample buffer (120 mM Tris [pH 6.8], 6% [wt/vol] SDS, 20% [vol/vol] glycerol, 10% ␤-mercaptoethanol, and 0.01% bromophenol blue) was used to collect the cells. For stable transfection, H9C2 cells were transfected as described above but in the presence of blasticidin at 8 g/ml. The cells were monitored over the next 2 weeks, replacing the medium and antibiotics every 3 days. At the end of 2 weeks, surviving cells stably expressing the protein were selected, transferred into a T75 flask, and grown normally under blasticidin selection at 8 g/ml. Ischemia buffer and cell toxicity assay. The recipe for the ischemia buffer was adapted from that described by Esumi et al. (23). Briefly, 50 ml of ischemic buffer (127 mM NaCl, 3.58 mM KCl, 0.49 mM MgCl 2 · 6H 2 O, 1.8 mM CaCl 2 · 2H 2 O, 4 mM HEPES, pH 7.4) was made up using 5 ml of 10ϫ basic stock solution and 45 ml of sterile deionized H 2 O; 10 mM 2-deoxy-D-glucose (2-DG) and 20 mM Na lactate were added to the 50-ml solution, filter sterilized using a 0.2-m syringe, and warmed to 37°C (the pH of the solution at this point is 6.8). Once the solution was warmed, 1 mM Na dithionite was added to the solution (the pH drops to 6.3) and mixed before quickly adding it onto the cells. The cells were then incubated in room air supplemented with 5% CO 2 in a water-saturated incubator at 37°C for 10 min. After 10 min, cells were collected using 200 l of 2ϫ SDS buffer and analyzed by Western blotting. To measure the cell toxicity caused by TAB1 peptide, an LDH assay kit (MAK066; Sigma) was used. H9C2 cells were incubated with 20 M TAB1 peptide in SFM overnight, and the cell medium was used to carry out the LDH test. The TAT-TAB1(384 -412) peptide was purchased from Activotech. The TAB1 sequences are as follows: TAT-TAB1 (wt), GYGRKKRRQRRRGEMSQPTPTPAPGGRVYPVSVPYSSAQSTSKTSVTLSLVMPSQGQMV; TAT-TAB1 (scrambled), GYGRKKRRQRRRGTVSQGPASATPAYPPGGRVESQPLPSQSTLGAPSKTSVTSAVQSMM.
Expression and purification of p38␣/p38␣(T185G) in Escherichia coli. p38␣ and p38␣(T185G) were cloned in the first multiple-cloning site of the pETDuet vector. The p38␣ gene was cloned with a His 6 tag followed by a TEV cleavage site at the N-terminal end. The DNA was transformed in E. coli strain Rosetta II cells (Novagen), and ampicillin was used for selection. The protein expression and purification were carried out as described in reference 24.
ITC. Isothermal titration calorimetry (ITC) was carried out in an ITC200 microcalorimeter from Microcal (GE Healthcare Life Sciences). Microcal Origin 7.0 data software was used to analyze the result. Integrated heat data obtained for the titrations corrected for heats of dilution were fitted with a nonlinear least-squares minimization algorithm to a theoretical titration curve. The data fitting gave us the values for binding stoichiometry (n), binding affinity (K), and molar enthalpy change (ΔH). The molar entropy (ΔS) was calculated using ΔS ϭ ΔH Ϫ ΔG, where ΔG ϭ ϪRT ϫ ln K (R ϭ 1.987 cal mol Ϫ1 K Ϫ1 and T ϭ 298 K).
Accession number(s). The X-ray structure of the p38␣(T185G)-TAB1 complex has been deposited under PDB code 5O90.