T Cell Receptor-mediated Activation of p38α by Mono-phosphorylation of the Activation Loop Results in Altered Substrate Specificity*

p38 MAPKs are typically activated by upstream MAPK kinases that phosphorylate a Thr-X-Tyr motif in the activation loop. An exception is the T cell antigen receptor signaling pathway, which bypasses the MAPK cascade and activates p38α and p38β by phosphorylation of Tyr-323 and subsequent autophosphorylation of the activation loop. Here we show that, unlike the classic MAPK cascade, the alternative pathway results primarily in mono-phosphorylation of the activation loop residue Thr-180. Recombinant mono-phosphorylated and dual phosphorylated p38α differed widely with regard to activity and substrate preference. Altered substrate specificity was reproduced in T cells in which p38 was activated by the alternative or classical MAPK pathways. These findings suggest that T cells have evolved a mechanism to utilize p38 in a specialized manner independent of and distinct from the classical p38 MAPK signaling cascade.

p38 MAPKs are typically activated by upstream MAPK kinases that phosphorylate a Thr-X-Tyr motif in the activation loop. An exception is the T cell antigen receptor signaling pathway, which bypasses the MAPK cascade and activates p38␣ and p38␤ by phosphorylation of Tyr-323 and subsequent autophosphorylation of the activation loop. Here we show that, unlike the classic MAPK cascade, the alternative pathway results primarily in mono-phosphorylation of the activation loop residue Thr-180. Recombinant mono-phosphorylated and dual phosphorylated p38␣ differed widely with regard to activity and substrate preference. Altered substrate specificity was reproduced in T cells in which p38 was activated by the alternative or classical MAPK pathways. These findings suggest that T cells have evolved a mechanism to utilize p38 in a specialized manner independent of and distinct from the classical p38 MAPK signaling cascade.
MAPK 2 are expressed in all eukaryotic cells and participate in responses to stimuli involved in cell activation, proliferation, differentiation, and death (1,2). Of the three major MAPK families, the ERKs play a part in mitogenic signal propagation, whereas the JNKs (c-Jun N-terminal kinases) and p38 isoforms participate in responses to pro-inflammatory cytokines and stress. MAPK activity is closely coupled to changes in gene expression, consistent with the fact that their targets include numerous transcription factors and kinases that themselves regulate transcription factors. Aberrant p38 activity has been implicated in inflammation and cancer, for which p38 inhibitors are being explored as therapeutic agents (3,4).
MAPK are activated by a three-tiered kinase signaling cascade. At the most membrane-proximal level are the MAPK kinase kinases (MAPKKK), serine/threonine kinases that phosphorylate and activate the second level, the MAPK kinases (MAPKK). MAPKK are dual specificity kinases that phosphorylate a MAPK Thr-X-Tyr motif (X being glycine for p38 family members) located on the flexible activation loop that borders the catalytic site and regulates access to substrate. The p38 MAPK family has four members as follows: ␣, ␤, ␥, and ␦. p38␥ expression is limited primarily to skeletal muscle (5), whereas the others are more widely distributed. p38␣ is the major isoform in T cells, which also express lesser amounts of p38 ␤ and ␦ (6). p38␣ and -␤ are the most closely related by amino acid sequence (74% homology), are subject to inhibition by the widely used inhibitor SB203580, and share a tripeptide motif, Pro-Tyr-Asp, involved in the T cell antigen receptor (TCR)initiated alternative activation pathway (7).
TCRs recognize peptide antigen presented by major histocompatibility molecules on the surface of antigen-presenting cells. Although it had generally been assumed that TCR-induced p38 activation utilized the canonical MAPK cascade (8), we found that this is not the case (7). Rather, stimulation via the TCR results in Lck-dependent activation of ZAP70, which in turn phosphorylates p38␣ and -␤ on Tyr-323. Phosphorylation of Tyr-323 leads to p38␣ and p38␤ autophosphorylation on the activation loop and increased activity toward third party substrates. Recent evidence suggests that the scaffold protein Dlgh1 bridges Lck/ZAP70 and p38, and in its absence Tyr-323 is not phosphorylated, and p38 is not activated in response to TCR signaling (9). Interestingly, despite the fact that B cells express the ZAP70 relative Syk, the alternative pathway does not appear to exist in antigen receptor-stimulated B cells (7,10). The importance of the alternative pathway has been demonstrated in mice lacking Gadd45␣, in which the alternative pathway is constitutively active, resulting in spontaneous T cell p38 activity and autoimmunity (11). More recently, we have found that p38␣ cannot be activated by TCR signaling in primary T cells from gene-targeted mice in which Tyr-323 has been replaced with Phe (10). An intriguing question is why T cells acquired a seemingly unique mechanism for p38 activation rather than utilizing the common signaling cascade. Here we find that the two activation pathways are qualitatively different, because unlike MAPKKs, p38␣ predominantly autophosphorylates a single residue in the activation loop, Thr-180, resulting in altered substrate selectivity.

EXPERIMENTAL PROCEDURES
Cells-Primary T cells were purified from lymph nodes and spleens of 6 -8-week-old C57/BL6 mice using negatively selecting T cell purification columns (CL101) from CedarLane. T cells and the Jurkat T lymphoma (ATCC) were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (Invitrogen), 2 mM glutamine, 50 M ␤-mercaptoethanol, and 100 M gentamicin.
In Vitro Kinase Assays-Substrate proteins (3-10 g) were prewarmed at 30°C in 30 l of kinase buffer (20 mM Tris, pH 7.5, 10 mM MgCl 2 , 2 mM dithiothreitol, 5 mM ␤-glycerophosphate, 5 mM NaF, and 0.2 mM Na 3 VO 4 ) with 1-5 Ci [ 32 P]ATP and the indicated amounts of ATP. Reactions were initiated by addition of 100 ng of semisynthesized p38␣ or MKK6-activated p38, as indicated. At the indicated times aliquots were removed and mixed with sample buffer. Phosphorylated products were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and visualized with a Storm PhosphorImager (GE Healthcare).
Protein Semisynthesis-To permit native chemical ligation, the natural histidine of the N-terminal residue of the synthetic peptide (aa 310 -360) was mutated to cysteine. This mutation had no effect on the spontaneous activity of bacterially expressed full-length p38 (data not shown). Residues 1-309 of mouse p38␣ were cloned into the vector pTYB1 (New England Biolabs) using the NdeI and SapI restriction sites. To overcome insolubility, the resulting fusion protein of p38␣ (aa 1-309) with the VMA1 intein and the chitin-binding domain was grown in BL21(DE3) cells harboring the plasmid pREP4-GROEL/S, which encodes the chaperone complex GROEL/S. Cultures grown in LB with 100 g/ml ampicillin and 25 g/ml kanamycin to an A 600 of 0.6 were induced with 0.1 mM isopropyl ␤-D-thiogalactopyranoside for 15 h at 12°C, and proteins were extracted by sonication in binding buffer (PBS, 0.5 M NaCl, 0.1% Triton X-100). Fusion proteins were immobilized on chitin beads, and p38␣ (aa 1-309) having a C-terminal thioester was released by overnight treatment at room temperature with 0.2 M mercaptoethanesulfonate. After elution with mercaptoethanesulfonate/binding buffer and concentrating to ϳ10 mg/ml with YM-30 columns, p38␣ (aa 1-309) was ligated to a 5-fold excess of the N-terminal cysteine of the C-terminal (aa 310 -360) peptides with Tyr-323 either unphosphorylated or fully phosphorylated. The ligation reaction mixtures were washed with YM-30 columns into PBS before use in in vitro kinase reactions.
Cell Stimulation-Following incubation overnight in complete medium supplemented with HL-1 (Lonza), T cells were stimulated through the TCR by transferring to wells coated with 5 g/ml anti-CD3 in PBS.

RESULTS
Phosphorylation of Tyr-323 Is Sufficient to Activate p38␣-Upstream MAPKKs activate p38␣ by phosphorylating the activation loop residues Thr-180 and Tyr-182 (16). Bacterially expressed p38␣ becomes activated when phosphorylated in vitro on Tyr-323 (7). However, p38␣ acquires a variable degree of activation loop phosphorylation during growth in bacteria (7) 3 due to autophosphorylation, because a kinase-inactive (K53M substitution) mutant of p38␣ lacks any detectable activation loop phosphorylation or kinase activity (data not shown). To determine whether Tyr-323 phosphorylation in the absence of prior Thr-180/Tyr-182 phosphorylation is sufficient to up-regulate kinase activity, we prepared synthetic p38␣ phosphorylated on only one site, Tyr-323 (Tyr(P)-323), using protein semisynthesis (17). Recombinant p38␣ residues 1-309 were ligated to synthesized C-terminal (310 -360) fragments, creating full-length p38␣ with Tyr-323 either phosphorylated or unphosphorylated. The effect of Tyr-323 phosphorylation on p38␣ activity was assessed with an in vitro kinase assay using the physiological substrate ATF2. Whereas semisynthesized unphosphorylated p38␣ Tyr-323 was inactive, the introduction of a phosphorylated Tyr-323 resulted in an active kinase that phosphorylated ATF2 (Fig. 1, left). Importantly, both semisynthesized molecules were successfully refolded and capable of kinase activity, because after incubation with MKK6, which phosphorylates Thr-180/Tyr-182 in the activation loop, they phosphorylated ATF2 equally well (Fig. 1, right). Thus, Tyr-323 phosphorylation in the absence of activation loop phosphorylation is sufficient to activate p38.
Characterization of Antibodies Recognizing Individually Phosphorylated p38␣ Activation Loop Residues-The site of p38␣ autophosphorylation by the TCR-mediated alternative pathway was previously mapped to the activation loop using an antiserum thought to recognize dual phosphorylated p38 (7). However, the report that a similar antibody against the phosphorylated activation loop could also recognize mono-phosphorylated Thr-180 (18) prompted us to revisit the question of what residues were involved in p38␣ autophosphorylation. We developed a rabbit polyclonal antibody against a peptide containing the mouse p38␣ activation loop in which only Tyr-182 was phosphorylated (anti-Tyr(P)-182). A dot blot of peptides comprising p38␣ residues 176 -186 with Thr-180 and Tyr-182 unphosphorylated, phosphorylated singly (Thr(P) or Tyr(P)), or phosphorylated together (Thr(P)/Tyr(P)) was immunoblotted with the putative dual phosphorylation-specific antibody previously used to evaluate p38 phosphorylation in T cells (9211) (7) or anti-P-Tyr-182. 9211 recognized Thr(P)-180 with or without concomitant Tyr(P)-182 ( Fig. 2A) but had little activity toward Tyr(P)-182. Anti-Tyr(P)-182 recognized Tyr(P)-182 but not Thr(P)-180. The specificity of these antibodies for the phosphorylated native protein was determined by immunoblotting kinase-inactive p38␣ that contained mutations of one or both phosphoacceptor sites and that had been phosphorylated by MKK6. Once again, 9211 recognized primarily p38␣ mono-phosphorylated on Thr-180 (Fig. 2B). Therefore, the 9211 antibody is hereafter referred to as anti-Thr(P)-180.
p38␣ Autophosphorylation Occurs on Thr-180-TCR-induced p38␣ autophosphorylation follows phosphorylation of the nonactivation loop residue Tyr-323 (7). We observed that p38␣ expressed in Escherichia coli grown at 37°C is phosphorylated on both Tyr-323 and the activation loop because of autophosphorylation, because these residues are not phosphorylated on the kinase-inactive p38␣ mutant (data not shown). Mutation of Tyr-323 to Phe also abolished activation loop autophosphorylation and activation of p38␣ in bacteria (data not shown), indicating that autophosphorylation requires phosphorylation of Tyr-323. This suggested that E. coli-expressed p38␣ can serve as a model for p38␣ activated by the alternative pathway. We asked if active p38␣ can phosphorylate kinaseinactive p38␣ and if so on which site(s). GST-tagged kinaseinactive p38␣ was offered as a substrate for active p38␣ in an in vitro kinase (IVK) assay. After incubation with active p38␣, GST-p38␣ was recognized by anti-Thr(P)-180 but not anti-Tyr(P)-18 (Fig. 3). Thus, p38␣ can phosphorylate itself in trans, which is perhaps the mechanism of its autophosphorylation in TCR-activated T cells, on the activation loop residue Thr-180 but not Tyr-182.
To determine whether the recombinant p38␣ reflects the behavior of the endogenous protein when activated in cells, we examined the phosphorylation status of p38 in TCR-stimulated T cells. Anti-Thr(P)-180 detected Thr-180 phosphorylation in cells stimulated through the TCR (the alternative pathway) or with PMA (which initiates the MAPK cascade) (Fig. 4). In contrast, anti-Tyr(P)-182 gave a strong signal with p38␣ from T cells stimulated with PMA but only a faint signal with TCR-   Kinase-dead GST-p38 (Lys-53-Met) was left unphosphorylated or phosphorylated by active His-p38 in an IVK reaction. Reactions were immunoblotted (IB) with the indicated anti-phospho-p38␣ antibodies (upper panels) and with antibodies recognizing total p38 (lower panels). stimulated cells. Therefore, unlike the MAPK cascade, TCRtriggered alternative p38 activation causes primarily monophosphorylation of Thr-180.
Mono-phosphorylated and Dual Phosphorylated p38␣ Have Different Substrate Specificities-Dual phosphorylated p38 is the only form of the activated enzyme that has thus far been implicated in cellular events. That TCR signaling causes p38␣ mono-phosphorylation raises the possibility that the enzyme functions differently in cells stimulated via the TCR compared with stimuli that set off the MAPK cascade. As a first approach, we asked if dual and mono-phosphorylated p38␣ differ in their ability to phosphorylate known p38␣ substrates. Dual and mono-phosphorylated p38␣ were generated by phosphorylating recombinant wild type and Y182F p38␣ with MKK6. To exclude potential confounding effects of Tyr-323 phosphorylation, Tyr-323 in both proteins was replaced with Phe. MKK6 phosphorylated both proteins on Thr-180 to yield mono-and dual phosphorylated p38␣ (Fig. 5A) that were then tested in an IVK assay with the substrate GST-ATF2-(1-109) (19). ATF2 contains two p38 target residues, Thr-69 and Thr-71 (20). Thr-69 and Thr-71 are equivalent targets for p38␣, but initial phosphorylation of Thr-71 favors the subsequent phosphorylation of Thr-69 to yield the doubly phosphorylated substrate (21). Although the mono-phosphorylated p38␣ was somewhat less efficient that the dual phosphorylated form, both generated phospho-ATF2 that resolved as two bands on SDS-PAGE (Fig.  5B, top row). Interestingly, whereas dual phosphorylated p38␣ generated predominantly the slower migrating ATF2 band over time, mono-phosphorylated p38␣ reproducibly yielded similar amounts of the slower and faster migrating species. Quantitation of incorporated 32 P in the experiment shown, for example, revealed that the upper to lower band ratio was ϳ1:1 at all time points using mono-phosphorylated p38␣ but increased from 1:1 at 3 min to 5:1 at 30 min using dual phosphorylated p38␣. To identify the different ATF2 species, we mutated the known sites of phosphorylation and determined how these proteins migrated when phosphorylated by dual phosphorylated p38␣ (Fig. 5C). The upper band was not found in any of the mutants, demonstrating that it represents dual phosphorylation of Thr-69 and Thr-71. Phosphorylation of Thr-69 alone (Fig. 5C,  lane 3) yielded the faster migrating species, and phosphorylation of Thr-71 alone (Fig. 5C, lane 2) resulted in a band with intermediate migration. Thus, three phosphorylated ATF2 species can be resolved; the lower band is phospho-Thr-69, and the upper is doubly phosphorylated ATF2, and the intermediate band (a transient species that cannot be distinguished in phosphorylated wild type ATF2, Fig. 5C, lane 1) is phospho-Thr-71. We therefore conclude that whereas both forms of p38␣ can phosphorylate ATF2, the mono-and dual phosphorylated p38␣ differ in their fine specificity, the latter being better able to convert mono-phosphorylated ATF2 to the doubly phosphorylated and slower migrating species.
We compared the activity of mono-and dual phosphorylated p38␣ on other p38␣ substrates (Fig. 5B). MAPKAP kinase 2 (2,22) and the transcription factor MEF2A (23,24) were each phosphorylated by mono-phosphorylated p38 but much less well than by the dual phosphorylated form. An even greater dichotomy was observed with the substrate STAT4 (Fig. 5B) (25,26). Bcl-2 and Bcl-X L were also examined, because it has been suggested that their phosphorylation by p38 can disable their anti-apoptotic function (14,27). Bcl-X L was a very poor substrate for mono-phosphorylated p38␣, and Bcl-2 was not phosphorylated at all. Thus, the activity of mono-phosphoryla-tedcomparedwithdualphosphorylatedp38␣wasverysubstratedependent, being almost equivalent for ATF2 (although with a difference in fine specificity) and being virtually nil for STAT4 and Bcl-2.  Alternatively Activated p38␣ Has Altered Substrate Specificity in Vivo-The question of whether the altered substrate specificity of recombinant mono-phosphorylated p38 is reflected in vivo was addressed by examining the phosphorylation of several substrates in TCR-activated cells. Lysates of primary mouse T cells activated with either anti-CD3 (the alternative pathway) or with PMA (the MAPK cascade) were immunoblotted with the antibodies that recognize the p38 phosphorylation sites on ATF2 and STAT4. Phosphorylation of p38␣ Thr-180 in response to both stimuli was similar (Fig. 6, 3rd row), but only PMA caused Tyr-182 phosphorylation (Fig. 6, 4th row). Consistent with the in vitro results, TCR and PMA stimulation induced similar phosphorylation of ATF2. Note that the antibody used to detect phospho-ATF2 recognizes Thr-71, and therefore one cannot detect Thr-69 (mono-phosphorylated) ATF2. STAT4, one of the substrates least favored by monophosphorylated p38 in vitro, was phosphorylated well after PMA but not TCR-mediated activation. Therefore, as for recombinant p38␣, the substrate specificity of endogenous TCR-activated (mono-phosphorylated) p38 differs from MAPK cascade-activated (dual phosphorylated) p38.

DISCUSSION
MAPKs are unusual among kinases in having two phosphoacceptor sites on their activation loops, and the role of the second (tyrosine) site is not well understood. The existence of phosphatases specific for one or the other of the sites has prompted the hypothesis that mono-phosphorylated species exist, but biological examples have not been described, and Thr-180 mono-phosphorylated p38␣ has been observed only in nonphysiological contexts (28,29). Constitutively active point mutants of the p38␣ homolog Hog1 were isolated in a genetic screen in yeast (30). When introduced into p38␣, some of these mutations induced kinase activity and autophosphorylation in trans on Thr-180 (18). Notably, these constitutively active mutants bearing an additional Y182F substitution retained activity, albeit at a reduced level (31,32). The majority of the Hog1-activating mutations were located along the C-terminal extension known as L16, a flexible loop that wraps around the N-terminal domain of all MAPK and which contains Tyr-323 (30). The many similar features between the Hog1-activating mutations (18,32) and Tyr(P)-323-activated p38␣ reported here strongly suggest that they activate kinase activity by a common mechanism. Such a mechanism could involve disruption of a hydrophobic core made up of the aromatic side chains of Tyr-69, Phe-327, and Trp-337 (31). The side chain of unphosphorylated Tyr-323 is oriented inward toward this core, and addition of a negative charge could perturb the pocket, twist L16, and, via an interaction between L16 and the activation loop analogous to that seen in dual phosphorylated ERK2, induce a rotation between the C-and N-terminal domains that remodels the catalytic site of p38 (18,(32)(33)(34).
The individual contributions of Thr-180 and Tyr-182 phosphorylation to p38 function have recently been addressed (29,35). Using ATF2 as a substrate, it was found that Thr-180 mono-phosphorylated p38␣ was an order of magnitude less active than dual phosphorylated p38␣ in vitro. We found that the difference between mono-and dual phosphorylated p38␣ became smaller as the concentration of ATP was increased. 3 mM ATP was used in this study because it approximates physiological intracellular levels (36), and under these conditions ATF2 was phosphorylated almost as well by mono-phosphorylated p38␣ as by the dual phosphorylated species. Deuterium exchange mass spectroscopy revealed that dual phosphorylation induces significant alterations in solvent accessibility to the activation loop, active site, and substrate docking domains, whereas Tyr-182 mono-phosphorylation has little effect (37). Therefore, and consistent with our data, it seems likely that Thr-180 phosphorylation has the major role in formation of the active site, and Tyr-182 phosphorylation enhances the activity and plays a role in determining specificity.
An important question is why have T cells acquired an alternative p38␣ and -␤ activation pathway that is not shared even by closely related B cells. The finding in this report that activation of p38␣ by TCR stimulation is qualitatively different raises several possibilities. T cells are equipped to participate in inflammatory responses and receive stimulation through receptors such as interleukin-12 and -18, which activate p38 by the classical MAPK cascade and thus increase the production of downstream pro-inflammatory molecules such as interferon-␥ (IFN-␥) (13, 38 -41). The responses a T cell makes to antigen via the TCR, such as clonal expansion and providing help to B cells, can occur independently of inflammatory conditions, and induction of pro-inflammatory downstream events might be deleterious. Therefore, the alternative pathway may be a means FIGURE 6. In vivo TCR-activated p38␣ displays altered substrate specificity. T cells were stimulated as in Fig. 4, and lysates were immunoblotted with antibodies recognizing the indicated phosphorylated proteins. of allowing p38 activation and some limited gene up-regulation without inflammatory sequelae. Another possibility is that MAPK cascade-activated p38␣ is itself harmful to T cell development and/or cell function. Phosphorylation of Bcl-2 by p38 has been shown to induce its release from mitochondria, thereby disrupting its ability to inhibit apoptosis (27,42). Mice expressing a constitutively active form of MKK6 in T cells, and therefore having constitutively active dual phosphorylated p38, have a reduction in the number of CD8 ϩ T cells because of their apoptosis, and the cells that survive proliferate less well in response to mitogen (43). It was suggested that the spontaneous apoptosis of CD8 ϩ cells was because of p38-mediated phosphorylation of the anti-apoptotic protein Bcl-2, a mechanism that has also been implicated in Fas-induced CD8 ϩ T cell death (14). Our finding that Bcl-2 is a very poor substrate for mono-phosphorylated, in contrast to dual phosphorylated, p38␣ is consistent with the notion that activation of p38 by the alternative pathway helps to maintain cell viability in TCR-signaled cells. IFN-␥ is a target for TCR and cytokine signaling pathways, either separately or in combination, in T cells. ATF2, which we found to be a reasonably good substrate for mono-phosphorylated p38␣, is implicated in TCR-mediated induction of IFN-␥ (44), whereas STAT4, which is not a target of monophosphorylated p38␣, is important for interleukin-12-induced (25) but not TCR-induced (26) IFN-␥ (via the MAPK cascade). Thus, the evolutionarily conserved loss of TCR-coupled MAPK cascade initiation and the acquisition of the alternative p38 activation pathway may be driven by the need of T cells to tightly control p38 activation and limit its substrates to those useful in immune responses.