Crystal Structure of the P38α-MAPKAP Kinase 2 Heterodimer*

The p38 signaling pathway is activated in response to cell stress and induces production of proinflammatory cytokines. P38α is phosphorylated and activated in response to cell stress by MKK3 and MKK6 and in turn phosphorylates a number of substrates, including MAPKAP kinase 2 (MK2). We have determined the crystal structure of the unphosphorylated p38α-MK2 heterodimer. The C-terminal regulatory domain of MK2 binds in the docking groove of p38α, and the ATP-binding sites of both kinases are at the heterodimer interface. The conformation suggests an extra mechanism in addition to the regulation of the p38α and MK2 phosphorylation states that prevents phosphorylation of substrates in the absence of cell stress. Addition of constitutively active MKK6-DD results in rapid phosphorylation of the p38α-MK2 heterodimer.

The p38 signaling pathway is activated in response to cell stress and induces production of proinflammatory cytokines. P38␣ is phosphorylated and activated in response to cell stress by MKK3 and MKK6 and in turn phosphorylates a number of substrates, including MAPKAP kinase 2 (MK2). We have determined the crystal structure of the unphosphorylated p38␣-MK2 heterodimer. The C-terminal regulatory domain of MK2 binds in the docking groove of p38␣, and the ATP-binding sites of both kinases are at the heterodimer interface. The conformation suggests an extra mechanism in addition to the regulation of the p38␣ and MK2 phosphorylation states that prevents phosphorylation of substrates in the absence of cell stress. Addition of constitutively active MKK6-DD results in rapid phosphorylation of the p38␣-MK2 heterodimer.
The p38 signal transduction pathway regulates cellular inflammatory responses. The pathway is activated by proinflammatory cytokines, tumor necrosis factor ␣, bacterial lipopolysaccharides, or osmotic stress (1) and results in increased levels of tumor necrosis factor ␣, interleukin 1␤, interleukin 6, interleukin 10 interferon-␥, and COX-2 (2,3). Inhibitors of this pathway have been demonstrated to reduce inflammation (4). The p38 kinase family has four members: P38␣, ␤, ␥, and ␦ (5). P38␣ and ␤ are the main transponders of the inflammatory response (6) and have been targets for anti-inflammatory therapy. Several p38␣ and ␤ inhibitors are in clinical trials for the treatment of chronic obstructive pulmonary disease, psoriasis, and rheumatoid arthritis (7). Activated p38␣ phosphorylates transcription factors such as ATF2, Elk-1, and MEF2A and downstream kinases such as MAPKAP kinase 2 (MK2), 2 MK3, PRAK, MNK1/2, and MSK1 (8,9). Several p38␣ substrates are phosphorylated in the nucleus, others in the cytoplasm. MK2 also phosphorylates proteins found in both the nucleus (cAMP-response element-binding protein, or CREB) (10) and cytoplasm (HSP25/27 and LSP-1) (11,12). Two splice variants of MK2 have been identified. MK2a contains the nuclear localization signal (NLS) and the p38 docking domain, whereas MK2b is a truncated variant of MK2 that lacks the NLS and p38 docking domain (13,14). MK2b is still phosphorylated by p38␣ but the signal transduction is less efficient (15). P38␣ forms a complex with MK2a even when the signaling pathway is not activated (16). This heterodimer is found mainly in the nucleus. Upon activation of the p38␣ signaling cascade, p38␣ and MK2a phosphorylate their nuclear substrates before translocating to the cytoplasm. P38␣ does not have a nuclear export signal (NES) and cannot leave the nucleus by itself but rather needs to be associated with MK2a. The NES of MK2a facilitates the transfer of both kinases from the nucleus to the cytoplasm but only after MK2 has been phosphorylated by p38␣ (17,18). P38␣ phosphorylates MK2 at four sites, Thr 25 , Thr 222 , Ser 272 , and Thr 334 . The phosphorylation of Thr 334 has been demonstrated to be critical for the nuclear export of the p38␣-MK2 complex, whereas the kinase activity of MK2 is not (18). Phosphorylation of Thr 334 is believed to induce a conformational change in the complex prior to interaction with the leptomycin B-sensitive nuclear export receptor (19).
In this report, we present a structural analysis of the p38␣-MK2 heterodimer. Crystal structures were determined for both the full-length p38␣-MK2 heterodimer and the p38␣-MK2 peptide (residues 370 -400) complex. The structures reveal that the heterodimer conformation buries the substrate binding grooves of both p38␣ and MK2a, thus preventing both kinases from phosphorylating their respective substrates. The structure suggests that substrate phosphorylation is dependent on the phosphorylation states of both kinases and the heterodimer conformation.

Protein Purification and Generation of Protein-Protein
Complex-Human p38␣ was expressed and purified as described (20). Full-length MK2 was expressed and purified according to the protocol used for N-terminal-truncated MK2 (residues 46 -400) (21). P38␣-MK2 heterodimer samples were prepared by incubating the purified proteins in equimolar concentrations (25 M each) for 1 h on ice. The p38␣-MK2 complex was separated from the monomers by Q-Sepharose anion exchange chromatography using a linear NaCl gradient (100 -500 mM), and the fractions containing the p38␣-MK2 heterodimer were pooled. Finally, the buffer was exchanged into 20 mM Tris-Cl, pH 7.8, 50 mM NaCl, and 2 mM dithiothreitol, and the protein sample was concentrated to 20 mg/ml for crystallization set-ups. * These studies were conducted within the protein kinase collaboration between Vertex Pharmaceuticals and Novartis Pharma AG. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Crystallization and Data Collection-Crystals were grown with the hanging drop method over a reservoir containing 3.75% polyethylene glycol 2000, 100 mM Hepes, pH 7.5. Prior to the crystallization set up, 1-s-nonyl-1-␤-D-thioglucoside (1ϫ critical micelle concentration) and heptanetriol (1.5%) were added to the protein sample. The crystallization drop was prepared by mixing the reservoir solution with the protein solution in a 1:1 ratio. Crystals grew overnight and were typically harvested after 72 h. The crystals were transferred for 1 min to a solution consisting of reservoir solution and 30% glycerol before being flash frozen in liquid nitrogen. Data collection at The data sets were indexed and scaled with D*trek. The structure was determined by molecular replacement (MOLREP) (22) using a peptide-bound monomer of p38␣ (1LEZ) (23) as the starting model. The asymmetric unit contained two p38␣ peptide complexes. The crystal structure was refined with multiple rounds of rebuilding and refinement using the refinement program (Auto)BUSTER (24). The final model contained p38␣ residues 4 -173 and 184 -353 and MK2a peptide residues 370 -393. The R and R free of the p38␣-MK2 peptide complex were 0.23 and 0.26, respectively. p38␣-MK2 peptide complex and the MK2 monomer (1KWP) were subsequently used to solve the structure of the full-length p38␣-MK2 heterodimer. This structure was refined using only the rigid body protocol of BUSTER, and rebuilding was limited to cleaning up the Ramachandran plot and building of the MK2 activation loop.
Phosphorylation of the p38␣-MK2 Complex and Western Blot Analysis-Phosphorylation of the p38␣-MK2 complex in the absence or presence of the p38␣ inhibitor SB203580 was performed by mixing the p38␣-MK2 complex and MKK6-DD in a 100:1 molar ratio. The phosphorylation was carried out in 20 mM Tris-Cl, pH 7.8, 50 mM NaCl, 10 mM MgCl 2 , and 2 mM dithiothreitol. The complex was preincubated with inhibitor for 30 min on ice prior to phosphorylation. The reaction was started by addition of 2.5 mM ATP, and aliquots were taken at 0-, 5-, 10-, 20-, 30-, and 60-min time intervals and quenched with SDS sample buffer. Western blots were performed using phospho-p38 MAP kinase and phospho-MK2 primary antibodies (1:5000 dilution) and Alexa Fluor 680-conjugated secondary antibody (Invitrogen) (1:10,000 dilution). Blots were scanned with an Odyssey Infrared Imaging System at 700-nm wavelength, and the integrated intensities were analyzed to determine the time course of phosphorylation.

RESULTS
General Description of the Structure-P38␣ and MK2 are both serine/threonine protein kinases. The structures of both P38␣ and MK2 have been determined previously as monomers (20,21,25). Both structures have typical kinase folds with a small N-terminal domain dominated by ␤-strands, a larger C-terminal domain consisting of ␣-helices, and an ATP-binding site wedged between the two domains. The ATP-binding site is formed by the flexible glycine-rich loop, the hinge that connects the small N-terminal domain with the large ␣-helical domain, and the catalytic loop. The kinase domains of p38␣ and MK2 also have a substrate binding groove. This groove is at the interface of the N-and C-terminal domains and binds the part of the substrate protein that contains the phosphorylation site. Both p38␣ and MK2 have additional unique features. P38␣ has an extra docking groove for activating kinases, substrate kinases, and inactivating phosphatases. The docking groove contains the previously identified glutamate-aspartate (ED) and common docking (CD) regions and is conserved among other MAP kinases such as ERK1/2 and JNK, but not in other kinases.
The kinase domain of MK2a is preceded by a proline-rich sequence (residues 1-44) that is not present in the published crystal structures (26). Beyond the kinase domain it has a C-terminal regulatory domain (residues 338 -400) that consists of helices ␣J and ␣K and a long loop. The C-terminal regulatory domain contains an NES and the bipartite NLS (17,18). The ␣K-helix occupied the MK2a substrate binding groove in an autoinhibitory conformation in one of the crystal structures of the monomer (21).
We have determined the crystal structure of a p38␣-MK2 heterodimer that contains full-length p38␣ and MK2a, and we have also determined a complementary structure of p38a complexed with only the docking peptide (residues 370 -400) of MK2. The p38␣ and MK2a molecules that make up the heterodimer are positioned "face to face" so that the ATP-binding sites of both kinases are at the heterodimer interface (Fig. 1). The C-terminal segment of MK2 (residues 368 -400), which follows the ␣K-helix, wraps around p38␣ and inserts into the docking groove. Residues 382-390, including the C-terminal part of the NLS, fold into an ␣-helix (␣L), which occupies the CD region of the p38␣ docking groove (Fig. 1). The ␣K-helix (residues 344 -367) of the C-terminal regulatory domain is sandwiched between the p38␣ and MK2a kinase domains. It occupies the MK2a substrate binding groove, as it does in the crystals of the MK2a kinase domain alone (21). The C-terminal end of the ␣K-helix (Fig. 1, A and C) that contains the NES of MK2a (residues 356 -365) is buried between the two kinase domains in close proximity to the p38␣ ATP-binding site. The MK2a activation loop (residues 207-233) also adapts to binding with p38␣. In the monomer crystal forms, the activation loop is disordered (not shown), whereas in the p38␣-MK2 heterodimer the complete MK2a activation loop backbone can be traced through the electron density map, as it is stabilized by interactions with the ␣-helical domain of p38␣ (supplemental Fig. S1). The phosphorylation site in the MK2a activation loop (Thr 222 ) does not make contact with p38␣.
The C-terminal regulatory domain and the activation loop of MK2a form most of the contacts between p38␣ and MK2a in the heterodimer (Fig. 2). The C-terminal regulatory domain contacts both the N-terminal ␤-strand and the ␣-helical domains of p38␣. The C-terminal end of the ␣K-helix (residue Glu 354 ) touches the tip of the glycine-rich loop (Ala 34 ) and the ␤1L0-␤2L0 loop, but only Glu 354 (MK2a) and Ala 34 (p38␣) are less than 4 Å apart. The C-terminal loop of the C-terminal regulatory domain occupies the docking groove of p38␣ and is responsible for the high affinity interaction between p38a and MK2a (16). This interaction will be discussed in more detail below. The activation loop of MK2a makes contacts with p38␣ in the heterodimer and a neighboring MK2a molecule. The conformation of the C-terminal portion of the activation loop (residues 232-237) is stabilized by the loop between the ␣Dand ␣E-helices in p38␣, but only the backbone density is visible in the electrodensity map. There is an additional contact between the MK2a loop of ␣H and ␣I and the p38␣ loop between ␣F and ␣G. The solvent-accessible surface area buried in the heterodimer is 1373 Å 2 for MK2a and 1477 Å 2 for p38␣ (Fig. 2).
The p38␣ activation loop contains two phosphorylation sites (Thr 180 and Tyr 182 ) that when modified stabilize its activated conformation and open the substrate binding groove. In the heterodimer described here, p38␣ is not phosphorylated and the p38␣ activation loop is largely disordered. Residues 174 -184 were omitted from the model due to lack of electron density. This disordered region was in close proximity to a disordered loop region from MK2a (residues 265-283). The MK2a residues 264 -281 are included in a published, mercury-derivatized crystal structure (1KWP) (21), but their positions are not compatible with the p38␣-MK2 heterodimer conformation.
The P38␣ Docking Groove-P38␣ and the other MAP kinases have a docking groove for binding with their activating kinases, inactivating phosphatases, and substrates (23,(27)(28)(29)(30)(31)(32)(33)(34)(35)(36). The docking groove is not to be confused with the substrate binding groove at which the phosphorylation reaction takes place. The groove was first identified by mutational studies as two separate regions (Fig. 3), the CD region and the ED region (27). The CD region is part of a shallow groove formed by the acidic residues Asp 313 , Asp 316 , Glu 81 , and the aromatic residues Phe 129 and Tyr 311 . The latter is part of a deeper groove formed by residues 159 -163 (159-VNEDCE-163) at one side and residues Gln 120 , His 126 , and Phe 129 at the opposite side.
The C-terminal end of the regulatory domain of MK2a (residues 369 -384) binds in the docking groove of p38␣ and makes key interactions with the ED and CD regions (Fig. 4). These interactions are a combination of hydrophobic and polar contacts. Earlier studies identified the MAP kinase binding motif as rich in positively charged residues, and because the ED and CD regions of p38␣ are mainly negatively charged, it was assumed that these motifs contact each other. Our structure shows that this is indeed the case.  Contacts between p38␣ and MK2 in the heterodimer. The distances between p38␣ (yellow) and MK2 (purple) residues were calculated. Those residues that were less than 4 Å separated were colored either red, blue, green, or orange. Residues 5-99 of p38␣ and the MK2 binding partners were colored red. Residues 100 -199 of p38␣ and the MK2 contacts were colored blue. Residues 200 -299 of p38␣ and MK2 contacts were colored green. Residues 300 -352 of p38␣ and MK2 contacts were colored orange.
The docking groove interactions are consistent with the binding studies of Lukas et al. (16). Using surface plasmon resonance, stopped flow fluorescence, and isothermal titration calorimetry, they observed that the docking groove interactions contribute most to the formation of the heterodimer. They also noticed that the binding of the MK2a to p38␣ depends on the ionic strength of the buffer, consistent with the many electrostatic interactions between p38␣ and MK2a that we see in the heterodimer.
The sequence of the MAP kinase binding motif is well conserved among proteins that bind p38␣, ERK, or JNK. The main difference is the location of the motif with respect to the catalytic domain. In activating kinases and inactivating phosphatases, the sequences of the MAP kinase binding motif come before the catalytic domains, whereas for the substrate kinases such as MK2a the binding motif comes after the kinase domain. Two crystal structures of p38␣ in complex with short peptides have been determined previously. One peptide is derived from the p38␣ activator MKK3b, and the second peptide is from the transcription activator MEF2c (23). These short peptides also bind in the p38␣ docking groove, but the major difference with the p38␣-MK2 complex is the reversed direction of the peptide backbones (Fig. 4). The peptides employed in these studies have very little sequence homology with each other or with MK2a, but the interactions between the peptide side chains and p38␣ at the ED domain are conserved despite the reversed direction of the peptide backbone. The residue pattern is similar but reversed: hydrophobic, polar, hydrophobic, polar. For instance, at the position of MK2a residue Ile 370 , both peptides also have an isoleucine with the side chains pointing in the same direction. Lys 371 overlaps with an arginine of the MKK3b peptide and both side chains extend into the solvent, away from p38␣. The peptides also have a hydrophobic residue at the position of Ile 372 that is buried by the loop of p38␣ consisting of residues 158 -363, and a polar residue that contacts Asp 161 .
Constitutively Active MKK6-DD Phosphorylates the p38a-MK2 Heterodimer-The ATP-binding sites and substrate binding grooves of p38␣ and MK2a are at the interface of the heterodimer, making them inaccessible to substrates. Thus, the conformation of the heterodimer seen here does not represent a p38␣ phosphorylation complex but rather a conformation for which the catalytic activity of p38␣ and MK2a are each sequestered from potential substrates. It probably represents the unactivated complex and reflects the state of both kinases when the cell does not need their phosphorylation capacity, i.e. when the cell does not receive any stress signals. As soon as stress signals do arrive at the cell, however, we expect the p38␣-MK2 heterodimer to alter such that both kinases are accessible to their substrates.
The p38␣-MK2 heterodimer is a stable, high affinity complex with a K d of 6 nM (16). When the p38 signaling pathway is activated due to cell stress, the activating kinases MKK6 or MKK3 phosphorylate p38␣. If the p38␣-MK2 heterodimer is the preferred conformation for both kinases, then p38␣ and MK2a should still readily be phosphorylated when activated MKK6 is added to the complex. According to what is known about the p38␣ signaling pathway, activated MKK6 will phosphorylate and activate p38␣ and p38␣ will subsequently phosphorylate MK2a.
MKK6-DD is a constitutively activated form of MKK6 in which the two aspartates mimic the phosphorylated residues (Ser 207 and Thr 211 ) (37) and is a very efficient activator of p38␣.  When a small amount of MKK6-DD (1:100 molar ratio) was added to the p38␣-MK2 complex, both p38␣ and MK2a became phosphorylated over time (Fig. 5). Phosphorylation of p38␣ was followed by Western blot analysis using an antibody that recognizes phosphorylated Thr 180 and phosphorylated Tyr 182 . Phosphorylation of MK2a was followed with two commercially available antibodies that detect phosphorylated Thr 222 and phosphorylated Thr 334 . The Western blots showed that when MKK6-DD was added to the p38␣-MK2 heterodimer the level of phosphorylated p38␣ increased steadily during the 60-min time course (Fig. 5A). At the starting point, t ϭ 0, no detectible levels of phosphorylated p38␣ were visible, but within 5 min a clear band of phosphorylated p38␣ became apparent and the levels increased until the end of the time course. The MK2a of the heterodimer also became phosphorylated in at least two sites (Fig. 5B). MK2a has four phosphorylation sites, and phosphorylated MK2a migrates with a slightly lower mobility than unphosphorylated MK2a. At t ϭ 0, a small amount of Thr 334 -phosphorylated MK2a was visible but migrated with higher mobility, suggesting that the protein was not fully phosphorylated. Indeed, Thr 222 was not phosphorylated at t ϭ 0. The levels of phosphorylated Thr 334 in MK2a increased more than 80-fold, whereas the levels of phosphorylated Thr 222 increased Ͼ20-fold during the course of the experiment. In summary, addition of constitutively activated MKK6 leads to phosphorylation at both p38␣ and MK2a in the p38␣-MK2 heterodimer, and we conclude that both kinases can readily become activated when activated MKK6 is present.
The signaling pathway of p38␣ and MK2a has been studied extensively, and it is commonly acknowledged that MKK6 phosphorylates p38␣ and p38␣ phosphorylates MK2a. Because we have shown that adding MKK6-DD to the p38␣-MK2 heterodimer leads to the phosphorylation of p38␣ and MK2a, we wanted to determine whether the phosphorylation was sequential, i.e. whether MKK6-DD phosphorylated p38␣ and p38␣ in turn phosphorylated MK2a. SB203580 is a potent and selective inhibitor of p38␣ that has been used frequently to study the p38␣ signaling pathway (38 -40). When SB203580 was added to the p38␣-MK2 heterodimer 30 min in advance of the addition of MKK6-DD, phosphorylation of MK2a was inhibited (Fig. 5B). Phosphorylation of Thr 334 did not exceed the t ϭ 0 levels in the first 20 min of the experiment, and after 60 min the levels of Thr 334 -phosphorylated MK2a were less than a quarter of that observed in the absence of inhibitor. The amount of phosphorylated Thr 222 was also reduced in the presence of SB203580. On the other hand, the activity of MKK6-DD was not affected by the addition of SB203580, as the levels of phosphorylated p38␣ did not change (data not shown); however, MKK6DD does not readily phosphorylate MK2 (Fig. 5C). Thus, when MKK6-DD is added to the p38␣-MK2 heterodimer in the presence of ATP and Mg 2ϩ , the hierarchy in the signaling cascade is maintained without the help of scaffolding proteins.

DISCUSSION
We have determined the crystal structure of the p38␣-MK2 complex. This is the first structure of two full-length kinases associated together in a complex. Because the resolution of the full-length p38␣-MK2 complex was only 4 Å, we also determined the structure of the p38␣-MK2 peptide complex, and together with the previously reported structures of the MK2a monomer we could visualize the docking groove interactions between p38␣ and MK2a.
We believe that the crystal structure represents the conformation of p38␣ and MK2a as it is present in the nucleus when the cell is devoid of stress. Although MK2a is a phosphorylation substrate of p38␣, the conformation of the heterodimer does not represent the phosphorylation reaction but rather prevents both p38␣ and MK2a from phosphorylating their substrates in the absence of stress signals. Tight regulation of the p38␣ signal transduction pathways is important for the proper functioning of a cell, as activation of this pathway increases the levels of proinflammatory cytokines and induces cell cycle arrest. There is evidence that the conformations of unphosphorylated and phosphorylated heterodimers differ. The K d of the unphosphorylated p38␣-MK2 heterodimer is 6 nM, whereas the K d of the phosphorylated heterodimer is 10 times weaker (16). A conformational change has also been observed in MK2a with fluorescence resonance energy transfer analysis and is triggered by the phosphorylation of Thr 334 (MK2a) (19). The conformational change likely makes the NES accessible, thus allowing the p38␣-MK2a heterodimer to be transported to the cytoplasm. The heterodimer structure presented here is not the translocation complex in which the two kinases migrate from the nucleus to the cytoplasm. The MK2a NES is buried between the two kinase domains of the heterodimer and not accessible to carrier proteins. The NLS is part of the p38␣ binding motif and is anchored in the CD domain of the docking groove and therefore not accessible to facilitate transport from the cytoplasm to the nucleus.
Cell stress activates the p38␣-MK2 signaling pathway, which must have major consequences for the heterodimer. Both The levels of phosphorylated p38␣ were followed via Western blot analysis with an antibody that recognizes pThr-Gly-pTyr (residues 180 -182). B, addition of MKK6-DD leads to the phosphorylation of MK2 in the p38␣-MK2 heterodimer, and this phosphorylation event is inhibited by SB203580. SB203580 was added to the p38␣-MK2 heterodimer 30 min prior to the addition of MKK6-DD and ATP. The amount of phosphorylated MK2 was followed with two antibodies that recognize Thr 222 or Thr 334 . C, MKK6-DD phosphorylates MK2 poorly. MKK6-DD and ATP were added to MK2, and the amount of phosphorylated MK2 was followed with two antibodies that recognize Thr 222 or Thr 334 .
kinases will be phosphorylated, p38␣ first by MKK3 or MKK6 and MK2a by p38␣. Both kinases will in turn phosphorylate transcription factors in the nucleus, move to the cytoplasm, and phosphorylate other substrates. Eventually the kinases will be dephosphorylated by a MAP kinase phosphatase. We have shown here that MK2a binds p38␣ in the docking groove. All other proteins that interact with p38␣ also bind in this groove (27,30,31,33,36). These proteins can be p38␣ substrates such as MEF2c, PRAK, and MK2a, p38␣ activators MKK3 and MKK6, or they can be the MAP kinase phosphatase MKP7 that inactivates p38␣. These substrates compete for the same binding site in p38␣ and consequently determine the stress response and subcellular location of p38␣ (Fig. 4). We have also shown that constitutively active MKK6 can phosphorylate the p38␣ in the heterodimer and that p38␣ in turn phosphorylates MK2a. We propose the following model (Fig. 6) for events that take place when the p38␣ signaling pathway is activated and reaches the nucleus.
In the nucleus of a normal cell, p38␣ and MK2a are in the heterodimeric state described here (Fig. 6A). Activated MKK3/6 displaces MK2a from p38␣ to phosphorylate and activate p38␣ (Fig. 6B). Activated p38␣ now has three options depending on the protein it will bind next. (i) It can bind one of its nuclear substrates such as transcription factor MEF2c (Fig. 6C) and activate gene expression in response to cell stress. The binding reaction is a typical catalytic reaction in which the kinase has much lower affinity for the product than for the substrate. Once phosphorylated, the transcription factor is released and p38␣ can bind a different protein. (ii) p38␣ encounters a MAP kinase phosphatase (Fig. 6E). This will result in dephosphorylation and inactivation of p38␣. The inactivated p38␣ can again form a heterodimer with MK2a and this heterodimer will remain in the nucleus (Fig. 6G). (iii) Activated p38␣ binds and activates MK2a at its phosphorylation sites, including Thr 334 (Fig. 6D). MK2a will undergo a conformational change and expose its nuclear export signal. The activated heterodimer of p38␣ and MK2a will then exit the nucleus and phosphorylate substrates in the cytoplasm (Fig. 6F). Most of the p38␣ and MK2a will eventually return to the nucleus. It is not clear whether they return as a heterodimer or as separate monomers but overexpressed p38␣ is mostly found in the nucleus (17), suggesting that p38␣ can reach the nucleus without the aid of MK2a. The bipartite NLS of MK2a is also the p38␣ docking motif, suggesting that binding by p38␣ prevents the binding of importin-␣ and therefore translocation to the nucleus. As a monomer, the MK2a NLS is completely exposed and importin-␣ could bind MK2a and transport it together with importin-␤ to the nucleus.
The p38␣ signaling pathway is a promising target for antiinflammatory therapy. Inhibition of the p38␣ signaling pathway reduces the levels of proinflammatory cytokines, and p38 inhibitors have shown beneficial effects in the treatment of rheumatoid arthritis, psoriasis, and chronic obstructive pulmonary disease (7). In this study we have provided evidence that p38␣ forms a stable heterodimer with MK2a that is likely the preferred conformation of both kinases in the nucleus of a cell that is not receiving stress signals. The crystal structures suggest aspects of the p38␣ signaling pathway that could be helpful for developing novel anti-inflammatory therapies.