Structural and dynamic insights into the allosteric activation of p38α MAP kinase via specific docking interactions

Abstract

Mitogen-activated protein kinases (MAPKs) bind their cognate activating kinases, inactivating phosphatases, substrates and scaffold proteins through docking interactions. Hematopoietic tyrosine phosphatase (HePTP) can remove phosphate from pTyr of activated p38α and ERK2 to generate partially active, mono-phosphorylated MAPKs (p38α/pT and ERK2/pT). Here, we reported that the enzymatic activity of p38α/pT, but not ERK2/pT, can be significantly increased through its interaction with the kinase interaction motif (KIM) of HePTP. To understand the allosteric activation of p38α/pT, we solved the crystal structure of inactive p38α complexed with a KIM peptide derived from HePTP (pepHePTPm). This study provides for the first time a high resolution structural comparison of the interactions for p38α and ERK2 with the same KIM peptide, and reveals a previously unrecognized interaction mode between MAPK and KIM peptide. Binding of pepHePTPm to the p38α C-lobe induces a distinct conformational change in the kinase hinge region which might influence binding of ATP to the active site. The molecular dynamics (MD) simulations further show that the KIM peptide binding increases the dynamic flexibility of the hinge region, facilitating ATP binding. Our results provide new insights into the regulation of MAPK signaling, highlighting the role of docking-induced allosteric changes in modulating kinase activity.

HePTP selectively activates mono-phosphorylated p38α via KIM docking-induced conformational changes in the kinase hinge region, providing new insights into MAPK regulation.

Introduction

The mitogen-activated protein kinases (MAPKs) are the major convergence points in many signaling pathways, and their activities are modulated by kinases and phosphatases^1–3. The three best-characterized MAPK cascades are the extracellular signal-regulated protein kinase (ERK) pathway, which responds to stimuli that induce proliferation and differentiation, the c-Jun N-terminal protein kinase (JNK) pathway, and the p38 kinase pathway, both of which are activated in response to non-mitogenic signals such as proinflammatory cytokines and environmental stress^4–6. Each cascade contains three sequentially acting protein kinases collectively known as a MAPK module. This module consists of a MAPK kinase kinase (MKKK), which activates a downstream MAPK kinase (MKK) that subsequently activates a particular set of MAPKs by phosphorylating the conserved threonine and tyrosine located within the TXY motif in the activation loop of MAPKs, leading to a more than 1000-fold increase of their catalytic activity. Activated MAPKs phosphorylate and regulate a large array of cellular proteins, including key regulatory enzymes, cytoskeletal proteins, regulators of apoptosis, nuclear receptors, and transcription factors, consistent with their diverse biological functions^7,8.

The activity and specificity of MAPKs is tightly regulated in a cell type-specific and temporal manner by the concerted action of multiple regulatory proteins, including upstream kinases, inactivating phosphatases, and scaffolding proteins⁹. Unlike activation of MAPKs, down-regulation of MAPK activity is performed by a large variety of phosphatases, including the serine/threonine-specific, tyrosine-specific and dual-specificity phosphatases^10,11. The hematopoietic tyrosine phosphatase (HePTP) is a classical non-receptor protein tyrosine phosphatase that is predominantly expressed in leukocytes, where it targets MAPKs and regulates T-cell-antigen-receptor signaling^12–15. HePTP can dephosphorylate the activation loop tyrosine of both ERK1/2 and p38α and also helps to sequester the kinases in their resting state in the cytosol¹⁶. It has been shown previously that in vitro, HePTP was able to efficiently remove phosphate from the phospho-tyrosine of bisphosphorylated ERK2 and p38α (ERK2/pTpY and p38α/pTpY) to generate mono-threonine phosphorylated form of MAP kinases (ERK2/pT and p38α/pT). The catalytic efficiencies of mono-threonine phosphorylated ERK2 and p38α are ~2–3 orders of magnitude higher than those of their unphosphorylated forms and only 1–2 orders of magnitude lower than those of the fully active bisphosphorylated ERK2 and p38α ^11,17–19. Consequently, the action of HePTP on dual-phosphorylated MAPKs would generate intermediate activity states of ERK2 and p38α, which may have distinct biological roles both in vitro and in vivo^20,21.

MAPK signaling cascades are mediated by protein-protein interactions of the kinases with their substrates and regulatory proteins^22–27. These interactions are regulated by docking sites on the MAPKs (D-sites) and conserved sequence motifs on their interaction partners (kinase interaction motif, KIM, also called D-site binding motif, D-motif). KIM is a conserved 13- to 16-amino-acid sequence, characterized by a group of basic residues (two to three lysines or arginines) and a Ø_A-X-Ø_B motif (where Ø_A and Ø_B are hydrophobic residues leucine, isoleucine, or valine)²⁶. A number of crystallographic studies showed that the binding of various KIM peptides to the D-sites induced a number of distal structural changes in different regions of MAPKs^28–36. Binding of transforming growth factor-beta-activated kinase 1-binding protein 1 (TAB1) via KIM to unphosphorylated p38α greatly increases the binding affinity for ATP ( ~ 2600-fold), allowing autophosphorylation of p38α in cis^37,38. The docking interaction between KIM peptide of MAP kinase-activated protein kinase 2 (MK2) and bisphosphorylated p38α enhances binding of ATP ( ~ 20 fold) and the phosphoacceptor to p38α, accelerating the phosphotransfer reaction³⁹. However, it is still unclear how docking interaction induced changes in protein conformation and how the conformational changes contribute to their biological functions.

In the present study, we showed that the catalytic activity of the monothreonine-phosphorylated p38α (p38α/pT) can be remarkably enhanced by direct binding of HePTP. To gain insight into the mechanism of HePTP-induced p38α/pT activation, we determined the crystal structure of p38α in complex with a KIM peptide derived from HePTP (pepHePTPm). In the complex, pepHePTPm binds to a docking site (D-site) in the C-terminal lobe of the kinase, which is situated in a non-catalytic region opposite to the activation loop. Binding of the peptide causes disorder in the activation loop and significant conformational change in the hinge region of p38α, which is different from those observed in other p38α-KIM peptide complexes. Molecular dynamics (MD) simulations suggest that the D-site docking interactions induce enhanced conformational flexibility in the p38α hinge region, resulting in increased access of ATP to the active site. Together with the previously determined crystal structure of the ERK2-pepHePTPm complex, our study also provides a high-resolution structural comparison of the interactions for p38α and ERK2 with the same KIM peptide.

Results

Catalytic activation of monothreonine-phosphorylated p38α by HePTP

HePTP consists of a C-terminal catalytic domain (residues 44–339) and a short N-terminal extension (residues 1–43) that functions to direct the enzyme to its physiological substrates (Fig. 1a). In an effort to understand the roles of phosphorylation at each site (Thr180 or Tyr182) with respect to the activation mechanism of p38α, we unexpectedly found that the catalytic activity of the monothreonine-phosphorylated p38α (p38α/pT) can be significantly enhanced by HePTP (Fig. 1b). The kinase activity of p38α was measured spectrophotometrically using a synthetic peptide substrate, EGFRtide^11,40. This assay couples the production of ADP with the oxidation of NADH by pyruvate kinase and lactate dehydrogenase⁴¹. It can be seen that in the absence of HePTP, only low kinase activity of p38α/pT was detected at 340 nm. A “flash activation” of the p38α/pT occurred when HePTP was added to assay system. Activation of p38α/pT was dose-dependent and saturable with half-maximal effect detected in the presence of 0.5 µM of HePTP (Fig. 1b, insert).

Fig. 1. Catalytic activation of p38α/pT by HePTP.

a Domain structures of p38α and HePTP.The N-lobe (slate), C-lobe (blue), and hinge region (orange) of p38α, along with the KIM (green) and catalytic domain (limon) of HePTP are shown. Sequences of the KIM and hinge region are provided, with key residues highlighted in green. b Time course of p38α/pT-catalyzed EGFRtide phosphorylation. Kinetics were monitored via continuous spectrophotometric measurement of ADP formation (PK/LDH coupled assay). The assay contained standard buffer, 43 nM p38α/pT, 1 mM ATP, and 200 µM EGFRtide. The addition of 1.8 µM HePTP is indicated by an arrow. The inset shows the concentration dependence of the initial velocity on HePTP. The experiment was performed in replicate for at least three times. The solid line represents the best fit of the initial rates versus HePTP concentration according to Eq. (1). c Time course of p38α/pT-catalyzed ATP hydrolysis monitored via ADP formation using the PK/LDH coupled assay (Method A). d Time course of p38α/pT-catalyzed ATP hydrolysis monitored via Pi release using the MESG/phosphorylase assay (Method B). For both methods, the reaction mixtures contained 480 nM p38α/pT, 2 mM ATP, and their respective standard buffers, with 1.2 µM HePTP added as indicated. e Dependence of p38α/pT-catalyzed ATP hydrolysis on HePTP and ATP concentrations. Initial rates were determined using the PK/LDH coupled assay at the indicated concentrations. The Michaelis-Menten model describing two-ligand binding was globally fitted to the data by nonlinear regression analysis. f Effects of HePTP mutants on the allosteric activation of p38α/pT. ATPase activity assays were performed in a standard PK/LDH coupled assay buffer containing 480 nM p38α/pT, 2 mM ATP and 7 µM wild-type or mutant HePTP proteins. Data are shown as activation relative to p38α/pT alone (mean ± SEM, n = 3), with dots representing experimental data points and error bars indicating SEM. g Allosteric activation of p38α/pT-catalyzed ATP hydrolysis by the HePTP KIM peptide (pepHePTP). The reaction included 480 nM p38α/pT and 2 mM ATP, with 140 µM pepHePTP added as indicated. The inset shows the initial velocity dependence on pepHePTP concentration. The experiment was performed in at least three independent replicates. The solid line represents the best fit of the initial rates versus pepHePTP concentration according to Eq. (1).

The activated p38α (p38α/pTpY and p38α/pT) displays robust ATPase activity¹¹. Therefore, we are able to monitor the ATP hydrolyzing activity of p38α/pT in the absence of any added protein substrate^11,40. ADP is a common product of kinase and ATPase reactions. Phosphate and phosphorylated peptide/protein, on the other hand, are unique products of the ATPase and kinase reactions, respectively. Since the rate of ADP release from the ATPase reaction should be stoichiometric with respect to the phosphate release, the kinetics of ATP hydrolysis can be monitored by one of the following two methods. In method A, the ADP generated during the reaction was quantified using pyruvate kinase and lactate dehydrogenase coupled-enzyme assay as described above. In method B, the phosphate that was generated by the hydrolysis reaction was quantitated by the nucleoside phosphorylation reaction. In this assay, the coupling enzyme purine nucleoside phosphorylase uses the inorganic phosphate to convert 7-methyl-6-thioguanosine to 7-methyl-6-thioguanine and ribose-1-phosphate, resulting in an increase in absorbance at 360 nm⁴². In the absence of HePTP, there is a considerable ATP hydrolysis activity as demonstrated from the detected ADP and phosphate release rate (Fig. 1c, d). Addition of HePTP resulted in a concomitant increase in the rates of ADP and phosphate production. The rate of ADP formation (0.041 ± 0.001 μM/s) from the ATPase reaction was identical to that of phosphate release within the error range (0.039 ± 0.001 μM/s). This result rules out the possibility that ADP’s production arose from transfer of phosphate from ATP to protein(s) (kinase activity)^11,40, and indicates that there was no detectable kinase activity of p38α towards either exogenous substrate (HePTP)¹⁶ or isogenous substrate (p38α itself) under the present experimental conditions^37,38. Furthermore, the phosphorylation states of p38α/pT at the beginning and the end of the reactions were also examined by Western blot analysis, which showed that no additional phosphorylation on p38α/pT was detected when incubated with HePTP and 2 mM ATP (Supplementary Fig. 1).

To investigate the mechanism of the HePTP-induced p38α/pT activation, the ATPase activity assay was performed for p38α/pT by varying the concentrations of both ATP and HePTP. In the presence of an allosteric activator, the general mechanism of the enzyme-catalyzed reaction can be represented as shown in Scheme 1

Scheme 1.

Kinetic mechanism for activation of p38α by HePTP.

where E represents enzyme p38α/pT, and A and S represent activator HePTP and substrate ATP, respectively. If we assume that the binary enzyme-substrate complexes remain in equilibrium with the free enzyme and substrate, a general initial velocity equation is given by the Eq. (1)

$$v=\frac{\left\{k_2{K}_A^{’}+k_2^{’}\left[A\right]\right\}{\left[E\right]}_0\left[S\right]}{K_A^{’}{K}_S+K_S^{’}\left[A\right]+K_A^{’}\left[S\right]+\left[A\right]\left[S\right]}=\frac{\left(K_A^{*}+\alpha \left[A\right]\right)v_0}{K_A^{*}+\left[A\right]}$$ 1

where

$$v_0=\frac{k_2{\left[E\right]}_0\left[S\right]}{K_S+\left[S\right]},K_A^{*}=\frac{K_A^{’}\left(K_S+\left[S\right]\right)}{K_S^{’}+\left[S\right]},\alpha =\frac{k_2^{’}\left(K_S+\left[S\right]\right)}{k_2\left(K_S^{’}+\left[S\right]\right)}$$

This equation involves five independent kinetic parameters, $K_A$, $K_A^{’}$, $K_S$, $k_2$ and $k_2^{\prime }$, to be determined experimentally. Figure 1e shows dependence of the initial velocity on HePTP concentration at four different fixed ATP concentrations. The data sets of initial velocities were analyzed with the Michaelis-Menten model describing two-ligand binding (Eq. (1)) by global fitting procedures. The kinetic parameters were determined to be $K_A=$ 3.2 ± 0.1 μM, $K_A^{’}=$ 0.95 ± 0.09 μM, $K_S=$1.69 ± 0.18 mM, $K_S^{’}=$0.50 ± 0.04 mM, $k_2=$0.051 ± 0.03 s^-1, and $k_2^{’}=$0.192 ± 0.004 s^-1, respectively. Thus, upon binding of HePTP to the non-catalytic region of p38α, the catalytic constant ($k_2^{’}$ vs.$k_2$) increases 4-fold and the Michaelis constant ($K_S^{’}$ vs. $K_S$) decreases 3-fold, resulting in a nearly 12-fold increase in the second order rate constant ($k_2^{’}/K_S^{’}$ vs.$k_2/K_S$). Our kinetic analysis suggests that HePTP-mediated activation of p38α/pT may involve mechanisms that promote both ATP binding and catalytic efficiency. In comparison with p38α/pT, the bisphosphorylated p38α/pTpY displays an 11-fold increase in k_cat for ATP hydrolysis (0.051 s^-1 vs 0.562 s^-1), with an ~8 fold decrease in $K_m$ (1690 μM vs 212 μM)¹¹, indicating that the role of the additional phosphorylation (pTyr182) is to optimally arrange the ATP binding site for the phosphoryl group transfer reaction. Thus, the partially active p38α/pT can employ biochemically distinct mechanisms (allosteric and covalent regulation) to achieve catalytic activation, albeit not to a similar degree, through kinetically similar means.

The N-terminal region of HePTP is important for p38α/pT activation

The physiological substrates of HePTP include ERK1/2 and p38α. These substrates bind HePTP via a 15 amino acid sequence, the kinase interaction motif (KIM, residues 16–31), within the HePTP N-terminus. In addition to KIM, the KIM-adjacent C-terminal region of HePTP (residues 32–79), which has been termed kinase-specificity sequence (KIS), also plays a key role in p38α/HePTP interactions^43,44. This region contains two phosphorylation sites for the active MAPKs ERK2 and p38α, Thr45 and Ser72, and their phosphorylation may block the docking interaction between MAPK and HePTP. Thus, there are at least three MAPK interaction regions in the HePTP molecule: (i) HePTP KIM interacting with MAPK docking site; (ii) the region around the HePTP phosphorylation site (Thr45 and Ser72) interacting with MAP kinase catalytic site; and (iii) HePTP catalytic site interacting with MAPK activation loop⁴⁵.

To elucidate the importance of these multiple interactions for p38α/pT activation, we generated various HePTP mutants via site-directed and truncation mutagenesis, and determined the effects of structural alterations on HePTP-induced p38α activation (Fig. 1f). Biochemical studies have shown that catalysis by the protein-tyrosine phosphatase (PTP) superfamily requires an essential nucleophilic Cys, and the catalytically inactive Cys-to-Ser mutant retains the ability to bind substrates, but is unable to carry out substrate dephosphorylation^46,47. We purified an inactive HePTP mutant in which Cys270 is changed to Ser (HePTP/C270S). Like wild-type HePTP, catalytically inactive HePTP/C270S increased the ATPase activity of p38α/pT to the same extent, demonstrating that HePTP enzymatic activity is not necessary for allosteric activation of p38α/pT (Fig. 1f). We also mutated the potential phosphorylation sites in HePTP, Thr45 and Ser72, to alanines and found these mutations do not affect HePTP-induced p38α/pT activation. Similar experiments conducted with HePTP/ΔN43 were unable to detect catalytic activation that enhances the ATPase activities of p38α/pT (Fig. 1f). These results support the notion that the N-terminal region of HePTP is important for allosteric activation of p38α/pT. Furthermore, we synthesized a peptide corresponding to the KIM sequence of HePTP (pepHePTP, ¹⁶RLQERRGSNVALMLDV³¹) and validated its role in the p38α/pT activation. It can be seen from Fig. 1g that upon the addition of pepHePTP into the assay system, a significant increase in p38α/pT ATPase activity was observed, and the activation of p38α/pT was dose-dependent and saturable with increasing concentrations of pepHePTP (Fig. 1g, insert). Direct fitting of the experimental data at a fixed ATP concentration (2 mM) to Eq. (1) yields an apparent $K_A^{*}$ of 43.6 ± 3.1 μM, which is about 30-fold higher than that of full-length HePTP ($K_A^{*}$= 1.4 μM), indicating that residues outside the KIM motif of HePTP also contribute to p38α binding^44,48.

Consistent with these data, the purified HePTP/ΔN43 did not interact with p38α/pT in vitro, as detected by gel filtration (Supplementary Fig. 2a). Our co-immunoprecipitation (co-IP) experiments also showed that full-length HePTP can physically interact with mono-threonine phosphorylated p38α and bisphosphorylated p38α, respectively, but the mutant HePTP/ΔN43 failed to bind to p38α (Supplementary Fig. 2b). In intact cells, PTP-SL and STEP distinctively regulated the kinase activity and nuclear translocation of ERK1/2 and p38α^43,49. PTP-SL associates with ERK1/2 and p38α in the cytoplasm of intact COS-7 cells, and regulates their activity by both protein localization (by retaining the MAPKs in the cytosol) and dephosphorylation of their active sites. However, co-expression of STEP and these MAPKs in COS-7 cells did not result in their physical association, and no inhibition of MAPK catalytic activity and nuclear translocation was observed. We therefore investigated the effect of HePTP on the intracellular localization of p38α. For these experiments, COS-7 cells were cotransfected with p38α-GFP and HA-HePTP, followed by immunofluorescence analysis. As shown in Supplementary Fig. 2c (a–c), p38α-GFP was distributed both in the nucleus and in the cytosol, whereas HA-HePTP was located exclusively in the cytosol but not in the nucleus. Notably, coexpression of p38α-GFP with HA-HePTP resulted in exclusion of p38α-GFP from the nucleus and its retention in the cytosol (Supplementary Fig. 2c, d–f). A similar result was observed for co-transfection of p38α-GFP with the inactive mutant HA-HePTP/C270S, indicating that cytoplasmic retention of p38α is independent of the catalytic activity of HePTP (Supplementary Fig. 2c, g–i). These results demonstrate that coexpression of HePTP and p38α in COS-7 cells effectively blocks p38α entry into the nucleus.

Taken together, our experimental observations give rise to a working model for the regulation of p38α by HePTP in cells. Upon dual phosphorylation by MAP kinase kinase, activated MAP kinase is translocated to the nucleus, where it regulates the expression of immediate-early genes. Subsequent dephosphorylation of MAP kinase is then thought to mediate its exit from the nucleus. In addition to phosphorylating transcription factors, MAP kinases can also phosphorylate cytosolic targets to regulate intracellular events. An additional layer of regulation is provided by HePTP, which is a cytosolic enzyme expressed in resting T-cells and forms a physical complex with ERK1, ERK2 and p38α. First, HePTP efficiently dephosphorylates the phosphotyrosine residue in the activation loop of p38α, thereby partially inactivating p38α, and second, HePTP can sequester it by forming a stable 1:1 complex, thereby preventing p38α from translocating to the nucleus. Binding of HePTP to mono-threonine-phosphorylated p38α allosterically activates its enzymatic activity toward some non-KIM containing substrates. Thus, despite the apparently simple architecture of the p38α MAP kinase pathway, this enzyme is capable of responding to diverse extracellular stimuli to produce exquisitely specific cellular outcomes. These responses depend on the kinetics of activation and inactivation, subcellular localization, the complexes in which p38α acts, and substrate availability. Our model provides a flexible mechanism for the regulation and fine-tuning of MAP kinases, which function as crucial signal integration and decision points in T cell activation as well as in growth and stress responses in many other cell types.

Structure of the p38α-pepHePTPm complex

To understand the structural basis for the HePTP-induced p38α/pT activation, we initially attempted crystallization of both p38α/pT alone and p38α/pT complexed with a KIM peptide derived from HePTP (pepHePTP). Unfortunately, despite extensive crystallization trials, we were unable to obtain crystals of p38α/pT for structural determination. To elucidate the molecular determinants that direct p38α-pepHePTP interactions, we then determined the crystal structure of inactive p38α in complex with pepHePTP. Because efforts to crystallize the complex of p38α with wild-type KIM peptide failed, we made a mutant peptide, pepHePTP/V31C (referred to below as pepHePTPm), which may form a disulfide bond with p38α in analogy with the disulfide bridge that occurs in the p38α-pepMKK3b complex, to produce a crystal suitable for structural analysis²⁸. The binding of pepHePTPm can cause similar allosteric effects on p38α/pT activation (Supplementary Fig. 3). The crystal structure of inactive p38α bound to pepHePTPm was solved at 2.14 Å resolution (Table 1). The final model includes all residues of p38α, except for twelve residues in the activation loop (174–184), and nine residues in the C-terminus (354–360) (Fig. 2a). The electron density for the associated pepHePTPm is well-defined from Arg¹⁶ to Cys³¹ (Supplementary Fig. 4), and this peptide forms a short helix involving ¹⁷LQER²⁰ at its N-terminus.

Table 1.

Data collection and refinement statistics

Data Collection	p38α-pepHePTPm
Space group	P3121
Cell dimensions
a, b, c (Å)	81.96, 81.96, 122.40
α, β, γ(°)	90, 90, 120
Wavelength (Å)	1.54
Resolution (Å)	50.0-2.14 (2.20-2.14)a
Unique Reflections	26,386 (1,741)
Completeness (%)	98.0 (80.5)
I/σ(I)	18.9 (2.2)
Redundancy	7.9 (3.3)
CC1/2	1.00 (0.89)
Rmerge	0.127 (0.89)
Refinement
Resolution (Å)	46.35-2.14
Rwork/Rfree	0.186/0.216
R.m.s deviations
Bond length (Å)	0.010
Bond angles (°)	1.48
Average B values (Å2)	37.70
Ramachandran plot (%)
Favored	97.45
Allowed	2.55
Disallowed	0.0

^a Numbers in parentheses donate the values in highest resolution shell.

Fig. 2. Crystal structure and docking interactions of the p38α-pepHePTPm complex.

a Overall structure of the p38α-pepHePTPm complex. The color scheme follows that in Fig. 1A. The dotted line indicates disordered region of the activation segment. Key structural elements are labeled and indicated by arrows. b pepHePTPm-induced conformational changes of p38α. Superimposition of p38α-pepHePTPm with unphosphorylated p38α (gray, PDB: 5UOJ) and bisphosphorylated p38α/pTpY (pale yellow, PDB: 3PY3) using the C-lobe as reference. Major regions of conformational change induced by pepHePTPm binding are indicated. The inset shows a close-up view of the active site changes. The side chains are highlighted as sticks, the K-E distances are indicated, and a water molecule is shown. c Details of the interactions between p38α and pepHePTPm. For clarity, pepHePTPm is shown as a ribbon. Residues involved in the interaction are shown as sticks, with p38α in cyan and pepHePTPm in green. Hydrogen bonds are depicted as magenta dotted lines, and the water molecule is represented as a yellow dot. d Effects of KIM sequence mutations on HePTP-catalyzed p38α/pTpY dephosphorylation. Assays were performed with 40 nM HePTP proteins (wild-type or indicated mutants) and varying concentrations of p38α/pTpY. Catalytic efficiencies ($k_{cat}/K_m$) for wild-type and indicated HePTP mutants toward p38α/pTpY were obtained by fitting the data to the Michaelis-Menten equation (mean ± s.e.m., n = 3).

The complex structure shows that the pepHePTPm binds to the so-called “KIM-binding site” of p38α (also referred to as D-site), a common docking site in MAPKs for interaction partners that have docking motifs (KIMs or D-motifs). Long-range conformational changes accompany formation of the complex between p38α and pepHePTPm. Unlike the unphosphorylated p38α structure, most residues of the activation loop in the p38α-pepHePTPm complex are crystallographically disordered. The structure of p38α in the present complex superimposes on the unphosphorylated and bisphosphorylated apo-p38α with root-mean-square deviations (RMSD) of 0.45 Å and 1.01 Å for all Cα atoms, respectively (Fig. 2b). There is a closure and approximation of the N- and C-lobes of the kinase most usually associated with phosphorylation of the activation segment. The glycine-rich loop in p38α-pepHePTPm is moved from its up position found in the inactive p38α structure to its down position. Furthermore, within the N lobe there was an inward swing of αC that brought Glu71 within 2.7 Å of Lys53 and enabled formation of the salt bridge typical of an active kinase (Fig. 2b). It is expected that these changes within the nucleotide-binding pocket will facilitate localization of the phosphate binding ribbon and increase affinity toward ATP. The catalytic base (Asp150 in the HRD motif) which is essential for proton transfer during catalysis, adopts a conformation similar to that observed in active p38α (Fig. 2b).

In our structure, the DFG motif does not fully align with the active “DFG-in” conformation. Instead, it exhibits a “DFG-intermediate” configuration closer to the inactive p38α (Fig. 2b). The crossover connection (¹⁰⁷HLMGA¹¹¹, also called the hinge region) of p38α is located at the back of the ATP binding site and interacts with the adenine ring of ATP, based on the structure of ATP-bound ERK2⁵⁰. The hinge region in the complex adopts a distinct conformation, which is different from either inactive or active apo-p38α (Figs. 2b and 3b). The side chain of His107 rotates about 90° toward the large lobe relative to inactive p38α, and residue Met109 moves toward the adenine pocket of the ATP binding site, preventing ATP access to the active site (Fig. 2b). These structural observations seem to be inconsistent with the p38α/pT activation data suggesting that binding of HePTP to p38α/pT significantly increases its ATP binding affinity. Thus, the molecular mechanism of the HePTP-induced p38α/pT allosteric activation remains largely unexplained at the atomic level (see below for further discussion).

Fig. 3. Comparison of interaction modes and conformational changes between p38α-pepHePTPm and ERK2-pepHePTPm complexes.

a Binding interactions in the p38α-pepHePTPm and ERK2-pepHePTPm complexes (PDB: 2GPH). The sequence of pepHePTPm is shown, with hydrophobic residues (Ø_U, Ø_L, Ø_A, Ø_B) labeled. The surface representations of p38α and ERK2 are colored according to electrostatic potential (positive, blue; negative, red), with a color bar showing the potential range ( ± 150 kT/e). PepHePTPm is shown in a ribbon (green in the p38α complex and forest green in the ERK2 complex) with side chains as sticks. The classical CD and ED regions that interact with basic residues, and the hydrophobic pockets (Φ_U, Φ_L, Φ_A, Φ_B) for binding hydrophobic residues are indicated. b Conformational changes in the hinge region of p38α induced by pepHePTPm binding. Structures of p38α (gray) and p38α/pTpY (pale yellow), as well as unphosphorylated ERK2 (gray, PDB: 1ERK) and bisphosphorylated ERK2/pTpY (cyan, PDB: 2ERK), were superimposed with their C-lobes onto the p38α-pepHePTPm (colored as that in Fig. 2a) and ERK2-pepHePTPm (light blue), respectively. Side chains of hinge region residues are shown as sticks. c Effects of HePTP on the ATP hydrolysis activity catalyzed by p38α/pT and ERK2/pT. The assays were performed in the presence of 480 nM p38α/pT or ERK2/pT, 2 mM ATP, and 1.2 µM HePTP (as indicated), with absorbance at 340 nm measured continuously during the reaction.

Docking interactions and mutational analysis

The KIM binding site spans across helices αD and αE in a crevice formed between these helices and β7-β8 reverse turn in the C-terminal lobe of p38α (Fig. 2a). This docking site is comprised of a highly acidic patch and a hydrophobic groove (Figs. 2c and 3b). The acidic patch contains two subsites: a CD (common docking) region involving Asp313, Asp315, and Asp316; and a Glu160-Asp161 ED region²⁸. The hydrophobic groove has four small hydrophobic pockets, named the Φ_A-X-Φ_B groove (the first two hydrophobic pockets identified), Φ_L (L, lower; also referred to as Φ_A − 2) and Φ_U (U, upper; also referred to as Φ_H)²⁷. Two residues that line the Φ_B pocket are also part of the hinge region (¹⁰⁷HLMGAD¹¹² in p38α and ¹⁰⁴DLMETD¹⁰⁹ in ERK2), which is a flexible connector between the N-terminal and C-terminal kinase lobes. KIMs (D-motifs) in MAPK binding partners can have different lengths. The KIM sequence of HePTP has a well-established consensus sequence “X-Ø_U-X₂-(K/R)_1-2-X_2-6-Ø_A-X-Ø_B” with longer intervening regions (Fig. 1a). Sequence alignments suggest that the Ø_A-X-Ø_B motif of HePTP is ²⁷LML²⁹ (number in superscript denotes the residues from the pepHePTPm)^23,29. Two basic residues ²⁰RR²¹ in the HePTP KIM have been previously proposed to form important functional interactions with the negatively charged CD region of MAPKs²³.

The structure of the p38α-pepHePTPm complex shows that the basic region of pepHePTPm interacts with the CD region and the Ø_A-X-Ø_B residues bind in the Φ_A-X-Φ_B groove, a mode conserved in p38α when binding to its substrates/activating kinases^27,33. In the complex, residues 17-19 of the pepHePTPm folded into a small one-turn helix at its N-termimus that facilitates the formation of intermolecular interactions between the p38α CD region and the pepHePTPm (Fig. 2c). The guanidinium group of Arg¹⁶ of pepHePTPm forms a salt bridge with Asp316 and hydrogen bonds with Tyr311, Tyr132, and Asp316. The adjacent hydrophobic residue, Leu¹⁷, occupies the Φ_U pocket of p38α, and Gln¹⁸ makes two hydrogen bonds with the main chain of Asp161 and side chain of Glu163. Unexpectedly, the basic submotif of pepHePTPm (²⁰RR²¹) makes direct contacts with residues outside the CD region of p38α. Arg²⁰ makes two hydrogen bonds with Gln310 of p38α; and the side chain guanidinium group of Arg²¹ makes a hydrogen bond with the main chain of Asp161. In addition, there is an extensive hydrogen bond network, involving four buried water molecules that bridge the Arg²¹ of pepHePTPm to the acidic residues of p38α (Asp161 and Asp125). KIMs containing an RR motif usually structured in a short helix that is required for binding to the CD region of MAPKs (an electrostatic surface depression)^29,32,51,52. By contrast, the ²⁰RR²¹ of pepHePTPm does not adopt a helical conformation in the p38α-pepHePTPm complex. Thus, the pepHePTPm docking to the p38α CD region represents a previously unrecognized interaction mode.

The interaction between the Ø_A-X-Ø_B sequence of pepHePTPm (²⁷LML²⁹) and the hydrophobic groove of p38α is highly similar to that observed in p38α in complex with their respective docking motifs^28,33,38,52. The ²⁶ALMLDC³¹ segment of pepHePTPm occupies a similar position to that in p38α-pepMKK3b, utilizing the same binding pockets (Φ_A and Φ_B) for the ²⁷LML²⁹ motif, forming the same disulfide bridge between Cys³¹ (Ø_B + 2) and Cys119 of p38α (Fig. 2c). The disulfide does not appear to influence the interactions of Ø_B, since this residue (Leu²⁹) adopts a conformation similar to that found in pepMKK3b²⁸. Additional interactions are also observed between residues adjacent to the Ø_A-X-Ø_B motif and the hydrophobic groove. Like pepMEF2A and pepMKK3b, the pepHePTPm backbone forms hydrogen bonds with the side chains of p38α Gln120, His126, and Glu160. In addition, the backbone carbonyl of Leu²⁹ forms water-mediated hydrogen bonds with both the backbone and the side chain of Asn115 of p38α.

To confirm the importance of the observed interactions in the p38α-pepHePTPm complex, we carried out site-directed mutagenesis in the KIM sequence of HePTP (residues 16-31) and determined the effects of these mutations on HePTP-catalyzed p38α/pTpY dephosphorylation (Fig. 2d). It was shown previously that deletion of the N-terminal region from HePTP (residues 1-43, HePTP/ΔN43) completely abolishes its phosphatase activity to p38α/pTpY, although the catalytic activity of this mutant is similar to that of the wild-type HePTP using pNPP as a substrate¹¹. These data indicated that the N-terminal truncation does not affect the intrinsic phosphatase activity of HePTP. Therefore, K_m can be considered as a measure of the binding affinity between p38α and the HePTP mutants, while the ATPase assay for p38α activation by HePTP involves more reaction events (see Eq. 1). When the Arg residues at positions 20 and 21 were replaced by Ala, the $k_{cat}/K_m$ of R20A and R21A for p38α dephosphorylation decreased 17 and 33-fold, respectively. Compared with wild-type HePTP, replacement of both Arg²⁰ and Arg²¹ with Ala (R20A/R21A) led to an 80-fold drop in the $k_{cat}/K_m$ value. Similarly, when the key hydrophobic residues Leu²⁷ and Leu²⁹ were individually replaced by Ala, their catalytic efficiencies ($k_{cat}/K_m$) were decreased approximately 10-20-fold, and substitution of both Leu²⁷ and Leu²⁹ with Ala (L27A/L29A) resulted in a further decrease ( ~ 40-fold) in $k_{cat}/K_m$. Mutating Leu¹⁷ at the N terminus of the KIM to Ala (L17A) led to marked enzymatic inactivation, whereas no significant effects were observed for the HePTP mutants R16A, Q18A and D30A. Collectively, these data indicated that both the hydrophobic and charged residues within KIM of HePTP are required for its efficient dephosphorylation.

Differences between p38α-pepHePTPm and ERK2-pepHePTPm complexes

The structure of unphosphorylated ERK2 in complex with the 16-mer KIM peptide (pepHePTPm) has been reported²⁹. To define the structural basis of the activation mechanism of p38α by HePTP, we compared our p38α-pepHePTPm crystal structure to the ERK2-pepHePTPm complex (Fig. 3a). Although pepHePTPm interacts with the docking sites through a similar bipartite mechanism, the residues of HePTP interacting with the two subsites in p38α and ERK2 are considerably different. In the ERK2-pepHePTPm complex, both Arg²⁰ and Arg²¹ of HePTPm bind intimately to Asp319 in the ERK2 CD region²⁹. By contrast, in p38α-pepHePTPm, there is no direct charge-charge interaction between the positively charged residues (Arg²⁰ and Arg²¹) and the negatively charged residues (Asp313, Asp315 and Asp316) in the p38α CD region. Instead, HePTP Arg¹⁶ forms extensive interactions with the CD region of p38α, whereas Arg²⁰ and Arg²¹ make direct contacts with other residues flanking the CD Asps (Figs. 2c and 3a). Moreover, there is a major difference in the docking interactions between the Ø_A-X-Ø_B motif and the hydrophobic groove in the two structures. In the p38α-pepHePTPm complex, the HePTP Ø_A-X-Ø_B residues (Leu²⁷-Met²⁸-Leu²⁹) bind to the Φ_A-X-Φ_B groove of p38α in a canonical manner (Fig. 2c). However, the crystal structure of the ERK2-pepHePTPm complex revealed a register shift in the Ø_A-X-Ø_B motif in which Leu²⁷ and Leu²⁹ occupy the Φ_L and Φ_A binding pockets, respectively, and no direct contact between the residue of pepHePTPm and Φ_B pocket in ERK2 is observed.

Structural comparison between the two complexes also shows that the bound pepHePTPm exhibits significant variability in main chain conformations and its docking interactions induce different conformational changes in the two MAPKs (Fig. 3b). For example, binding of HePTPm peptide causes a significant conformational change in the crossover connection of p38α, whereas no conformational changes occur along this region in ERK2. In p38α MAPK, a small glycine residue (Gly110) is directly adjacent to the linker residue (Met109), and this arrangement does not exist in ERK2 where the glycine residue is replaced with a bulkier amino acid residue (Glu) (Fig. 3b). L-glycine is able to undergo a backbone flip, which gives rise to an enhanced p38α flexibility of the hinge region. The differences observed in the conformational flexibility of this region may be linked to different responses of p38α and ERK2 to HePTP binding.

In addition to the leukocyte-expressed HePTP, its brain-specific homologs, striatal-enriched protein tyrosine phosphatase (STEP) and STEP-like protein tyrosine phosphatase (PTP-SL), can also inactivate ERK and p38 MAP kinases by dephosphorylating pTyr in the TXY activation loop^19,53. These MAPK-specific tyrosine phosphatases have in common the existence of a highly conserved KIM and therefore presumably share a common function as negative regulators of the ERK and p38 MAP kinase pathways. To examine the effects of these tyrosine phosphatases on the catalytic activity of p38α/pT and ERK2/pT, the ATPase activities of p38α/pT and ERK2/pT were measured in the absence and presence of the KIM-containing PTPs (Supplementary Fig. 5). The ATPase activity of p38α/pT was specifically increased in the presence of 1.2 μM recombinant HePTP, PTP-SL, and STEP, indicating that all three tyrosine phosphatases allosterically induced catalytic activation of p38α/pT (Fig. 3c and Supplementary Fig. 5 left). In contrast, although the mono-threonine phosphorylated ERK2 (ERK2 phosphorylated on Thr183, ERK2/pT) has very similar kinetic behavior to that of p38α/pT^11,17,18, the ATPase activity of ERK2/pT was unaffected by these KIM-containing PTPs under the identical experimental conditions (Fig. 3c and Supplementary Fig. 5 right). These results indicate that the KIM-containing PTP-mediated catalytic activation appears to be limited strictly to p38α MAP kinase.

Comparison of peptide-induced conformational changes among p38α-peptide complexes

The first crystal structures of the MAPK-KIM peptide complexes (p38α-pepMEF2A and p38α-pepMKK3b) were described twenty years ago²⁸. Since then, a body of data has been accumulated that docking motifs in MAPK kinase substrates and processing enzymes mediate MAPK interactions. Binding of the peptides to the D-site induces conformational changes in the active site and disorder in the phosphorylation loop²⁷. Comparisons of our structure with the previously reported complex structures reveal that despite the strong similarities between p38α-pepHePTPm, p38α-pepMEF2A and p38α-pepMKK3b, there are differences (Figs. 3a and 4a). The pepHePTPm clearly interacts with both the CD region and the hydrophobic groove, while docking interactions observed in the structures of pepMEF2A and pepMKK3b were limited to the hydrophobic groove and the Ø_A-X-Ø_B motif. In addition to the Φ_A and Φ_B pockets, a third hydrophobic pocket, Φ_L, binds the KIM peptides found in both the p38α-pepMEF2A and p38α-pepMKK3b complexes, but no interactions with the (Φ_L) site were observed in the p38α-pepHePTPm structure. In the p38α-pepHePTPm complex, the hinge region of p38α adopts a different conformation than previously seen in the complexes with either the transcription factor substrate pepMAF2A or the activating enzyme pepMKK3b (Figs. 3b and 4b). Chang et al. suggested that the conformational change induced by pepMEF2A could perhaps activate the enzyme, and in the case of MKK3b, binding of the activating enzyme may release the activation loop of p38α for binding in the MKK3b active site allowing for phosphorylation²⁸. However, crystallographic data from doubly phosphorylated p38α complexed with pepMKK3b (p38α/pTpY-pepMKK3b) reveal a structure similar to unphosphorylated p38α-MKK3b, and distinct from bisphosphorylated active apo-p38α (p38α/pTpY)⁵⁴. Thus, the functional relevance of these conformational changes remains to be tested.

Fig. 4. Comparison of interaction modes among p38α-KIM peptide complexes.

a Surface representation of the docking groove with bound pepMKK3b (PDB: 1LEZ), pepMEF2A (PDB: 1LEW), and pepMK2 (PDB: 2OKR). Colored arrows indicate the N → C-terminal direction of the docking peptides. The p38α is shown as a surface colored by electrostatic potential, with hydrophobic pockets (Φ_L, Φ_A, Φ_B) indicated. The KIM peptides are shown in cartoon representation with C-terminal hydrophobic residues as sticks. b Peptide-induced conformational comparison of the hinge region in p38α. The structure of p38α is shown in slate, with the hinge region highlighted in orange. All structures were superimposed based on their corresponding Cα atoms within the C-lobe. The side chains of His107 and Met109 and the Ø_B residue of the KIM are shown as sticks. c Allosteric activation of p38α/pT ATPase activity by KIM peptides. The assays included standard buffer, 480 nM p38α/pT, 2 mM ATP, and KIM peptides (pepMKK3b, pepMEF2A, or pepMK2), added as indicated. ATPase activity was measured by monitoring ADP production at 340 nm. The insets show the initial rates of p38α/pT-catalyzed ATP hydrolysis versus the pepMKK3b, pepMEF2A and pepMK2 concentrations. Each experiment was performed in replicate for at least three times. The solid lines represent the best fit of the initial rates versus peptide concentration according to Eq. (1).

A number of recent studies have shown that p38α is also capable of binding KIM peptides/proteins in a reversed N- to C-terminal orientation compared to the canonical KIM peptide binding mode. MAPK-activated protein kinase-2 (MK2), a non-transcription factor substrate of p38α, was the first structure to show that the KIM peptide (residues 369–384, pepMK2) binds to p38α in the opposite direction (the hydrophobic residues are situated N-terminal of the basic residues)^30,51. An overlay of p38α complexed with MK2 and p38α complexed with pepMEF2A and pepMKK3b shows spatial conservation of the motif Ø_A-X-Ø_B despite the reversed direction of the peptide backbone (Fig. 4a). The pepMK2 binding to p38α also induces a side-chain displacement of the hinge region Met109, a rearrangement that is similar to the pepMEF2A-induced p38α changes (Fig. 4b).

Together, the previous and present studies demonstrate that binding of the KIM peptides to the same docking groove can induce distinct conformational changes in the hinge region of p38α. The conformational changes observed in this region appear to require occupation of the Φ_B binding pocket, which makes intimate interaction with Ala111 in the p38α-KIM complexes (Fig. 4b). Because the Φ_B interactions between KIM peptides and p38α are so highly conserved, we are wondering if the p38α/pT activity can be activated by distinct KIM peptides. To address this question, we examined catalytic activation of p38α/pT by distinct KIM peptides, using the activation-based assay developed above. Figure 4c shows that all KIM peptides were able to activate p38α/pT ATPase activity in a dose-dependent and saturable fashion, demonstrating that the canonical KIM docking interactions cause similar allosteric effects on p38α function. The apparent dissociation constants for the KIM peptides were determined by fitting the observed rates of ATP hydrolysis versus activator concentrations to Eq. (1). One possible explanation for the observed discrepancies between the kinetic and X-ray results might be that the KIM peptide binding induces dynamic conformational changes of the p38α hinge region, and enhanced flexibility in the active site may contribute to increased specific activity by facilitating nucleotide binding.

Molecular dynamics simulations of apo-p38α and pepHePTPm-bound p38α

We hypothesize that the hinge regions of p38α-KIM complexes fluctuate between multiple conformations in solution and the observed conformations in the crystal structures were distinct “snapshots” of states under various crystallization conditions. To test this hypothesis, we thus turned to long-timescale MD simulations, which can reveal conformational changes on ns-μs timescales⁵⁵. The 1 µs unbiased MD simulations for apo-p38α and pepHePTPm-bound p38α were performed in explicit solvent starting with their crystal structures. The simulations of apo p38α and p38α-pepHePTPm complex are illustrated in Supplementary Movies S1 and S2 with the side chain of Met109 highlighted. Figure 5a illustrates the time evolution of root-mean-square-deviation (RMSD) of heavy atoms on the backbone from the initial starting structures (inactive state, PDB ID: 5UOJ; and the p38α-pepHePTPm complex) in the two free MD trajectories. After ~10 ns of simulations, the systems reached equilibrium and pepHePTPm indeed bound stably at the D-site of p38α. The trajectory with the inactive state is seen to reach a plateau value of 3.54 ± 0.30 Å, whereas the RMSD of the peptide bound state reaches a value as high as 4.35 ± 0.95 Å. After 450 ns of simulations, the p38α-pepHePTPm complex exhibited larger fluctuations.

Fig. 5. MD simulations of apo p38α and p38α-pepHePTPm complex.

a RMSD (Cα) of p38α (green) and p38α-pepHePTPm (blue) reaches plateau values of 3.54 ± 0.30 Å and 4.35 ± 0.95 Å during the simulation. The time point at 450 ns is marked by a vertical red line. b RMSF of the p38α Cα atoms for all residues of p38α and p38α-pepHePTPm. Highly flexible regions involved in kinase regulation are indicated. c Ramachandran plots of DFG-Asp and DFG-Phe in p38α and p38α-pepHePTPm. The Ramachandran regions are marked A (alpha), B (beta), L (left), and E (epsilon). The areas showing a trend toward the BLAminus cluster are encircled by orange dashed lines. d RMSD (Cα) of the hinge region (residues 107-112) in p38α and p38α-pepHePTPm. Conformational transitions at specific time points are marked by vertical red lines. e Structural snapshots of the hinge region in p38α-pepHePTPm. Snapshots at indicated time points during the MD simulation are illustrated, with p38α-pepHePTPm shown as a cartoon representation (p38α in slate and pepHePTPm in pale green). The insets show close-up views of the hinge region changes. For reference, the apo p38α structure is shown in gray. At 9 ns, the equilibrated model of p38α-pepHePTPm. At 16 ns, the hinge region undergoes a conformational change, with His107 and Met109 shifting to an inactive-like conformation. At 203 ns, the hinge region transitions to an active-like conformation, with His107 and Met109 adopting a conformation that facilitates activation.

Figure 5b illustrates the root-mean-square fluctuation (RMSF) of each residue around the average structure from 1 µs MD simulation. This plot reveals the regions that have undergone major deviations from the initial starting structures during the free MD. For both apo and peptide-bound forms, the residues of p38α with higher fluctuation values are observed in several loop regions, including the Gly-rich loop (residues 31–36), the αC helix (residues 62–77), the hinge region (residues 107–112), the activation loop (residues 168–192), the αG helices region (residues 228–261), and the C-terminal tail (residues 306–348). Looking at the differences between the systems, one notable difference consists of the peaks whose whole mobility is significantly reduced upon pepHePTPm binding, suggesting that specific interactions of the KIM tend to stabilize the D-site of the C-terminal domain. Another change is observed in p38α corresponding to the activation loop, whose mobility is increased upon pepHePTPm binding. The kinase activation loop is crucial for catalysis and begins with the conserved DFG motif (residues ¹⁶⁸Asp-Phe-Gly¹⁷⁰ in p38α). The transition between active and inactive states of kinases involves a characteristic movement of the DFG motif which undergoes substantial changes in backbone dihedral angles during activation⁵⁶. Recently, Modi and Dunbrack have developed a clustering of kinase conformations based on the location of the Phe side chain (DFGin, DFGout, and inter) and the backbone dihedral angles of Asp and Phe in the DFG motif. According to their classification of kinase conformations, we analyzed the DFG motif conformation by calculating the distances between the DFG-Phe side chain and two conserved residues. The scatterplot of distances shows a clear cluster of points characteristic of the DFGin conformations for both systems (Supplementary Fig. 6). We then measured the φ/ψ angles of Asp168 and Phe169 to characterize their distributions within the DFGin. The backbone conformations of the DFG-Asp and DFG-Phe occupy the beta (B), left (L), and alpha (A) regions of the Ramachandran map, respectively (as defined in Fig. 5c). It can be seen that peptide binding significantly alters the φ/ψ angle distributions, with DFG-Asp transitioning from a broad distribution across the E/L/B regions to confined L/B regions, and DFG-Phe shifting towards the A region. The peptide-induced dynamic changes in φ/ψ angles observed along the simulation exhibit a trend toward the features of the BLAminus cluster, a conformation that represents the active DFGin state of kinases⁵⁶. This process is accompanied by subtle adjustments in activity-regulating elements that likely cooperate to drive the transition to an active conformation.

Notably, the hinge region in the peptide-bound state also showed a larger conformational flexibility than the inactive state during the MD simulations (Supplementary Movies S1 and S2). Figure 5d shows the plot of RMSD for the hinge region vs simulation time from the initial starting structure in the two free MD trajectories. The structural fluctuations of the hinge region in the p38α-pepHePTPm complex were much larger than those observed for apo p38α. It can be seen from Movies S1 and S2 that in the pepHePTPm-bound state, the hinge region of p38α was able to flip back and forth from the observed apo-state. Figure 5e illustrates snapshots of partial structures of the p38α-pepHePTPm complex (residues ¹⁰⁷HLMGA¹¹¹) during various times in the MD simulations. The representative snapshots (9 ns, 16 ns and 203 ns) were selected from the 1 µs simulation trajectories based on their similarities to the three distinct conformations observed in crystal structures of p38α-pepMKK3b, uncomplexed p38α and p38α-pepMAF2A, respectively. As shown in Fig.5e in the snapshot at 16 ns, the conformations of the hinge region in p38α-pepHePTPm closely resembled that observed in the crystal structure of inactive apo p38α. In the snapshot at 9 ns, the hinge region moved towards the active site and the side chains of His107 and Met109 rotated into the adenine pocket of the ATP binding site, whereas in the snapshot at 203 ns, the hinge region shifted away from the active site as Met109 occupied a new pocket formed in the β7-β8 reverse turn. In summary, the binding of KIM peptides induces conformational changes that maintain some features of the active kinase conformation. These changes involve critical structural elements, such as the hinge region, the glycine-rich loop, and the DFG motif, which likely work in concert to facilitate the transition towards an active conformation. The hinge region, which connects the N- and C-lobes, exhibits increased flexibility that may promote ATP access to the active site. As a comparison, we also performed additional MD simulations of the pepHePTPm-p38α complex without the Cys31-Cys119 disulfide bond (Supplementary Fig. 7). Although the complex became unstable after ~450 ns due to peptide dissociation (Supplementary Movie S3), the overall conformational features prior to dissociation remained largely similar to those observed for the complex with the Cys31-Cys119 disulfide bond. These results indicate that the increased hinge region flexibility is a dynamic feature of the peptide-bound state, independent of the disulfide bond.

Based on the MD trajectory of the peptide-bound state, we performed shortest-path analysis to identify a potential allosteric pathway connecting the KIM peptide-binding site to the hinge region. The top ten most likely pathways were obtained from the set of residues connecting the CD region of the KIM-binding interface to the hinge region (Supplementary Table 1). The Paths are not randomly distributed but instead show a high degree of convergence toward the hinge region. The optimal pathway is defined through the connections Gln310-Ala309-Gln133-Ile134-Lys165-Asn159-Ala111 (Supplementary Fig. 8). Interestingly, most of these residues are traversed in close proximity to the Φ_A and Φ_B pockets, spatially connecting the KIM-binding site to the hinge region. Thus, the enhanced hinge flexibility is likely mediated by the allosteric pathway connecting docking interface residues to the hinge region.

Discussion

Full activation of MAPKs requires dual phosphorylation of the threonine and tyrosine residues in the activation loop by specific MAPK kinases. Depending on the phosphorylation status, MAPKs can bind ATP and catalyze the transfer of the γ-phosphate from ATP to hydroxyl groups of downstream protein targets. Kinetic studies showed that the major effect of phosphorylation and activation of ERK2 is to increase k_cat, with minor effects on the binding affinity for ATP ($K_m$)¹¹. However, for the p38α-catalyzed reaction, both k_cat and $K_m$ values change substantially upon activation of p38α¹¹. Structural comparison of ERK2 and p38α provides an explanation for this observation. Both inactive and active X-ray structures of ERK2 show a properly formed ATP binding site and no significant conformational changes around the hinge upon phosphorylation^50,57. In contrast, the structure of inactive p38α reveals that the ATP site is malformed^58,59. Upon phosphorylation, p38α goes through characteristic global conformational changes of domain closure and rotation. The crossover connection of p38α is optimally arranged for accommodating ATP during the kinase activation¹¹. These previous works, taken together with the present results, support the view that both direct phosphorylation and D-motif docking interactions switch p38α from its low affinity ATP state to a high affinity ATP state through the rearrangements in the hinge region.

Selective recognition of substrates and effectors by MAPKs involves electrostatic and hydrophobic interactions between key residues at the MAPK docking grooves and at the interaction sites of the substrates and effectors. To understand the molecular mechanism of substrate specificity, Piserchio et al. compared the docking interactions of the HePTP peptide with ERK2 and p38α using solution NMR spectroscopy and isothermal titration calorimetry (ITC)⁴⁴. Their results demonstrated that the KIM-mediated interactions for the two kinases, while similar overall, do show some specific differences. In the present study, we report a crystal structure of inactive p38α in complex with the KIM peptide derived from HePTP. Together with prior structural knowledge, our structure provides, for the first time, a high-resolution crystallographic comparison for the interactions of ERK2 and p38α with the same KIM peptide. Direct comparison of the two complex structures revealed that although pepHePTPm binds p38α and ERK2 at the same locus, the peptide adopted different main chain conformations between the contact points corresponding to the KIM consensus sequence. Particularly, the charge interaction patterns between the D-site Asp and the pepHePTPm Arg are completely different. These findings could explain some controversial results regarding the specificity of HePTP towards p38α, and also suggest a possible reason for the difference in affinity as measured using ITC⁴⁴.

Due to the central role of p38α in regulating the immune response, the design and development of a small molecule that selectively inhibits p38α MAP kinase has received significant attention as a therapeutic strategy for the treatment of inflammatory diseases^60–62. Over the past two decades, tremendous efforts have been focused on the discovery and development of small-molecule p38α MAPK inhibitors. The most promising approach is the development of linear binders, designed to align to the hinge region of p38α, which form a tight complementary surface with the hinge region of the enzyme and induce a glycine flip at Gly110 by a carbonyl oxygen atom of the respective ligand^63–65. However, despite numerous inhibitors with diverse structures and binding modes having been developed, to date only a few compounds are in clinical trials, highlighting the need for new approaches to design an isoform selective inhibitor. The present structure of p38α kinase in complex with pepHePTPm provides useful clues that could be exploited to design selective p38α inhibitors.

Methods

Materials

Tris, ATP, NADH, phospho(enol)pyruvate (PEP), lactate dehydrogenase (LDH), pyruvate kinase (PK), purine nucleotide phosphorylase (PNPase), and 2-amino-6-chloro-purine ribonucleoside, were purchased from Sigma (St. Louis, MO, USA). 3-(N-morpholino) propanesulfonic acid (MOPS), BSA, DTT, NaCl were from Amresco (Solon, OH, USA). EGF receptor peptide (KRELVEPLTPSGEAPNQALLR), pepHePTP (¹⁶RLQERRGSNVALMLDV³¹), pepHePTPm (¹⁶RLQERRGSNVALMLDC³¹), pepMKK3b (¹SKGKSKRKKDLRISCNSK¹⁸), pepMEF2A (¹RKPDLRVVIPPS¹²), pepMK2 (³⁷⁰IKIKKIEDASNPLLLKRRKKARALEAAALAH⁴⁰⁰) were commercially synthesized by Scilight Biotechnology (Beijing, China). 7-Methyl-6-thioguanosine (MESG) was synthesized from 2-amino-6-chloro-purine ribonucleoside as described previously⁶⁶. All other chemicals were local products of analytical grade used without further purification. Double-deionized water was used throughout.

Plasmids, protein expression, and purification

The cDNAs of human HePTP, mouse STEP, and rat PTP-SL (147-549) were generously provided by Dr. Tomas Mustelin (Burnham Institute) and Dr. Rafael Pulido (Instituto de Investigaciones Citológicas). The full-length human HePTP was amplified by standard PCR and cloned into vectors pET15b (Novagen) and pGEX-4T (GE Healthcare) using NdeI and XhoI restriction sites. Full-length mouse STEP and rat PTP-SL (147-549) were cloned into the pET15b vector with the restriction sites Nde I and Xho I. The recombinant proteins derived from pET15b or pGEX-4T carried an N-terminal His₆- or GST- tag, respectively. Bisphosphorylated MAP kinases (ERK2/pTpY, phosphorylated at T183 and Y185, or p38α/pTpY, phosphorylated at T180 and Y182) were obtained by coexpressing MAP kinases (ERK2 or p38α) with their constitutively active upstream kinases (MEK1EEΔN4 or MKK6EE) in vivo¹⁹. Both ERK2/MEK1EEΔN4 and p38α/MKK6EE were subcloned into the same pETDuet-1 plasmid (Novagen) containing an N-terminal His₆ tag. Mutants of HePTP were generated by site-directed mutagenesis using polymerase chain reaction and confirmed by DNA sequencing.

All proteins were expressed in Escherichia coli BL21 (DE3). Cells were grown in LB medium at 37 °C to an OD₆₀₀ of 0.6-0.8 and induced with 0.2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 20 °C for 14 h. Cells were harvested by centrifugation at 4000 rpm for 10 min at 4 °C. The cell pellets were resuspended in buffer containing 50 mM Tris (pH 8.0) and 300 mM NaCl. Cells were lysed by sonication in the presence of 1 mg/mL lysozyme and 0.1 mM phenylmethylsulfonyl fluoride (PMSF) at 4 °C. The lysates were clarified by centrifugation at 13,000 rpm for 40 min at 4 °C, and the supernatants were collected for purification. His₆- or GST- tagged proteins were purified using Ni-NTA (Qiagen, Hilden, Germany) or GS4B (GE Healthcare, Uppsala, Sweden) affinity columns, and eluted with 300 mM imidazole or 5 mM reduced glutathione. The eluted proteins were subsequently purified by Source 15Q ion-exchange chromatography and eluted with a NaCl gradient up to 1 M, followed by Superdex 200 Increase size-exclusion chromatography equilibrated with buffer containing 10 mM HEPES (pH 7.5), 150 mM NaCl, and 2 mM DTT, both performed on an AKTA FPLC system (GE Healthcare, Uppsala, Sweden) at 4 °C. Monophosphorylated ERK2/pT (ERK2 phosphorylated on Thr-183) and p38α/pT (p38α phosphorylated on Thr-180) were generated by mixing bisphosphorylated ERK2/pTpY or p38α/pTpY with GST-HePTP as described previously¹¹. The N-terminal His₆-tagged bisphosphorylated proteins (0.5 mg in 500 μL) were incubated with GST-HePTP (final concentration 1 μM) in standard buffer (100 mM MOPS, pH 7.0, 100 mM NaCl, and 10 mM MgCl₂) at 25 °C for 60 min. The monophosphorylated ERK2 or p38α proteins were separated from GST-tagged HePTP by a Ni-NTA column. All prepared proteins were assessed by SDS-PAGE and stained with Coomassie Brilliant Blue. Protein concentrations were determined spectrophotometrically based on theoretical molar extinction coefficients at 280 nm. All proteins were stored at −80 °C, and stocks for enzymatic assays were supplemented with glycerol to a final concentration of 20% (v/v).

Enzyme assays for MAPKs

The kinase activity of p38α/pT was measured spectrophotometrically using EGF receptor peptide as a substrate^41,54. This assay coupled the production of ADP with the oxidation of NADH through PK and LDH. Standard assays were performed with 43 nM p38α/pT and indicated amounts of EGF receptor peptide at 25 °C in 1.8 mL reaction mixture containing 50 mM MOPS (pH 7.0), 100 mM NaCl, 0.1 mM EDTA, 10 mM MgCl₂, 0.2 mM NADH, 1 mM PEP, 20 units/mL LDH, 15 units/mL PK, 1 mM ATP, and various concentrations of HePTP. The progress of the reaction was monitored continuously by following the formation of NAD⁺ at 340 nm, using a Lambda 45 spectrophotometer (PerkinElmer Life Sciences, Waltham, MA, USA) equipped with a magnetic stirrer in the cuvette holder. The concentrations of ADP formed in the p38α/pT-catalyzed reaction were determined using an extinction coefficient for NADH of 6220 cm^-1M^-1 at 340 nm.

The ATPase activity of p38α/pT and ERK2/pT was also measured by quantifying ADP production coupled to NADH oxidation as described above. Standard assays were carried out at 25 °C in 1.8 mL reaction mixture with assay buffer, various concentrations of ATP, and synthetic peptides or PTP proteins containing the KIM sequence as indicated in each figure. The reaction was initiated through the addition of 480 nM p38α/pT or ERK2/pT, followed by the addition of synthetic peptides or PTP proteins. The absorbance at 340 nm was measured to monitor the reaction progress. The concentrations of ADP generated in the p38α/pT- or ERK2/pT- catalyzed ATP hydrolysis reactions were calculated with an extinction coefficient for NADH of 6220 cm^-1M^-1 at 340 nm.

Western blotting

To evaluate the phosphorylation states of p38α/pT at the end of the reactions, p38α/pT (1 μM) was incubated with HePTP (5 μM) and 2 mM ATP at 25 °C in a buffer containing 50 mM MOPS (pH 7.0), 100 mM NaCl, and 10 mM MgCl₂, and 0.1 mM EDTA for 30 min. The samples were subjected to 12% SDS-PAGE and electrotransferred onto PVDF membranes, and then incubated with anti-phospho-Thr (1:1000 for immunoblotting, Cell Signaling Technology, #9381), anti-phospho-Tyr (1:1000 for immunoblotting, Cell Signaling Technology, #9411), and anti-bisphosphorylated p38 antibodies (1:1000 for immunoblotting, Cell Signaling Technology, #9211). Specific immunocomplexes were detected by chemiluminescence using ECL reagents.

Transfections, coimmunoprecipitation, and immunofluorescence

HEK293T cells (ATCC) were transfected with HA-tagged HePTP (HePTP/C270S and HePTP/ΔN43) and/or Flag-tagged p38α. After 48 h, cells were treated with 0.5 M sorbitol for 30 min. Cells were lysed in lysis buffer containing 20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% NP-40, 1 mM PMSF, 1 μg/mL aprotinin, 1 μg/mL leupeptin, 25 mM NaF, and 2 mM Na₃VO₄. The lysates were subjected to immunoprecipitation with protein A/G plus-agarose beads (Santa Cruz Biotechnology, Dallas, USA, cat. no. sc-2003) and anti-Flag (Sigma-Aldrich, St. Louis, MO, USA, cat. no. H9658) overnight at 4 °C. Immunoprecipitates were analyzed by 12% SDS-PAGE gel and electrotransferred onto polyvinylidene difluoride (PVDF) membranes (Merck Millipore, Darmstadt, Germany). After incubation of PVDF membranes with anti-bisphosphorylated p38 (Cell Signaling Technology, Danvers, MA, USA, cat. no. 9211), anti-phospho-Tyr (Cell Signaling Technology, cat. no. 9411), anti-Flag or HA antibodies (Sigma-Aldrich, cat. no. F3165, H9658) specific immunocomplexes were detected by chemiluminescence using ECL reagents (Beyotime, Shanghai, China).

COS-7 cells (ATCC) were transfected with p38α-GFP and/or HA-HePTP (HA-HePTP/C270S) expression vectors and processed for immunofluorescence analysis after 36 h of culture. Cells were rinsed with PBS (1.5 mM KH₂PO₄, 4.3 mM Na₂HPO₄, 137 mM NaCl, pH 7.4) and then fixed with methanol at −20 °C for 5 min. Samples were incubated in blocking solution (PBS containing 3% bovine serum) at room temperature for 15 min. HA‑HePTP (HA-HePTP/C270S)-expressing cells were then incubated with mouse anti‑HA antibody at 37 °C for 90 min, followed by incubation with Alexa Fluor 568‑conjugated goat anti‑mouse antibody for 1 h at room temperature, washed with PBS, and mounted for imaging. p38α-GFP was visualized directly by GFP fluorescence.

Gel filtration assay

Size exclusion chromatography was performed to determine the apparent molecular weight of p38α/pT alone, HePTP/ΔN43 alone, and a mixture of the two. Each sample (0.1 mL) was loaded onto a Superdex 200 column, pre-equilibrated in buffer containing 50 mM HEPES (pH 7.5), 150 mM NaCl, and 2 mM DTT. Chromatography was carried out at 4 °C with a flow rate of 0.5 mL/min. The column was calibrated using molecular weight markers. Fractions of 0.5 mL each were collected, and aliquots of relevant fractions were analyzed by SDS-PAGE and stained with Coomassie Brilliant Blue.

Enzyme assays for HePTP

The phosphatase activity of HePTP and its mutants toward the substrate p38α/pTpY was measured using a continuous spectrophotometric assay described previously¹¹. This assay employed a coupled enzyme system that uses PNPase to catalyze the transfer of phosphate to its chromogenic substrate MESG for monitoring the generation of inorganic phosphate in the phosphatase reaction. All kinetic measurements were conducted at 25 °C in 1.8 mL reaction mixture containing 50 mM MOPS (pH 7.0), 100 mM NaCl, 0.1 mM EDTA, 10 mM MgCl_2, 100 μM MESG, 0.1 mg/mL PNPase, and varying concentrations of p38α/pTpY. Reactions were initiated by adding 40 nM HePTP or indicated HePTP mutants to the reaction mixture, and the dephosphorylation of p38α/pTpY was monitored by recording the time courses of absorbance change at 360 nm using a Lambda 45 spectrophotometer. The change in absorbance reflects the conversion of MESG to 7-methyl-6-thioguanine in the presence of inorganic phosphate. Phosphate release was quantified using an extinction coefficient of 11,200 M^-1cm^-1 for the phosphate-dependent reaction at 360 nm⁴². Initial rates were determined from the linear slope of progress curves, and the data were analyzed using nonlinear regression. Kinetic parameters ($k_{cat}$ and $K_m$) were obtained by fitting the data to the Michaelis-Menten equation.

Crystallography

Purified p38α was mixed with HePTP peptides in a 1:3 molar ratio and subjected to crystallization trials. All protein complexes were crystallized by the vapor-diffusion technique in hanging drops, prepared by mixing equal volumes of protein and reservoir solution at 18 °C. Crystals of the p38α-pepHePTPm complex were obtained with a reservoir solution containing 20% PEG 3350 and 0.2 M sodium citrate. Fresh crystals were quickly transferred to a cryoprotectant buffer containing 20% glycerol and flash-frozen under liquid nitrogen. Diffraction data sets were collected at beamline 17U of the Shanghai Synchrotron Radiation Facility (SSRF, Shanghai, China) and processed using HKL2000⁶⁷. The structure was solved by molecular replacement using Phaser⁶⁸ with p38α (PDB: 5UOJ) as search model. Standard refinement was performed with REFMAC5⁶⁹ and Coot⁷⁰. Ramachandran analyses were carried out using MolProbity⁷¹, and all residues are in the favored and allowed regions of the Ramachandran plot. Data processing and refinement statistics were summarized in Table 1. The structural representations were prepared with PyMOL (http://www.pymol.org).

Molecular dynamics simulations and allosteric pathway analysis

The initial structure of p38α was obtained from the crystal structure (PDB: 5UBJ), and the initial structure of the p38α-HePTP complex was based on our resolved structure, with missing residues modeled in Coot. Molecular dynamics (MD) simulations for the apo p38α and p38α-HePTP complex models were carried out using explicit-solvent periodic boundary conditions prepared by CHARMM-GUI^72–74. Each model was solvated in a rectangular box with a 10 Å distance between the solutes and the sides of the solvent box. Potassium and chloride ions were added to neutralize the system and achieve a final salt concentration of 150 mM. The system minimization, equilibration, and production were carried out by GROMACS v2023 with the CHARMM36m force field. The solvated systems were first energy-minimized using the steepest descent algorithm until the maximum force on any atom was below 1000 kJ/mol/nm. And then, the systems were equilibrated under NVT conditions at 303.15 K using the velocity rescaling thermostat, followed by NPT conditions at 1 bar using Parrinello-Rahman barostat, for 100 ps each. Production MD simulations were then performed for 1 µs with a 2-fs time step under NPT conditions, maintaining a constant temperature of 303.15 K and a pressure of 1 bar. The particle mesh Ewald method (PME) was used for long-range electrostatic interactions, with a cutoff of 1.0 nm for van der Waals interactions and 1.4 nm for electrostatics. Trajectory snapshots were saved every 100 ps for subsequent analysis. All simulation analyses were implemented using Chimera 1.19⁷⁵ and VMD 1.9.4⁷⁶.

To identify potential allosteric communication pathways, residue interaction networks were constructed from the MD trajectory using MD-TASK⁷⁷. Residues were represented as nodes and edges were defined based on a heavy-atom distance cutoff of 7.0 Å. Edges with contact frequencies below the threshold of 0.6 were excluded to retain reliable interactions. Shortest paths between selected source and target residues were defined as those with the minimum number of edges in the unweighted network. When multiple shortest paths of equal length existed, the most probable pathway was selected by maximizing the product of contact frequencies along each path. This strategy prioritizes communication routes that are both topologically efficient (shortest in steps) and dynamically stable (highest frequency product). All analyses were performed using the NetworkX Python package.

Statistics and reproducibility

Kinetic data were globally fitted to a Michaelis-Menten model describing two-ligand binding using SigmaPlot (v12.0; Systat Software, CA, USA). Data in Figs. 1f and 2d are presented as mean ± SEM (n = 3), with dots representing individual experimental replicates and bars indicating the mean. All experiments were independently repeated.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.-Y.Z. and C.-J.P. contributed equally to this work. Y.-Y.Z. purified recombinant proteins and carried out in vitro biochemical assays. Y.-Y.Z. and C.-J.P. determined the crystal structure and performed data analysis with advice from Z.P.L. Z.W.Z. contributed to data analysis. Z.-X.W. and J.-W.W. conceived the project and wrote the manuscript together with all other authors.

Peer review

Peer review information

Communications Biology thanks Zheng Ruan, Ziyuan Jiang, and the other, anonymous, reviewer for their contribution to the peer review of this work. Primary Handling Editors: Dr. Janesh Kumar and Dr. Nilanjan Banerjee.

Funding

This work was supported in part by grants 2019YFA0802500 and 2022YFA0806500 from the Ministry of Science and Technology of China, the Project of Biomedical Basic Research Center (BBRC) of Jiangsu (BK20255001), and the Basic Research Program of Jiangsu (BK20243030).

Data availability

The atomic coordinates and structure factors for the p38α-pepHePTPm complex have been deposited in the Protein Data Bank with accession code 8YP8. All source data are provided as a supplementary data file (Supplementary Data 1).

Competing interests

The authors declare that they have no competing interests. Data and materials availability.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10502-7.