2026 Volume 26 Pages 1-10
All seven human mitogen-activated protein kinase kinases (MAP2Ks) are essential for cellular processes such as cell proliferation and apoptosis, and dysfunction of MAP2Ks is associated with cancers and autoimmune diseases. 5Z-7-oxozeaenol (5Z7O) strongly inhibits MAP2K1, 2, 3, and 6, and weakly inhibits MAP2K4, 5, and 7. In this study, the potencies of 5Z7O toward MAP2K2, 3, and 5 were explored by computational methods including docking and molecular dynamics (MD) simulations using homology models. Docking simulations showed that the α, β-unsaturated ketone moiety of 5Z7O bound close to the conserved cysteine residue located in front of the DFG motif (DFG-1) in MAP2K2 and 3 as in previous studies of MAP2K1 and 6; in strong inhibition, 5Z7O binds covalently with this cysteine. MD simulations of MAP2K2 and 3 in the apo state showed that the high flexibility of a conserved “gatekeeper” methionine residue enabled access of 5Z7O to this binding site. However, in docking simulation of MAP2K5, the α, β-unsaturated ketone moiety of 5Z7O was far from the conserved DFG-1 cysteine. A threonine residue at the gatekeeper position in MAP2K5 likely prevents access of the ketone moiety to that cysteine. These findings, together with previous data, provide guidance for the development of inhibitors selective for each type of MAP2K.
Mitogen-activated protein kinase kinases (MAP2Ks), of which there are seven in humans (MAP2K1–7), play crucial roles in essential cellular processes such as cell proliferation and apoptosis [1-3]. MAP2K1 and MAP2K2 are 87% identical in amino acid sequence in the kinase domain (KD) and biologically equivalent; MAP2K3 and MAP2K6 are 87% identical in amino acid sequence in the KD and physiologically, almost equivalent; MAP2K4 and MAP2K7 are 57% identical in amino acid sequence in the KD and activate some parts in common but different signaling pathways [1-3]. MAP2K5 is less like the other MAP2Ks and activates a unique pathway [1-3]. Dysfunction of any MAP2K is associated with serious diseases, including cancers and autoimmune diseases [4-7]. Therefore, MAP2Ks are regarded as important drug targets. However, high homology of amino acid residues among MAP2Ks makes it challenging to develop inhibitors that are highly selective for only one MAP2K.
5Z-7-oxozeaenol (5Z7O) (Figure 1) is a natural compound that possesses an α, β-unsaturated ketone moiety that forms a covalent bond with noncatalytic cysteine residues [8]. 5Z7O inhibits all MAP2Ks, but to different extents, with inhibitory concentrations in the nanomolar-to-micromolar range (Table 1) [9,10]. Four MAP2K crystal structures complexed with 5Z7O have been reported and registered in the Protein Data Bank (PDB): MAP2K1 (PDB ID: 8YP4), MAP2K4 (PDB ID: 8YP5), MAP2K6 (PDB ID: 9M1Z), and MAP2K7 (PDB ID: 3WZU) [9-11]. 5Z7O covalently bound to the conserved cysteine residue located in front of the DFG motif (DFG-1) in MAP2K1 and MAP2K6, but to a non-conserved cysteine in MAP2K7; it bound noncovalently in MAP2K4 [9-11]. Detailed structural comparison and molecular dynamics (MD) simulations of apo-MAP2K1, 4, 6, and 7 revealed that the high mobility of a conserved methionine residue at the gatekeeper position [12] was associated with the accessibility of 5Z7O to the ATP-binding site deep within the protein [9,10].
To extend the above analysis to all MAP2Ks and help develop highly selective inhibitors for each type of MAP2K, the structural basis of the selectivity of 5Z7O as an inhibitor of MAP2K2, 3, and 5 was explored here by computational methods, including docking and MD simulations using homology models.
Figure 1. 5Z-7-oxozeaenol (5Z7O)
Table 1. Inhibitory activities of 5Z7O toward human MAP2Ks and the positions of noncatalytic cysteine residues
| DFG-1 | αD helix | P-loop | A-loop | IC50 (μM) | |
|---|---|---|---|---|---|
| MAP2K1 | Cys207 | Ser150 | Gly79 | Thr226 | 0.08 |
| MAP2K2 | Cys211 | Ser154 | Gly83 | Thr230 | 0.06 |
| MAP2K3 | Cys207 | Ser156 | Thr75 | Cys227 | 0.005 |
| MAP2K4 | Cys246 | Ser184 | Tyr113 | Cys266 | 4.7 |
| MAP2K5 | Cys300 | Ser248 | Gly177 | Thr319 | 8.2 |
| MAP2K6 | Cys196 | Ser135 | Tyr64 | Cys216 | 0.008 |
| MAP2K7 | Cys276 | Cys218 | Cys147 | Cys296 | 1.3 |
IC50 values of 5Z7O for MAP2K1, 4, 6, and 7 have been reported previously [9,10,11], and those for MAP2K2, 3 and 5 are newly reported in this study.
2.1 Model preparation of MAP2K2, 3, and 5
The structural models of MAP2K2, 3, and 5 for MD and docking simulations were prepared by homology modelling using MODELLER [13] and the SWISS-MODEL web server (https://swissmodel.expasy.org). The sequences and template coordinates are shown in Table 2. To remove the bias in the bonding state of 5Z7O, we chose the structure of 5HZE instead of 8YP4 as a template for MAP2K2 homology modeling. Additionally, we selected 3SLS among the available MAP2K1 structures as a template for MAP2K5 homology modelling because the activation loop of 3SLS was different from typical MAP2K1 structure. We chose the MAP2K2 and MAP2K3 models with as high a molpdf score and as low a DOPE score as possible and the MAP2K5 model was generated automatically.
Table 2. Templates and sequences of MAP2K2, 3, and 5 used in homology modelling
| MAP2K | Method | Template | Sequence | aRMSD (Å) |
|---|---|---|---|---|
| MAP2K2 | MODELLER | 5HZE (MAP2K1) | baa 47–400 | 0.70 Å |
| MAP2K3 | MODELLER | 9M1Z (MAP2K6) | baa 49–347 | 0.18 Å |
| MAP2K5 | SWISS-MODEL | 3SLS (MAP2K1) | baa 166–448 | 0.10 Å |
aRMSD value is calculated between the template (query) and homology models. baa, amino acids.
2.2 Molecular dynamics simulations of MAP2K2, 3, and 5 in the apo state
We performed MD simulations of MAP2K2, 3, and 5 in the apo state, i.e., in the absence of inhibitor. The initial models of MAP2K2, 3, and 5 were obtained by homology modelling as described in section 2.1. For the simulation model of MAP2K2 in solution, a simulation box was constructed with a margin of 12 Å to the boundary of the box, resulting in dimensions 94 Å × 94 Å × 94 Å. The solution system contained 23,972 TIP3P water [14] molecules together with 71 potassium ions and 68 chloride ions corresponding to ~0.15 M ion concentration, resulting in 77,629 atoms in total. AMBER ff14SB was applied for the potential energy of the all-atom protein [15, 16]. The MD simulations were performed by AMBER20 [17] in constant temperature and pressure conditions at T = 300 K and P = 1 atm using Berendsen’s thermostat and barostat [18] at a relaxation time of 1 ps and using the particle mesh Ewald method for electrostatic interactions [19]. The simulation length was 1 μs, together with a 2-fs time step using constraining bonds involving hydrogen atoms via the SHAKE algorithm [20]. The coordinates were saved every 10 ps and used for analysis. Analysis of MD trajectories was performed by using the AMBER cpptraj module [21]. The MD simulations of MAP2K3 and MAP2K5 were also carried out as described above (the solution systems contained 43,705/48,039 atoms overall in a box of 73 Å × 84 Å × 91 Å/80 Å × 80 Å × 80 Å, respectively).
2.3 Docking simulations of 5Z7O
We performed docking simulations of 5Z7O into MAP2K2, 3, and 5 models using AutoDock Vina to obtain protein–ligand Michaelis complexes [22, 23]. The homology model structures of MAP2K2 and MAP2K3 were used as receptor models for docking. These models are reliable because the template and model are highly homologous (see Introduction). Homology modeled MAP2K5 was obtained using MAP2K1, which is 50% identical in amino acid sequence to MAP2K5, as the template. The lower homology between the template and the model suggests that this model structure may be unreliable. Hence, we performed MD simulations of apo-MAP2K5 to obtain diverse structures for receptor models. Finally, receptor models of MAP2K5 were obtained by k-means clustering of whole 1-µs MD trajectories by the cpptraj module, which resulted in 10 representative structures [21]. Searching for docking sites of 5Z7O was conducted within a cubic grid box of 25 Å on each side, centered on the DFG-1 cysteine residue.
2.4 Enzyme-linked immunosorbent assays of MAP2K2, 3, and 5
Enzyme reactions were conducted in solution containing activated MAP2Ks (Carna Biosciences), unphosphorylated glutathione-S-transferase-tagged substrate protein (ERK2 for MAP2K2, p38α for MAP2K3, and ERK5 for MAP2K5), Mg2+, and ATP, at room temperature. Assays were conducted at ATP concentrations near the corresponding Km values: 15 µM for MAP2K2, 0.5 µM for MAP2K3, and 1 µM for MAP2K5, respectively ( www.carnabio.com). Phosphorylated products were detected using respective anti-phospho-substrate monoclonal antibodies as primary antibodies. The immunocomplexes were quantified by measuring absorbance at 450 nm using the chromogenic substrate 3,3′,5,5′-tetramethylbenzidine. The IC₅₀ values of 5Z7O for each MAP2K were determined by fitting a four-parameter logistic curve to the plot of percentage inhibition vs. the logarithmic concentration of 5Z7O, using nonlinear least squares regression. The IC₅₀ values determined for MAP2K2, 3, and 5 are reported in Table 1.
The binding mode between 5Z7O and MAP2K2 with the highest score in the docking calculation was almost the same as that between 5Z7O and MAP2K1, a physiological equivalent of MAP2K2, observed in the crystal structure [9] (Figure 2a). The β-carbon atom of the α, β-unsaturated ketone moiety of 5Z7O was sufficiently proximal to the Sγ atom of Cys211 at the DFG-1 position (distance = 3.5 Å) to form a covalent bond. This indicates the covalent bond formation between these atoms, as observed in the crystal structure of the 5Z7O/MAP2K1 complex [9]. Additionally, the amino acid residues that participate in interaction with 5Z7O are fully conserved between MAP2K1 and MAP2K2 (Figure 3). These observations are consistent with the fact that the inhibitory activities of 5Z7O toward MAP2K1 and MAP2K2 are similar (Table 1). Furthermore, manual observation of the MD trajectory of apo MAP2K2 along with the changes in RMSD values revealed that gatekeeper residue Met147 were classified into three conformations: MD-1, MD-2, and MD-3 (Figure 4a). They mainly adopted two configurations (MD-1/MD-3 and MD-2) (Figure 4a), as also shown in previous study of apo-MAP2K1 (Figure 4a) [9]. Considering the crystal structure of the 5Z7O/MAP2K1 complex, 5Z7O seems to selectively bind to the MD-2 state defined in Figure 2a and form the covalent adduct but not to MD-1 and MD-3 because of steric hindrance with 5Z7O [9].
The binding mode between 5Z7O and MAP2K3 with the highest score in the docking calculation was almost the same as that between 5Z7O and MAP2K6, a physiological analogue of MAP2K3, observed in the crystal structure of the 5Z7O–MAP2K6 complex [10] (Figure 2b). The distance between the β-carbon atom of the α, β-unsaturated ketone moiety of 5Z7O and the Sγ atom of Cys207 at the DFG-1 position of MAP2K3 (4.2 Å) was slightly longer than that observed in the 5Z7O/MAP2K2 model. However, this distance is reminiscent of the formation of a covalent bond between 5Z7O and Cys207 from this Michaelis complex. Additionally, the amino acid residues that participate in interaction with 5Z7O are fully conserved between MAP2K3 and MAP2K6 (Figure 3). These observations are consistent with the fact that the inhibitory activities of 5Z7O toward MAP2K3 and MAP2K6 are similar (Table 1). In addition, manual observation of the MD trajectory of apo MAP2K3 along with the changes in RMSD values revealed that gatekeeper residue Met140 mainly adopts two configurations (Figure 4b), as also observed in MD simulation of apo-MAP2K6 [10]. Referring to the crystal structure of the MAP2K6/5Z7O complex [10], 5Z7O seems to selectively bind to the MD-1 state of MAP2K3 and form the covalent adduct but not to MD-2 and MD-3 (Figure 4b). Thus, as in MAP2K1 and MAP2K6, there is a conformation of GK in which 5Z7O can access the DFG-1 cysteine of MAP2K2 and MAP2K3.
Figure 2. Close up view of docking simulation results of 5Z7O in ATP-binding sites of MAP2K2 and MAP2K3.
(a) Michaelis complex of MAP2K2 with 5Z7O and superposition of 5Z7O binding mode in MAP2K1 and MAP2K2 (cyan: MAP2K1, blue: MAP2K2). (b) Predicted Michaelis complex of MAP2K3 with 5Z7O and superposition of 5Z7O binding mode in MAP2K3 and MAP2K6 (purple: MAP2K3, yellow: MAP2K6). Proteins and 5Z7O are represented in ribbon and stick models, respectively.
Figure 3. Multiple sequence alignment of MAP2K family members and typical kinase fold with region representation
(a) Multiple sequence alignment of MAP2K family. P-loop is the loop that binds the phosphate group of ATP, GK is the gatekeeper, and HRD is the conserved sequence in the catalytic loop [12]. (b) Typical kinase fold with the presentation for the regions important for 5Z7O binding. Protein structure is shown by white ribbon model. P-loop, Hinge, DFG, αD helix, HRD, 310 helix and DFG are shown by green, blue, cyan, orange, yellow and salmon pink, respectively. GK, αC-β4 loop and DFG-1 are shown by black circles.
Figure 4. Root-mean-square deviation (RMSD) from homology models and representative conformations of the “gatekeeper” methionine residue in MAP2K2 and MAP2K3
(a) Molecular dynamics (MD) simulation result for MAP2K2. (b) MD simulation result for MAP2K3. The gatekeeper methionine residues of MAP2K2 and MAP2K3 are represented as blue and purple stick models, respectively. Reference structures are represented as white stick models.
5Z7O was docked into clustered apo-state MAP2K5 models derived from MD simulations (Figure 5). All the poses were different from those for the other MAP2Ks. The β-carbon atom of the α, β-unsaturated ketone moiety of 5Z7O was not proximal to Cys300 at the DFG-1 position of MAP2K5. This indicates that a covalent bond does not form between 5Z7O and the DFG-1 cysteine, explaining the relatively low inhibitory activity of 5Z7O toward MAP2K5 (Table 1); similar was observed for MAP2K4 and MAP2K7 [9]. Superimposition of the MAP2K5 model and the crystal structure of the MAP2K1/5Z7O complex suggests that the unique residue Ile225 in the αC–β4 loop and gatekeeper residue Thr241 of MAP2K5 prevent the binding of 5Z7O in a similar manner to that in MAP2K1 (Figures 3, 6). The additional methyl group of Ile225 compared with Val127 in MAP2K1 and the β-branched methyl group of Thr241 likely create repulsive interactions with 5Z7O if 5Z7O is oriented in MAP2K5 in the position in which it binds to MAP2K1 (Figure 6). Therefore, these unique residues of MAP2K5 seem key to the difference in specificity of 5Z7O as an inhibitor of this enzyme compared with other MAP2Ks.
Figure 5. Close up views of the binding mode of 5Z7O in 10 MAP2K5 structures generated by k-means clustering of MD trajectories
MAP2K5 is represented by orange ribbon models; 5Z7O is shown in cyan stick models.Relative position of DFG-1 cysteines is shown in green circles.
Figure 6. Superposition of MAP2K1/5Z7O complex and apo-MAP2K5
MAP2K5 is represented by orange ribbon models. 5Z7O is shown using cyan stick models. Residues located in deep ATP-binding sites of MAP2K1 and MAP2K5 are represented in cyan and orange stick models, respectively. Distances between the atoms indicated by black dotted lines are also represented.
In this study, molecular modeling, MD calculations, and docking simulations of MAP2K2, 3, and 5 with 5Z7O were performed. Docking poses of 5Z7O toward MAP2K2 and MAP2K3 were nearly identical to the binding modes in the closest homolog; MAP2K1 and MAP2K6, respectively. Docking poses of 5Z7O toward MAP2K5 were dissimilar to that in MAP2K1/5Z7O crystal structure. In addition, molecular dynamics simulation of apo MAP2K2 and MAP2K3 revealed that gatekeeper methionine residue tends to adopt two conformations. These results are essentially consistent with the structural basis for the development of specific inhibitors obtained by X-ray crystallography and MD calculations for MAP2K1, 4, 6, and 7 in previous studies [9,10,11]. Our findings will reinforce the overall structural basis for developing selective inhibitors for each MAP2K by filling the gap in the structural information.
Combining the structural insights from this and previous studies [9,10,11], we propose a structure-based strategy for producing highly selective inhibitors for each MAP2K (Figure 7). As a first step, the seven MAP2Ks are divided into two groups mainly by the difference at the gatekeeper position (Figures 3, 7): MAP2K5 has threonine at the gatekeeper position, while the other MAP2Ks have methionine. The latter MAP2Ks are further divided into two groups based on the structural dynamics of the gatekeeper methionine (Figure 7) [9, 10]. In MAP2K4 and MAP2K7, the gatekeeper methionine tends to be fixed in a position that interferes with covalent bonding with the DFG-1 cysteine residue, while in MAP2K1, 2, 3, and 6, this methionine adopts the conformation allowing 5Z7O to bind covalently with the cysteine. MAP2K4 and MAP2K7 can be discriminated by difference in the first residue of the αD helix; Ser184 in MAP2K4 and Cys218 in MAP2K7 (Figures 3, 7) [11, 24]. MAP2K1, 2, 3, and 6 can be divided into two groups (MAP2K1 and 2, and MAP2K3 and 6, respectively) based on the conformational diversity of the conserved arginine residue in the HRD motif and difference in the P-loop hydrogen bond donor (Figures 3, 7) [10]. Physiological similarities make it less necessary to distinguish MAP2K1 from MAP2K2, or MAP2K3 from MAP2K6 [1-3]. Taken together, our structural insights are expected to promote the development of selective inhibitors for each type of MAP2K.
Figure 7. Schematic classification of MAP2Ks for production of highly selective inhibitors
All MAP2Ks are represented by ribbon models. MAP2K1, 2, 3, 4, 5, 6, and 7 are shown in cyan, blue, purple, green, orange, yellow, and magenta, respectively.
This work was supported by JST, the Establishment of University Fellowships Towards the Creation of Science Technology Innovation, grant number JPMJFS 2138. We acknowledge the support of the Research Support Project for Life Science and Drug Discovery (BINDS) from AMED under grant number JP25ama121023. We thank Edanz Inc. (https://jp.edanz.com/ac) for editing a draft of this manuscript.
