JNK-IN-8 – Bay63 2521 Chemical

In silico-in vitro discovery of untargeted kinase–inhibitor interactions from kinase-targeted therapies: A case study on the cancer MAPK signaling pathway

Li Menga, Zhijun Huangb
a College of Pharmaceutical Sciences, Jiangsu Vocational College of Medicine, Yancheng 224008, China
b Department of General Surgery, Yancheng First People’s Hospital, Yancheng 224005, China

A B S T R A C T
Protein kinase inhibitors have been widely used as therapeutic agents to treat a variety of diseases, but many of them may cause oﬀ-target eﬀects by unexpectedly targeting other noncognate kinases due to high conversion across the protein kinase family. The mitogen-activated protein kinase (MAPK) signaling pathway plays an essential role in tumorigenesis, which has been recognized as a high priority in the druggable target candidates of anticancer therapy. Here, we attempt to investigate the untargeted kinase–inhibitor interactions (UKIIs) of kinase-targeted therapies for the cancer MAPK signaling cascade via an integration of biomolecular modeling, cell viability assay and kinase inhibition analysis. A systematic kinase–inhibitor interaction proﬁle is created for 28 FDA-approved kinase inhibitor drugs across 9 caner-related MAPK kinases. The created proﬁle is analyzed at structural, energetic and dynamic levels and, consequently, totally 18 promising UKII pairs with high theoretical aﬃnity are derived, from which the noncognate inhibitors Cabozantinib, Regorafenib and Crizotinib are selected to test their cytotoXic eﬀects on human epithelial colorectal adenocarcinoma Caco-2 cell line and inhibition activity against the recombinant protein of human p38α kinase domain. The obtained results are compared with two cognate MAPK inhibitors JNK-IN-8 and BIRB796. As might be expected, the Regorafenib, Crizotinib and Cabozantinib exhibit high, moderate and low cytotoXicities, respectively. In addition, the Regorafenib is de- termined to have a potent p38α-inhibitory activity. This is basically in line with the test results of positive controls JNK-IN-8 and BIRB796 and can be well conﬁrmed by computational modeling.

1. Introduction
Mitogen-activated protein kinase (MAPK) signaling composes a fa- mily of serine/threonine protein kinases that transduce extracellular signals from the cell membrane to the nucleus via a cascade of phos- phorylation events (Son et al., 2011), which participates in a diverse array of cellular processes, including cell growth, division, movement and death (Schaeﬀer and Weber, 1999). Abnormalities in the signaling impinge on most, if not all these processes, and play a critical role in the development and progression of cancer (Dhillon et al., 2007); its aberrant activation has been implicated in numerous cancers such as colorectal cancer, ovarian cancer, breast cancer and lung cancer, making it an attractive target for antitumor therapy (Germann et al., 2017). The MAPK has also been reported to drive acquired drug re- sistance in melanoma through mutation, ampliﬁcation or over- expression of genes encoding the pathway regulators by stimulating ERK activation through alternative routes (Wellbrock, 2014; Lu et al.,2017). Protein kinases are a family of enzymes that catalyze the transfer of phosphate groups from high-energy, phosphate-donating molecules to speciﬁc substrates (Hanks et al., 1988). A majority of kinases share homologous sequence, analogous folding and similar function. In par- ticular, their catalytic active sites are highly conserved so that many kinase inhibitors exhibit low selectivity and broad speciﬁcity that may cause oﬀ-target eﬀects by unexpectedly targeting noncognate kinases (Zhang and Loughran, 2011). The oﬀ-target eﬀects have been observed as an untargeted consequence of targeted therapies with kinase in- hibition (Alemán et al., 2014). On the other hand, the oﬀ-target eﬀects associated with kinase inhibitors can be exploited as new and potential agents to treat other diseases, namely, new uses for old drugs (Sachs et al., 2017). Previously, Cui et al. performed structure-based grafting and modiﬁcation to identify the noncognate kinase inhibitors of mTOR signaling pathway as potential therapeutics for glioblastoma (Cui et al., 2015). In addition, commercially available kinase inhibitors have also been successfully used to target TGF-β and c-Abl for the treatment of scleroderma and Alzheimer’s disease, respectively (Cong et al., 2014; Zhu et al., 2014). Computational discovery of novel druggable protein targets attract great interest in the drug design community (Pennisi et al., 2016; Pappalardo et al., 2016), and inverse in silico screening has been successfully described to identify potential kinase inhibitor targets (Zahler et al., 2007). In this study, we created a systematic interaction proﬁle between all the 9 cancer-related MAPK kinases and 28 FDA- approved kinase inhibitors (Wu et al., 2016), by using a synthetic strategy that integrated in silico modeling (ligand grafting, molecular docking, dynamics simulation and aﬃnity scoring) and in vitro analysis (cytotoXic assay and kinase assay). The created kinase–inhibitor inter- action proﬁle was also examined in detail regarding their complex structures and binding energetics.

2. Materials and methods
2.1. MAPK family kinases
The human MAPK family comprises four main subgroups of pro- teins: extracellular signal-related kinases ERK1/2, ERK5, p38α/β/γ/δ and c-Jun amino-terminal kinases JNK1/2/3 (Séverin et al., 2010); they share a similar kinase domain at N-terminus but possess diﬀerent ad- ditional portions at C-terminus. Most (9 kinases) of these MAPK members are expressed in normal tissues and diverse tumors (Table 1), whereas only JNK3 is found speciﬁcally in brain and, in few cases, heart and testis (Bode and Dong, 2007). Therefore, the JNK3 was excluded from this study as we herein primarily concerned cancer MAPK kinases. All these investigated MAPK kinases, except p38γ, are currently available to their crystal structures in complex with small-molecular inhibitors in the protein data bank (PDB) database (Berman et al., 2000). For the p38γ, it has been cocrystallized with its speciﬁc substrate PTPN3, but not inhibitor ligand, by Chen et al. (2014). However, this p38γ–substrate complex is cooperated to promote Ras-induced onco- genesis and thus can be considered as a good start in structure-based drug design targeting the p38γ–PTPN3 interaction for anticancer ther- apeutics. Next, the cocrystallized inhibitors or substrate were stripped from these curated crystal complex structures, which were then mini- mized with 3Dreﬁne server (http://sysbio.rnet.missouri.edu/3Dreﬁne) (Bhattacharya et al., 2016) to eliminate bad atomic contacts and overlaps in the kinase domains. This method utilizes iterative optimi- zation of hydrogen bonding network combined with atomic-level en- ergy minimization on the optimized model using a composite physics and knowledge-based force ﬁelds for protein structure reﬁnement.

2.2. FDA-approved kinase inhibitors
Various kinase inhibitors have been developed to target a variety of protein kinases. Here, we only considered those existing inhibitors that have been licensed to the market. This is because the marketed drugs underwent rigorous evaluation and have been demonstrated to be safe and eﬃcient. In addition, these inhibitor drugs are readily commer- cially available. Currently, a total of 28 small-molecule kinase inhibitor drugs (Fig. 1 and Table 2) were retrieved from the approval list of US Food and Drug Administration (FDA), which are actively pursued as promising targeted therapeutics (Wu et al., 2015; Wu et al., 2016). These inhibitor compounds are diverse in terms of their chemical structures, including quinoline, indole, imidazole, carboXamide, qui- noXaline, etc; they are also diverse in terms of their cognate kinases and treated diseases, such as EGFR (lung cancer), VEGFR (thyroid tumor), BCR-ABL (breast cancer), B-Raf (melanoma) and JAK (myeloﬁbrosis and rheumatoid arthritis) (Levitzki, 2013; Xu et al., 2016). Some drugs (e.g. Regorafenib and Sorafenib) are pan-kinase inhibitors that exhibit a broad-spectrum eﬀect on an array of kinases, whereas some others (e.g. Vemurafenib and Trametinib) are highly speciﬁc that can even selec- tively inhibit disease-related kinase mutants (Anastassiadis et al., 2011; Davis et al., 2011).

2.3. Modeling noncognate MAPK kinase–inhibitor complex structures
Twenty-ﬁve out of the 28 kinase inhibitor drugs were reported to co-crystallize with their cognate kinase targets, and the crystal complex structures can be retrieved from the PDB database (Berman et al., 2000) as templates to model the complex structures of these inhibitor ligands with their noncognate MAPK kinase receptors by using a ligand grafting strategy (Cui et al., 2015). For other three inhibitor drugs (Regorafenib, Cabozantinib and Trametinib) with no available crystal structures, molecular docking calculations (Meng et al., 2011) were employed to predict their binding modes to noncognate MAPK kinases. Conse- quently, a total of 252 (28 × 9) noncognate MAPK kinase–inhibitor complex structures were obtained, which represent a systematic com- bination between the 28 kinase inhibitor drugs and 9 cancer MAPK kinases.
2.3.1. Ligand grafting
For the 25 inhibitors co-crystallized with their cognate kinase tar- gets (see Table 2), their binding modes can be directly ‘grafted’ from the cognate crystal complex templates into the active sites of noncognate MAPK kinases (Cui et al., 2015). The grafting strategy should be rea- sonable if considering that the inhibitor-bound active sites are basically conserved across kinase family members. For example, the schematic representation of using ligand grafting strategy to model the non- cognate p38α–Imatinib complex is shown in Fig. 2. First, the p38α crystal structure (PDB: 5WJJ) is superposed onto Abl–Imatinib complex crystal structure (PDB: 2HYY) by using Swiss-PdbViewer (Guex, 1996) to generate the superposed system of Abl/Imatinib/p38α. Second, the Abl is removed from the superposed system to obtain the modeled p38α–Imatinib complex structure.

2.3.2. Molecular docking
For other 3 kinase inhibitors with no available co-crystallized structures, molecular docking was employed to predict their binding modes to MAPK kinases. Before docking calculations the protonation state was added to kinase proteins using H++ server (http:// biophysics.cs.vt.edu) (Gordon et al., 2005), respectively, respectively. The docking calculations were performed with AutoDock Vina (Trott and Olson, 2010) to generate a number of potential complex models for a kinase–inhibitor interaction, from which we manually selected one of the most promising models in terms of their ranked docking scores and, more importantly, the binding mode of kinase cocrystalized with a chemically similar inhibitor. For example, the Regorafenib is a chemical analog of Sorafenib, and therefore the complex crystal structure of Sorafenib with its cognate kinase B-Raf (PDB: 5HI2) can be used to help the selection of promising one from the docking-generated complex models of Regorafenib with MAPK.

2.4. Molecular dynamics simulation and binding aﬃnity scoring
2.4.1. Dynamics simulation
The 252 noncognate MAPK kinase–inhibitor complex structures modeled by ligand grafting or molecular docking were separately sub- jected to molecular dynamics (MD) simulations using Amber ﬀ03 force ﬁeld (Duan et al., 2003) implemented in the AMBER14 package (Case et al., 2005). The general amber force ﬁeld (GAFF) (Wang et al., 2004) was applied to compound ligands. Partial charges of ligand atoms were determined with the RESP ﬁtting technique (Bayly et al., 1993). An implicit generalized Born (GB) model (igb = 2) (Tsui and Case, 2000a) was used to describe solvent eﬀect. Each complex was subjected to 10- ns MD production simulations in an isothermal isobaric ensemble (Zhou et al., 2016). A cut-oﬀ distance of 10 Å was applied for short-range electrostatics and van der Waals interactions, while the particle mesh Ewald method (Darden et al., 1993) was used to calculate long-range interactions. The SHAKE algorithm (Ryckaert et al., 1977) was em- ployed to constrain hydrogen-involving bonds.
2.4.2. Aﬃnity scoring
For each kinase–inhibitor complex, a total of 100 conformational snapshots were collected from the last 5-ns dynamics trajectory (Jiang et al., 2017), which were then used to derive the binding free energy of inhibitor ligand to kinase receptor by using molecular mechanics/ generalized Born surface area (MM/GBSA) analysis (Mikulskis et al., 2014). The kinase–inhibitor binding free energy ΔG (Score) consists of interaction potential and desolvation penalty, and the latter contains polar and nonpolar contributions. The polar aspect was described by GB model with the parameters developed by Tsui and Case (2000b). The nonpolar contribution was estimated using an empirical surface area model (Hou et al., 2009).

2.5. Cytotoxic assay
The cell-suppressing eﬀects of three investigated kinase inhibitors Cabozantinib, Regorafenib and Crizotinib as well as two positive con- trols JNK-IN-8 (pan-JNK inhibitor) and BIRB796 (pan-p38 inhibitor) on human epithelial colorectal adenocarcinoma Caco-2 cell line (ATCC HTB-37) were tested using a protocol described previously (Jindal et al., 2015; Lum et al., 2015). Brieﬂy, the Caco-2 cell line was main- tained in Dulbecco’s modiﬁed eagle medium (DMEM) containing 10% fetal calf serum at 37 °C. The cells seeded in 96-well microplates were treated with inhibitors at diﬀerent concentrations. Plates were in-cubated with 5% CO2 for 24 h. Cell viability was measured by using cell proliferation assay using an amicroplate reader: a/b × 100%, where a and b are the absorbance treated with inhibitor and medium, respec- tively.

2.6. Kinase assay
The p38 kinase assay described previously (Wadsworth et al., 1999; Underwood et al., 2000) was used to determine the inhibitory activity of Regorafenib as well as positive control BIRB796 and negative control Cabozantinib against the GST-tagged recombinant protein of human p38α kinase domain. Brieﬂy, p38α proteins were incubated in a reac- tion buﬀer (25 mM HEPES, pH 7.5, 10 mM MgCl2, 1 mM dithiothreitol and 0.1% BSA) containing 50 μM ATP (2 μCi [γ-32P]ATP) and 1 mM myelin basic protein as substrate. The phosphorylated substrate was captured on a phosphocellulose 96-well plate and counted in a Beck- menCoulter liquid scintillation counter. IC50 was deﬁned as the con- centration of the test compound that caused a 50% decrease in the maximal level of inhibition of kinase activity and was calculated from replicate curves using GraphPad Prism.

3. Results and discussion
3.1. Sequence homology analysis of MAPK family kinases
The MAPK family kinase members can be classiﬁed into four sub- groups, namely, ERK1/2, ERK5, p38α/β/γ/δ and JNK1/2/3, where the JNK3 was not considered in this study as we herein primarily concerned cancer MAPK kinases. Here, the primary sequences of ERK1/2/5, p38α/ β/γ/δ and JNK1/2 kinase domains were retrieved from the UniProt database (UniProt, 2015) and then compared to each other by MView multiple sequence alignment (Brown et al., 1998). As shown in Fig. 3, ERK1 and ERK2 share a high homology with sequence identity of 88.9% between them, but the ERK5 has only a low conservation compared to ERK1/2, with identity < 20%. Similarly, the p38α, p38β and p38γ can be attributed to a congeneric panel with identity > 60%, whereas the p38δ is speciﬁc relative to p38α/β/γ with identity < 30%. The JNK1 and JNK2 possess a high conservation with identity of 87.8%. The se- quence alignment revealed a varied homology among these MAPK ki- nases and small-molecule inhibitors would therefore be expected to exhibit a distinct selectivity proﬁle across the kinase family. 3.2. Binding mode comparison of co-crystallized, grafted and docked inhibitor In order to examine the accuracy and reliability of ligand grafting and molecular docking in predicting kinase–inhibitor binding modes, we herein compared the complex structures of p38α receptor with in- hibitor ligand Imatinib obtained from X-ray crystallography as well as computational modeling with ligand grafting and molecular docking. The two computational methods have been successfully applied for the structural modeling of kinase–inhibitor complexes previously. Fortunately, the noncognate complex structure of p38α kinase domain co-crystallized with Imatinib has already been solved by Namboodiri et al. (2010) at a high resolution (PDB: 3HEC), which was used as ex- perimental reference for the computational modeling. The co-crystal- lized, grafted and docked binding modes of Imatinib were superposed into p38α active site. As shown in Fig. 4, the Imatinib ligand exhibits an extended bound conformation in p38α active site and is partially out of the site. The inhibitor binding modes predicted by ligand grafting and molecular docking are highly consistent with that in co-crystallized structure, indicating that the two methods can well reproduce the crystallographic interaction between p38α and Imatinib. As might be expected, the grafted binding mode appears to be more accurate as compared to docked mode, with their root-mean-square deviations (rmsd) of 0.15 and 0.28 Å relative to the cocrystallized conformation of Imatinib, respectively. This is expected if considering that many active inhibitors can bind to diﬀerent protein kinases in a roughly similar manner (Patel and Doerksen, 2010; Zhou et al., 2013) and the prior knowledge utilized in ligand grafting would eﬀectively improve the method’s reliability. 3.3. Creating a systematic (noncognate) kinase–inhibitor interaction proﬁle The structures of 9 MAPK kinase domains were either retrieved from crystallographic data or modeled by virtual residue mutagenesis. Ligand grafting and molecular docking were employed to predict the binding modes between 9 MAPK kinases and 28 inhibitor compounds, totally resulting in 252 noncognate kinase–inhibitor complexes. Each complex was then subjected to 10-ns MD simulations and, based on hundreds of conformational snapshots extracted from the dynamics trajectory (Bai et al., 2017), an aﬃnity score matriX of 28 inhibitors across 9 kinases was created by MM/GBSA calculations. The Score is a physical quantity that characterizes the change in Gibbs free energy (ΔG) upon the kinase–inhibitor binding and, thus, its negative and positive values represent favorable and unfavorable binding interac- tions, respectively (Yang et al., 2015, 2016). The aﬃnity score matriX is visualized as a heatmap that char- acterizes the systematic (noncognate) kinase–inhibitor interaction proﬁle (Fig. 5). Evidently, most intermolecular interactions show a moderate or weak aﬃnity (Score < − 15 kcal/mol, highlighted by green and black colors); this is expected if considering that these compounds were originally not developed as the cognate inhibitors of MAPK kinases. However, there are also few interaction pairs with strong aﬃnity (Score > − 15 kcal/mol, highlighted by red color). As might be expected, many inhibitors exhibit a consistent binding cap- ability with the kinase conservation described in Fig. 3, that is to say, homologous kinases can generally be bound (or unbound) with the same inhibitors. For example, the Crizotinib and Ceritinib were pre- dicted as a dual-binder of JNK1 and JNK2 (Score = − 19.3/−17.7 and −17.0/−18.5 kcal/mol); the two kinases belong to the JNK subfamily and share a high homology between them (identity = 87.8%). Inter- estingly, a previous study found that the Imatinib, a cognate BCR-ABL inhibitor used to treat leukemia, can directly bind and inhibit non- cognate p38α kinase with a moderate activity (Kd = 34 μM and IC50 = 70 μM) (Namboodiri et al., 2010). This is basically line with the created proﬁle that the p38α–Imatinib interaction was calculated to possess a potent aﬃnity (Score = − 16.3 kcal/mol). Here, totally 18 promising UKII pairs with high theoretical aﬃnity (Score > − 15 kcal/ mol) were derived from the proﬁle, that is, ERK2–Bosutinib (−16.0 kcal/mol), ERK1–Cabozantinib (−18.9 kcal/mol), ERK2–Cabozantinib (−17.0 kcal/mol), ERK2–Ibrutinib (−15.8 kcal/ mol), p38α–Imatinib (−16.3 kcal/mol), p38β–Imatinib (−19.9 kcal/ mol), p38α–Imatinib (−18.6 kcal/mol), p38γ–Imatinib (−16.8 kcal/ mol), p38α–Regorafenib (−21.8 kcal/mol), p38β–Regorafenib (−15.8 kcal/mol), p38γ–Regorafenib (−16.6 kcal/mol), p38α–Da- brafenib (−17.7 kcal/mol), p38β–Dabrafenib (−19.8 kcal/mol), p38γ–Dabrafenib (−15.7 kcal/mol), JNK1–Crizotinib (−19.2 kcal/ mol), JNK2–Crizotinib (−17.7 kcal/mol), JNK1–Ceritinib (−17.0 kcal/ mol) and JNK2–Ceritinib (−18.5 kcal/mol).

3.4. Untargeted cytotoxic eﬀect of noncognate inhibitors on caco-2 cells
It is known that the MAPK signaling pathway plays an important role in the proliferation, migration, invasion and malignancy of tumors (Dhillon et al., 2007). Thus, suppression of the pathway would address cytotoXicity on cancer cells; this could be an untargeted consequence of the targeted therapies by these noncognate kinase inhibitors. Here, we compared the cell viability of human epithelial colorectal adenocarci- noma Caco-2 treated with three inhibitor compounds Cabozantinib, Regorafenib and Crizotinib as well as two known MAPK inhibitors JNK- IN-8 and BIRB796 at diﬀerent concentrations. Activation of the MAPK signaling has been reported to play an essential role in the maintenance in Supporting information, Table S1. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this concentration ≥10 μM. By comparison, the putative p38 inhibitor Re- gorafenib exhibits a similar cell-suppressing proﬁle with that of the cognate p38 inhibitor BIRB796, imparting an analogous cytotoXicity mechanism of the two kinase inhibitors at molecular level. In addition, the putative noncognate JNK inhibitor Crizotinib and known cognate JNK inhibitor JNK-IN-8 also display a roughly consistent proﬁle, al- though JNK-IN-8 possesses a relatively higher toXicity than Crizotinib against Caco-2. For the putative ERK1/2 inhibitor Cabozantinib, it seems to have only a modest eﬀect on Caco-2, suggesting that the noncognate inhibition of ERK kinases may not inﬂuence the cell via- bility substantially.

3.5. Untargeted inhibitory activity of noncognate inhibitors against p38α kinase
According to above computational analysis and cell viability assay, the Regorafenib was considered as a potent noncognate inhibitor of p38α kinase with a high aﬃnity Score of −21.8 kcal/mol, which can eﬀectively suppress the Caco-2 colorectal cancer cell viability with a similar proﬁle with the cognate p38α inhibitor BIRB796. In fact, the Regorafenib has been shown to have anti-angiogenic activity due to its targeted VEGFR2 tyrosine kinase inhibition (Wilhelm et al., 2011). In addition, the noncognate inhibitor Cabozantinib of ERK1/2 kinases was estimated to have only a modest aﬃnity (Score = −5.7 kcal/mol) for p38α kinase and exhibits a low cytotoXicity in Caco-2 viability assay. Therefore, the Regorafenib, BIRB796 (positive control) and Cabo- zantinib (negative control) were selected here to measure their in- hibitory activity targeting the recombinant protein of human p38α ki- nase domain (Table 3). The BIRB796 was determined to have a high potency against p38α (IC50 = 79 nM), which is basically in line with its previously reported value (IC50 = 38 nM) (Kuma et al., 2005). As might be expected, the Regorafenib can also exert a moderate or high in- hibition against the p38α (IC50 = 110 nM), which is close to that of the cognate p38 inhibitor BIRB796, whereas the Cabozantinib has no de- tectable activity for the kinase (IC50 = n.d.). The modeled complex structure of human p38α kinase domain with its noncognate inhibitor Regorafenib is shown in Fig. 6A. As can be seen, the inhibitor ligand can interact with the kinase in an extended conformation that is tightly packed against the p38α active site to de- ﬁne diverse nonbonded interactions across the complex interface. The inhibitor ligand is found to form ﬁve geometrically satisfactory hy- drogen bonds separately with the kinase residues Glu71, Thr106, Met109, Gly110 and Asp168, conferring strong speciﬁcity to the non- cognate kinase–inhibitor recognition, and intensive nonbonded con- tacts such as hydrophobic forces and van der Waals interactions are also identiﬁed at the interface that largely stabilize the complex archi- tecture. In addition, the complex structure of p38α with its cognate inhibitor BIRB796 is also shown in Fig. 6B. As can be seen, the Re- gorafenib and BIRB796 inhibitors adopt a similar binding mode to in- teract with p38α; both of them occupy at kinase active site and exhibit an extended conformation along the site. A number of kinase residues such as Glu71, Met109 and Asp168 are shared by the two inhibitors to form hydrogen bonds, which can be regarded as speciﬁc anchor residues that confer selectivity to the kinase inhibitors.

4. Conﬂicts of interest
No potential conﬂicts of interest.

References
Alemán, J.O., Farooki, A., Girotra, M., 2014. Eﬀects of tyrosine kinase inhibition on bone metabolism: untargeted consequences of targeted therapies. Endocr. Relat. Cancer 21, R247–R259.
Anastassiadis, T., Deacon, S.W., Devarajan, K., Ma, H., Peterson, J.R., 2011. Comprehensive assay of kinase catalytic activity reveals features of kinase inhibitor selectivity. Nat. Biotechnol. 29, 1039–1045.
Bai, Z., Hou, S., Zhang, S., Li, Z., Zhou, P., 2017. Targeting self-binding peptides as a novel strategy to regulate protein activity and function: a case study on the proto-oncogene tyrosine protein kinase c-Src. J. Chem. Inf. Model. 57, 835–845.
Bayly, C.I., Cieplak, P., Cornell, W.D., Kollman, P.A., 1993. A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges –the RESP model. J. Phys. Chem. 97, 10269–10280.
Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov, I.N., Bourne, P.E., 2000. The protein data bank. Nucleic Acids Res. 28, 235–242.
Bhattacharya, D., Nowotny, J., Cao, R., Cheng, J., 2016. 3Dreﬁne: an interactive web server for eﬃcient protein structure reﬁnement. Nucleic Acids Res. 44, W406–W409.
Bode, A.M., Dong, Z., 2007. The functional contrariety of JNK. Mol. Carcinog. 46,591–598.
Brown, N.P., Leroy, C., Sander, C., 1998. MView: a Web compatible database search or multiple alignment viewer. Bioinformatics 14, 380–381.
Buzzi, N., Colicheo, A., Boland, R., de Boland, A.R., 2009. MAP kinases in proliferating human colon cancer Caco-2 cells. Mol. Cell. Biochem. 328, 201–208.
Case, D.A., Cheatham 3rd, T.E., Darden, T., Gohlke, H., Luo, R., Merz Jr, K.M., Onufriev, A., Simmerling, C., Wang, B., Woods, R.J., 2005. The amber biomolecular simulation programs. J. Comput. Chem. 26, 1668–1688.
Chen, K.E., Lin, S.Y., Wu, M.J., Ho, M.R., Santhanam, A., Chou, C.C., Meng, T.C., Wang, A.H., 2014. Reciprocal allosteric regulation of p38γ and PTPN3 involves a PDZ do- main-modulated complex formation. Sci. Signal. 7, ra98.
Cong, L., Xia, Z.K., Yang, R.Y., 2014. Targeting the TGF-β receptor with kinase inhibitors for scleroderma therapy. Arch. Pharm. 347, 609–615.
Cui, Y.H., Chen, J., Xu, T., Tian, H.L., 2015. Structure-based grafting and identiﬁcation of kinase-inhibitors to target mTOR signaling pathway as potential therapeutics for glioblastoma. Comput. Biol. Chem. 54, 57–65.
Darden, T., York, D., Pedersen, L., 1993. Particale mesh Ewald and N.log(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089–10092.
Davis, M.I., Hunt, J.P., Herrgard, S., Ciceri, P., Wodicka, L.M., Pallares, G., Hocker, M., Treiber, D.K., Zarrinkar, P.P., 2011. Comprehensive analysis of kinase inhibitor se- lectivity. Nat. Biotechnol. 29, 1046–1051.
Dhillon, A.S., Hagan, S., Rath, O., Kolch, W., 2007. MAP kinase signalling pathways in cancer. Oncogene 26, 3279–3290.
Duan, Y., Wu, C., Chowdhury, S.S., Lee, M.C., Xiong, G.M., Zhang, W., Yang, R., Cieplak, P., Luo, R., Lee, T.S., Caldwell, J., Wang, J.M., Kollman, P., 2003. A point-charge force ﬁeld for molecular mechanics simulations of proteins. J. Comput. Chem. 24, 1999–2012.
Germann, U.A., Furey, B.F., Markland, W., Hoover, R.R., Aronov, A.M., RoiX, J.J., Hale, M., Boucher, D.M., Sorrell, D.A., Martinez-Botella, G., Fitzgibbon, M., Shapiro, P., Wick, M.J., Samadani, R., Meshaw, K., Groover, A., DeCrescenzo, G., Namchuk, M., Emery, C.M., Saha, S., Welsch, D.J., 2017. Targeting the MAPK signaling pathway in cancer: promising preclinical activity with the novel selective ERK1/2 inhibitor BVD- 523 (UliXertinib). Mol. Cancer Ther. 16, 2351–2363.
Gordon, J.C., Myers, J.B., Folta, T., Shoja, V., Heath, L.S., Onufriev, A., 2005. H++: a server for estimating pKas and adding missing hydrogens to macromolecules. Nucleic Acids Res. 33, W368–W371.
Guex, N., 1996. Swiss-PdbViewer: a new fast and easy to use PDB viewer for the Macintosh. EXperientia 52, A26.
Hanks, S.K., Quinn, A.M., Hunter, T., 1988. The protein kinase family: conserved features and deduced phylogeny of the catalytic domains. Science 241, 42–52.
Hou, T., Li, N., Li, Y., Wang, W., 2009. Characterization of domain-peptide interaction interface. Mol. Cell. Proteomics 8, 639–649.
Jiang, L., Tang, C., Rao, J., Xue, Q., Wu, H., Wu, D., Zhang, A., Chen, L., Shen, Z., Lei, L., 2017. Systematic identiﬁcation of the druggable interactions between human protein kinases and naturally occurring compounds in endometriosis. Comput. Biol. Chem. 71, 136–143.
Jindal, H.M., Le, C.F., Mohd Yusof, M.Y., Velayuthan, R.D., Lee, V.S., Zain, S.M., Isa, D.M., Sekaran, S.D., 2015. Antimicrobial activity of novel synthetic peptides derived from indolicidin and ranalexin against streptococcus pneumoniae. PLoS One 10.
Kuma, Y., Sabio, G., Bain, J., Shpiro, N., Márquez, R., Cuenda, A., 2005. BIRB796 inhibits all p38 MAPK isoforms in vitro and in vivo. J. Biol. Chem. 280, 19472–19479.
Lee, M., Kim, S.Y., Kim, J.G., Kim, H.S., Kim, S.M., Kim, E.J., 2013. Mitogen-activated protein kinase phosphatase-1 inhibition and sustained extracellular signal-regulated kinase 1/2 activation in camptothecin-induced human colon cancer cell death. Cancer Biol. Ther. 14, 1007–1015.
Levitzki, A., 2013. Tyrosine kinase inhibitors: views of selectivity, sensitivity, and clinical performance. Annu. Rev. Pharmacol. ToXicol. 53, 161–185.
Lu, H., Liu, S., Zhang, G., Bin Wu Zhu, Y., Frederick, D.T., Hu, Y., Zhong, W., Randell, S., Sadek, N., Zhang, W., Chen, G., Cheng, C., Zeng, J., Wu, L.W., Zhang, J., Liu, X., Xu, W., Krepler, C., Sproesser, K., Xiao, M., Miao, B., Liu, J., Song, C.D., Liu, J.Y., Karakousis, G.C., Schuchter, L.M., Lu, Y., Mills, G., Cong, Y., Chernoﬀ, J., Guo, J., Boland, G.M., Sullivan, R.J., Wei, Z., Field, J., Amaravadi, R.K., Flaherty, K.T., Herlyn, M., Xu, X., Guo, W., 2017. PAK signalling drives acquired drug resistance to MAPK inhibitors in BRAF-mutant melanomas. Nature 550, 133–136.
Lum, K.Y., Tay, S.T., Le, C.F., Lee, V.S., Sabri, N.H., Velayuthan, R.D., Hassan, H., Sekaran, S.D., 2015. Activity of novel synthetic peptides against candida albicans. Sci. Rep. 5, 9657.
Meng, X.Y., Zhang, H.X., Mezei, M., Cui, M., 2011. Molecular docking: a powerful ap- proach for structure-based drug discovery. Curr. Comput. Aided Drug Des. 7, 146–157.
Mikulskis, P., Genheden, S., Ryde, U., 2014. Eﬀect of explicit water molecules on ligand- binding aﬃnities calculated with the MM/GBSA approach. J. Mol. Model. 20, 2273. Namboodiri, H.V., Bukhtiyarova, M., Ramcharan, J., Karpusas, M., Lee, Y., Springman, E.B., 2010. Analysis of imatinib and sorafenib binding to p38α compared with c-Abl and b-Raf provides structural insights for understanding the selectivity of inhibitors targeting the DFG-out form of protein kinases. Biochemistry 49, 3611–3618.
Pappalardo, F., Russo, G., Candido, S., Pennisi, M., Cavalieri, S., Motta, S., McCubrey, J.A., Nicoletti, F., Libra, M., 2016. Computational modeling of PI3 K/AKT and MAPK signaling pathways in melanoma cancer. PLoS One 11, e0152104.
Patel, R.Y., Doerksen, R.J., 2010. Protein kinase-inhibitor database: structural variability of and inhibitor interactions with the protein kinase P-loop. J. Proteome Res. 9, 4433–4442.
Pennisi, M., Russo, G., Di Salvatore, V., Candido, S., Libra, M., Pappalardo, F., 2016. Computational modeling in melanoma for novel drug discovery. EXpert Opin. Drug Discovery 11, 609–621.
Ryckaert, J.P., Ciccotti, G., Berendsen, H.J.C., 1977. Numerical integration of the carte- sian equations of motion of a system with constraints: molecular dynamics of n-al- kanes. J. Comput. Phys. 23, 327–341.
Séverin, S., Ghevaert, C., Mazharian, A., 2010. The mitogen-activated protein kinase signaling pathways: role in megakaryocyte diﬀerentiation. J. Thromb. Haemost. 8, 17–26.
Sachs, R.E., Ginsburg, P.B., Goldman, D.P., 2017. Encouraging new uses for old drugs. JAMA 318, 2421–2422.
Schaeﬀer, H.J., Weber, M.J., 1999. Mitogen-activated protein kinases: speciﬁc messages from ubiquitous messengers. Mol. Cell. Biol. 19, 2435–2444.
Son, Y., Cheong, Y.K., Kim, N.H., Chung, H.T., Kang, D.G., Pae, H.O., 2011. Mitogen-activated protein kinases and reactive oXygen species: how can ROS activate MAPK pathways? J. Signal Transduction 2011, 792639.
Trott, O., Olson, A.J., 2010. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, eﬃcient optimization, and multithreading. J. Comput. Chem. 31, 455–461.
Tsui, V., Case, D.A., 2000a. Theory and applications of the generalized Born solvation model in macromolecular simulations. Biopolymers 56, 275–291.
Tsui, V., Case, D.A., 2000b. Molecular dynamics simulations of nucleic acids with a generalized Born solvation model. J. Am. Chem. Soc. 122, 2489–2498.
Underwood, D.C., Osborn, R.R., Kotzer, C.J., Adams, J.L., Lee, J.C., Webb, E.F., Carpenter, D.C., Bochnowicz, S., Thomas, H.C., Hay, D.W., Griswold, D.E., 2000. SB 239063, a potent p38 MAP kinase inhibitor, reduces inﬂammatory cytokine pro- duction, airways eosinophil inﬁltration, and persistence. J. Pharmacol. EXp. Ther. 293, 281–288.
UniProt, C., 2015. UniProt: a hub for protein information. Nucleic Acids Res. 43, D204–D212.
Wadsworth, S.A., Cavender, D.E., Beers, S.A., Lalan, P., Schafer, P.H., Malloy, E.A., Wu,
W., Fahmy, B., Olini, G.C., Davis, J.E., Pellegrino-Gensey, J.L., Wachter, M.P., Siekierka, J.J., 1999. RWJ 67657, a potent, orally active inhibitor of p38 mitogen- activated protein kinase. J. Pharmacol. EXp. Ther. 291, 680–687.
Wang, J., Wolf, R.M., Caldwell, J.W., Kollman, P.A., Case, D.A., 2004. Development and testing of a general amber force ﬁeld. J. Comput. Chem. 25, 1157–1174.
Wellbrock, C., 2014. MAPK pathway inhibition in melanoma: resistance three ways. Biochem. Soc. Trans. 42, 727–732.
Wilhelm, S.M., Dumas, J., Adnane, L., Lynch, M., Carter, C.A., Schütz, G., Thierauch, K.H., Zopf, D., 2011. Regorafenib (BAY 73–4506): a new oral multikinase inhibitor of angiogenic, stromal and oncogenic receptor tyrosine kinases with potent preclinical antitumor activity. Int. J. Cancer 129, 245–255.
Wu, P., Nielsen, T.E., Clausen, M.H., 2015. FDA-approved small-molecule kinase in- hibitors. Trends Pharmacol. Sci. 36, 422–439.
Wu, P., Nielsen, T.E., Clausen, M.H., 2016. Small-molecule kinase inhibitors: an analysis of FDA-approved drugs. Drug Discov. Today 21, 5–10.
Xu, J., Wang, P., Yang, H., Zhou, J., Li, Y., Li, X., Xue, W., Yu, C., Tian, Y., Zhu, F., 2016. Comparison of FDA approved kinase targets to clinical trial ones: insights from their system proﬁles and drug?target interaction networks. BioMed Res. Int. 2016.
Yang, C., Zhang, S., He, P., Wang, C., Huang, J., Zhou, P., 2015. Self-binding peptides: folding or binding? J. Chem. Inf. Model. 55, 329–342.
Yang, C., Zhang, S., Bai, Z., Hou, S., Wu, D., Huang, J., Zhou, P., 2016. A two-step binding mechanism for the self-binding peptide recognition of target domains. Mol. Biosyst. 12, 1201–1213.
Zahler, S., Tietze, S., Totzke, F., Kubbutat, M., Meijer, L., Vollmar, A.M., Apostolakis, J.,
2007. Inverse in silico screening for identiﬁcation of kinase inhibitor targets. Chem. Biol. 14, 1207–1214.
Zhang, R., Loughran, T.P., 2011. Oﬀ-target eﬀects of tyrosine kinase inhibitors: beauty or the beast? Leuk. Lymphoma 52, 556–557.
Zhou, P., Yang, C., Ren, Y., Wang, C., Tian, F., 2013. What are the ideal properties for functional food peptides with antihypertensive eﬀect? A computational peptidology approach. Food Chem. 141, 2967–2973.
Zhou, P., Zhang, S., Wang, Y., Yang, C., Huang, J., 2016. Structural modeling of JNK-IN-8/peptide/carbamazepine/T-cell receptor complex architecture: implication for the molecular mechanism of carbamazepine-induced Stevens-Johnson syndrome/ toXic epidermal necrolysis. J. Biomol. Struct. Dyn. 34, 1806–1817.
Zhu, Q., Chen, J., Wu, X., Jin, X., Ruan, B., 2014. Repurposing of kinase inhibitors to target c-Abl as potential therapeutics for Alzheimer’s disease. J. Pharm. Innov. 9, 331–340.