Design, synthesis, and biological evaluation of polo-like kinase 1/eukaryotic elongation factor 2 kinase (PLK1/EEF2K) dual inhibitors for regulating breast cancer cells apoptosis and autophagy
ABSTRACT: Both PLK1 and EEF2K are serine⁄threonine kinases that play important roles in the proliferation and programmed cell death of various types of cancer. They are highly expressed in breast cancer tissues. Based on the multiple-complexes generated pharmacophore models of PLK1 and homology models of EEF2K, the integrated virtual screening is performed to discover novel PLK1/EEF2K dual inhibitors. The top ten hit compounds are selected and tested in vitro, and five of them display PLK1 and EEF2K inhibition in vitro. Based on the docking modes of the most potent hit compound, a series of derivatives are synthesized, characterized and biological assayed on the PLK1, EEF2K as well as breast cancer cell proliferation models. Compound 18i with satisfied inhibitory potency are shifted to molecular mechanism studies contained molecular dynamics simulations, cell cycles, apoptosis and autophagy assays. Our results suggested that these novel PLK1/EEF2K dual inhibitors can be used as lead compounds for further development breast cancer chemotherapy.
1.Introduction
According to the cancer epidemiology data collected in the cancer statistics 2017, breast cancer is estimated to account about 14.4% of cancer related death and 29.6% of new cases in female, and which is the top type of female new cancer cases and the second cause of female cancer related death worldwide after lung cancer.[1] Breast cancer can be categorized as over twenty histologic subtypes, and the typical four subtypes that are identified by the biological markers, including the hormone receptors (HR) expression levels as well as HER2 proteins.[2] The main subtypes of breast cancer cases are HR+/HER2- (luminal A) and HR+/HER2+ (luminal B), which are responsible to the endocrinotherapy. About 10-15% of breast cancer cases are HR-/HER2- (triple negative), and 4-6% are HR-/HER2+ (HER2-enriched), in these subtypes, triple-negative breast cancers are estimated to be poor prognosis and higher aggressive because of lacking efficiently targeted therapies.[3-9] In recent years, small molecule kinase inhibitors have been identified as a novel type of emerging therapeutic approach for breast cancers. Except the classic receptor tyrosine kinase inhibitors, e.g. gefitinib,[10, 11] lapatinib,[12] tivozanib[13-15] and foretinib,[16-18] serine/threonine protein kinase (STK) inhibitors are also interesting therapeutic targets of breast cancers in resistance to cytotoxic therapy.[19-23] Recently, many STK inhibitors have displayed promising potency in the clinical applications and clinic trials for breast cancer treatment, i.e. CDK 4/6 inhibitor palbociclib,[24] Akt inhibitor ipatasertib,[25] Raf inhibitor vemurafinib,[26] etc. Therefore, the discovery of novel STK inhibitors targeted to the over-expressed STKs in breast cancer would be an attractive therapy
approach.
PLK1 is the main member of a family of serine/threonine kinases, called Polo-like kinase family, plays as an important regulator of cell cycles, mitotic exit and cytokinesis.[27] There are five members in the human PLK family, i. e. PLK1 to PLK5, all PLKs have one or more C-terminal ‘polo-box’ domains, and the N terminal kinase domain (truncated in PLK5).[28] The regulation function of PLK1 is involved in cell division, DNA replication/repair, DNA checkpoint regulation and microtubule dynamics.[29] As a proto-oncogene, the expression levels of PLK1 are elevated in various cancers, e. g. breast cancer, cutaneous melanoma, colorectal cancer, lung adenocarcinoma cancer, gastric cancer and glioblastoma.[30] Most reported PLK1 inhibitors binding to the ATP pocket of catalytic domain, similar to the conventional ATP competitive inhibitors of STKs or receptor tyrosine kinases (RTK).[31] Volasertib is the first PLK1 inhibitors approved to mono- or combined therapy of AML (acute myeloid leukemia).[32] Other PLK1 inhibitors, such as GSK461346, TAK960, NMS-P937 and ON 01910-Na are still in different stages of clinic trials.[28, 33] Moreover, a series of polo-box domain (PBD) inhibitors called Poloxins could significantly suppressed tumor cell growth both in vitro and in vivo.[34-38]EEF2K is an atypical STK, also called as calcium/calmodulin-dependent protein kinase-III (CaMK-III), regulates the protein synthesis of chain elongation by phosphorylation of EEF2 at Ser56 and Thr58.[39] EEF2K can response to the endoplasmic reticulum (ER) stress, and then controls the balance between apoptosis and autophagy in cancer cells.[39-41] The expression level of EEF2K is elevated in various cancer cells including breast cancer, malignant glioma, leukemia and pancreatic cancer, etc.[42-45] Although there are several EEF2K inhibitors reported in the past several years, such as Rottlerin, NH125, A-484854 and 1,3-selenazine analogues, etc.[46-49] Only limited EEF2K inhibitors demonstrate potent tumor suppressed activity in cellular or xenograft models and none inhibitor enters into clinic trial stage.We are interested in discovering potent and selective PLK1/EEF2K dual inhibitors to regulate the programmed cell death of breast cancer, especially triple-negative breast cancer. In the current study, we report the integrated virtual screening of PLK1/EEF2K dual inhibitors by multi-complexes based on pharmacophore models [50-54], homology modeling and molecular docking. The hit compound is structurally optimized by drug-protein interactions and then evaluated by enzymatic and cellular assays. Moreover, the detailed mechanisms of novel PLK1/EEF2K dual inhibitors are probed by molecular dynamics simulations and protein immunoblottings.
2.Results and discussion
To avoid diagnostic bias, the positive expressions of PLK1 and EEF2K on a breast cancer tissue microarray (TMA) are determined by two independent pathologists. The relative expression levels of PLK1 and EEF2K in both tumor and matched adjacent tissues of representative case are illustrated in Figure 1A. There are significant difference in both PLK1 and EEF2K protein expression levels between tumor and matched adjacent tissues. Moreover, the mRNA levels of PLK1 and EEF2K are retrieved and analyzed in the RNA sequencing (RNA-Seq) dataset of the Cancer Genomics Atlas (TCGA) breast cancer cohort (Figure 2B).[55-57] The relative RPKM (Reads Per Kilo-bases per Million reads) of PLK1 and EEF2K in breast cancer tissues (n=1091) are 16.7 and 1.52 fold than that of matched adjacent tissues (n=113). Next, we have determined whether PLK1 and EEF2K expression in tumor tissues are associated with the clinical prognosis. As shown in Figure 1C, the individually elevated expression of PLK1 or EEK2K in tumors exhibited worse survival rates compared with those in the low-expression group. There are little difference in clinical characteristics between the PLK(+)/EEF2K(-) group and PLK(-)/EEF2K(+) group. In addition, patients with synchronously high expression of PLK1 or EEK2K had the worst survival profile than the other groups.
To discover novel small-molecule PLK1/EEF2K dual-inhibitors, the four-step strategy is used: Generation of the multicomplexes based on pharmacophore models of PLK1, homology modeling of EEF2K kinase domain, the hybrid protocol virtual screening of fragment-like database, and the synthesis, characterization and bioassay of novel PLK1/EEF2K dual inhibitors.There are 11 crystallography structures of PLK1-inhibitor complex deposited in the PDB (protein databank) database (Table S1).[58] All PLK1-inhibitor complexes are aligned by the main chain α-carbon atoms and superposed together by the Modeller package reported previously.[59-61] The statistical analysis of structure-based pharmacophore features are shown in Table 1 and Figure 2. There are 15 pharmacophore features extracted from aligned PLK1-inhibitor complexes, including five hydrophobic (H1-H5), one positive chargeable (P1), two ring aromatic (R1 and R2) features, three hydrogen bond acceptors (A1-A3) and four hydrogen bond donors (D1-D4). Among all of the pharmacophore features, there are four features beyond 50% probability, including A1, D1, H1 and H2. In general case, the common pharmacophore features with higher probability might be more efficient in virtual screening, therefore the most frequent features (A1, D1, H1 and H2) are selected and merged into a multicomplex-based phamacophore (Fig. 2). Since some ligand-based pharmacophore models for PLK1 have been reported by Chekkara et al.,[62] Lu et al.[22] and Yang et al.[63], these ligand-based pharmacophore models have not cover all inhibitor-protein interactions (Table 1).
Because there is no atomic protein structure of EEF2K kinase domain available, we have modelled the structure of EEF2K kinase domain by homology modeling method in advance. The Accelrys Discovery Studio (DS3.5, Accelrys, San Diego, CA) is used and the brief procedures are listed as follows: In brief, the protein sequence of EEF2K kinase domain (Ile107 to Phe326) is retrieved from the UniProt database (UniProt entry O00418).[64] The PSI-BLAST search module in Discovery Studio is used for the template search, and two protein, the Alpha-kinase Domain of Mysosin heavy chain kinase A (PDB No. 5DYJ and 3LMH) showed good identity to EEF2K kinase domain (41% and 40%, respectively), the former is chosen as template for homology modeling and the alignment profile is shown in supplementary materials. Then the initial model is optimized by the standard molecular dynamics protocol in Discovery Studio. The equilibrated model is checked by the Profile 3D and Ramachandran Plot analysis (in Supplementary data, Supplementary Figs. S1– S3).The hybrid protocol virtual screening is carried out using the developed multicomplex-based phamacophore and homology model of EEF2K kinase domain. The fragment-like library of the ZINC database (a free database of commercially-available compounds for virtual screening) is selected for the first pharmacophore based virtual screening, which contains about 1.6 million compounds, and the top-scored compounds are docked into EEF2K kinase domain by the Libdock program, a rigid docking method. Then the top 500 hit compounds are subsequently screened by the LigandFit program, a flexible docking method considering induced-fit effects. The top ten scoring compounds are selected for bioassay, and five of them display potent inhibitory activities at 50µM against PLK1 and EEF2K (Figure 3). The most active one, compound 7 is structurally optimized and biological evaluated in subsequent studies.
To further optimize the structure of hit-compound 7, the 2D diagrams and 3D binding modes to PLK1 and EEF2K are shown in Figure 4. According to Figure 4A,the N-methylpyridin-2-amine fragment form stable hydrogen bonds to Asp194 of PLK1, and the Asp184, Gln286 and Gly298 of EEF2K, respectively. Moreover, the pyridine ring, N-methyl and phenyl ring contribute to the hydrophobic interactions. In the hydrogen bond affinity contoured surface, the amino group nears hydrogen bond donor areas in both proteins, however, the hydrophobicity of these areas in PLK1 and EEF2K are differed (see Figure 4B and 4C). Therefore the 4-(o-substituted phenyl) pyridin-2-amine fragment is kept to maintain the hydrophobic and hydrogen bond interactions to Asp194 of PLK1, and the Asp184, Gln286 and Gly298 of EEF2K, respectively. The N-methyl group is changed to ethyl, propyl or cyclopropyl group to better adapt the binding pocket of EEF2K, and the o-methyl on phenyl ring is change to fluorine, methoxyl or trifluoromethyl group, and a series of aromatic urea groups are substituted on the terminal amino group.
The synthetic route of target compounds is depicted on the scheme 1. In brief, the amino group of 4-bromopyridin-2-amine is substituted by alkyl halide by the alkaline catalyst sodium hydride. The 4-bromo-N-substituted-2-amine intermediates are coupling to 1-bromo-2-substitute-4-nitrobenzene via canonical Suzuki reaction to from intermediates 14a-f. Then the 2-amino group of compound 14a-f is protected by t-butyloxycarboryl group and the nitro group is reduced to amine by ferrous reductant to form intermediates 16a-f. Finally, the amino groups of 16a-f are substituted by a series of aromatic isocyanate to produce the target compounds 18a-w and 19a-d. 1H NMR, 13C NMR and mass spectrum are employed to characterize the chemical structures of target compounds.
The 2D diagrams of interactions; B. The surface of binding pocket contoured by the hydrogen bond affinity of PLK1 or EEF2K; C. The solvent accessible surface (SAS) of proteins contoured by the hydrophobicity, blue contour represents hydrophilic area, yellow to white represents hydrophobic area.Reagents and conditions: (a) NaH, 1,4-dioxane, 25℃, 16h; (b) bis(pinacolato)diboron, Pd2dba3, 1,4-dioxane, 12h; (c) Boc2O, DMAP, pyridine, 50℃, 36h; (d) Fe/HOAc,ethanol, 4h; (e) DCM, 25℃, 8 h.The inhibitory potencies of the synthesized compounds are shown in Table 2, including the kinase activity of PLK1 and EEF2K, as well as the cellular proliferation on breast cancer cell lines MCF-7 and BT-483. The bioactivity data for SAR analysis are inhibition ratios with a constant concentration of 20µM for each compounds. From Table 2, we can see that compound 18i, which contains an ethyl group at the R1 position, a trifluoromethyl group with the R2 position and the 3-chlorine atom at the R3 position, is the most active one. It is noticeable that compound 18e, 18s and 18l bearing inhibition ratios of -13.7, -4.29 and -23.0% on EEF2K, respectively. Thesubtractive inhibition ratios do not suggest that these compound are EEF2K activators. It is only inferred that these compounds have not significantly influence on the kinase activities of EEK2K. The compounds display minimal EEF2K inhibitory activities with cyclopropyl group on the R1 position or the naphthalene group on the R3 position. These results suggested that the binding pocket of EEF2K at R1 position is limited, and its hydrophilic area at R3 position cannot tolerates bigger hydrophobic groups. The proliferation assay on MCF-7 and BT-483 cells also reveals that compounds with good kinase inhibitory activities indeed demonstrate higher cytotoxicity on breast cancer cell lines.
Finally, the compound 18i displays the strongest inhibitory capacity on the PLK1 and EEF2K with IC50 values of 0.085 and 0.762 µM, as well as the proliferation of MCF-7 and BT-483 breast cancer cells with IC50 values of 1.24 and 13.3µM, respectively. Moreover, the KINOMEScan kinase profiling on compound 18i were performed to identify its selectivity on a panel of kinases. The results were shown in Table S2, and only PLK1 EEF2K were efficiently suppressed at 1.0 µM concentration. Therefore compound 18i is shifted to subsequently studies.It is obviously that the treatment of 5µM compound 18i for 24 hours could induce the formation of apoptotic nuclear fragments. According to figure 5C, the percentage of apoptotic cells in 10µM 18i group is 21.3% ± 5.20%, which is significantly higher than that in 5µM 18i group (13.3% ± 2.9%, p < 0.05), and NS (2.4% ± 0.1%, p < 0.05) groups. These results indicate that the apoptosis induced by compound 18i with a concentration-dependent manner. As the EEF2K is identified as an autophagy regulator, the changes of autophagy levels before and after compound 18i treatment are detected on the GFP-LC3 transfected BT-483 cells.[65, 66] According to Figure 5D, the formation of autophagosome are observed by the aggregation of GFP-LC3 punctures.[67-69] The cell cycle analysis results indicate that the PLK1/EEF2K dual inhibitor 18i potently induce cell cycles arrested into subG1 and G2/M phases. Both 24 and 72 hours treatment display common trends, and the 72h treatment group shows higher cell cycle arrest effect.In order to further validate the PLK1/EEF2K dual inhibitory effects and apoptosis/autophagy regulatory activity of compound 18i, the expression levels of total and phosphorylated PLK1, EEF2K, EEF2, as well as the autophagy and apoptosis relative proteins are determined by western blotting in BT-483 cells (seeFigure 6). According to Figure 6, compound 18i indeed inhibits theautophosphorylation of PLK1 and EEF2 phosphorylation by EEF2K in BT-483 cells without influences on the levels of total PLK1, EEF2K and EEF2, above results indicate that compound 18i specifically blocked of activation of PLK1 and EEF2K.[47, 70, 71] Moreover, the changes of autophagy relation proteins compound 18i suggest that 18i could activate the autophagic flux of BT-483 cells by the activation of LC3 and degradation of SQSTM1 (p62).[65, 72-74] To the apoptosis related proteins, the apoptosis inhibitory protein Bcl-2 is suppressed, the cytochrome C, Bad and Bax are up-regulated, and the cleavage of Caspase-3 is activated after compound 18i treatment. These results demonstrate that the PLK1/EEF2K dual inhibitor activates the mitochondrial apoptosis pathway. Meanwhile, the representative proteins involved in death receptor apoptosis pathway, such as caspase-8, Fas-Associated protein with Death Domain (FADD) and Fas are not obviously changed after compound 18i treated on Western blotting assay. The activation of autophagy and mitochondrial apoptosis are mainly involved in 18i-regulated programmed cell death, the detailed signaling pathway crosstalk between PLK1, EEF2K and apoptosis/autophagy might be a complicated way and to be further studied in the further.[75-77]The MD simulations are performed between compound 18i and PLK1 and EEF2K complexes for 100 ns (nanosecond) scale. The simulation stability of MD trajectories are evaluated by the peptide chain backbone atoms RMSD (root mean standard deviation, see supplementary information) of PLK1 or EEF2K, relative to their initial conformations. It is noteworthy that the complexes of compound 18i and PLK1 or EEF2K have reached equilibrium after 20 ns in the simulation. The average RMSD of the compound 18i -PLK1 and EEF2K complexes are 0.31 and 0.37 nm, respectively. In addition, the RMSD fluctuations of the two complexes are lower than 0.05nm, above results suggest that the MD simulations are reliable stability. Therefore the binding modes and hydrogen bonds analysis are performed based on the snapshots of simulation trajectories extracted from 21 to 100 ns. The average binding conformations of compound 18i to the binding sites of PLK1 and EEF2K are displayed in Figure 7A and 7B, respectively. From Figure 7A and 7B it can be observed that the urea fragment could formation stable hydrogen bond to Cys133 of PLK1 and Arg144 of EEF2K, respectively. Moreover, the 3-chlorophenyl group can form hydrophobic interactions with Leu59 of PLK1 and Tyr236 of EEF2K, and Phe183 of PLK1 forms stable π-π stack with 18i, which might be a potential reason of the better inhibition activity of PLK1 of compound 18i than that of EEF2K.The quantitative analysis of hydrogen bonding network between inhibitor and PLK1 or EEF2K are based on the calculation of MD trajectories. The detailed information hydrogen bonds are summarized in Table 3, there are two remarkable results are obtained: (1) Compound 18i forms stable hydrogen bonds with the Lys82 and Asp194 of PLK1, as well as Gly298 and Phe295 of EEF2K, respectively. It is similar to the hit-compound 18i, the core hydrogen bond interactions are remained after structure optimization; (2) the urea group of compound 18i forms hydrogen bond with Cys133 of PLK1 as a donor, and meanwhile it forms hydrogen bond with Arg144 of EEF2K as an acceptor. In consideration of the urea group locates on a hydrogen bond acceptor area of PLK1 and a donor area of EEF2K, its dual-characterin hydrogen bond formation plays an important to keep stable binding affinity of PLK1 and EEF2K proteins. The two results explain in part why compound 18i produces similar inhibitory effects on PLK1 and EEF2K with different interaction contacts.Figure 7. The binding modes of compound 18i to PLK1 and EEF2K kinase domains based on molecular dynamics simulations. A.. the 2D diagrams of interaction residues between compound 18i and PLK1 and EEF2K, respectively; B. the surface of binding pocket contoured by the hydrogen bond affinity of PLK1 or EEF2K. 3.Conclusions In conclusion, a series of 1-(4-(2-substituted-pyridin-4-yl)-3-substituted-phenyl)-3-phenylurea derivatives are designed, synthesized and biologically evaluated as novel PLK1/EEF2K dual inhibitors against breast cancer cells. The most potent compound 18i inhibited PLK1 and EEF2K with the IC50 values at sub micromole level, inhibits the proliferation of breast cancer cells, induced autophagy and mitochondrial apoptotic cell death. In addition, its kinase inhibitory activities are validated by western blotting in the BT-483 cells, suggesting that the PLK1/EEF2K dual inhibition might be an efficient approach to cancer chemotherapy. Molecular dynamics simulation and docking studies suggest the detailed interaction modes of compound 18i with PLK1 and EEF2K, and the SAR analysis provide valuable information for further study of these PLK1/EEF2K inhibitors. Collectively, our results manifest that the integrative computational methods, synthesis and biological discovery of synthesized PLK1/EEF2K dual inhibitor is an NMS-P937 attractive lead compound for breast cancer chemotherapy.