An investigation of binding interactions of tumor-targeted peptide conjugated polyphenols with the kinase domain of ephrin B4 and B2 receptors

Abstract

Abstract

Recent studies have shown that Ephrin receptors may be upregulated in several types of cancers including breast, ovarian and endometrial cancers, making them a target for drug design. In this work, we have utilized a target-hopping approach to design new natural product-peptide conjugates and examined their interactions with the kinase-binding domain of EphB4 and EphB2 receptors. The peptide sequences were generated through point mutations of the known EphB4 antagonist peptide TNYLFSPNGPIA. Their anticancer properties and secondary structures were analyzed computationally. Conjugates of most optimum of peptides were then designed by binding the N-terminal of the peptides with the free carboxyl group of the polyphenols sinapate, gallate and coumarate, which are known for their inherent anticancer properties. To investigate if these conjugates have a potential to bind to the kinase domain, we carried out docking studies and MMGBSA free energy calculations of the trajectories based on the molecular dynamics simulations, with both the apo and the ATP bound kinase domains of both receptors. In most cases binding interactions occurred within the catalytic loop region, while in some cases the conjugates were found to spread out across the N-lobe and the DFG motif region. The conjugates were further tested for prediction of pharmacokinetic properties using ADME studies. Our results indicated that the conjugates were lipophilic and MDCK permeable with no CYP interactions. These findings provide an insight into the molecular interactions of these peptides and conjugates with the kinase domain of the EphB4 and EphB2 receptor. As a proof of concept, we synthesized and carried out SPR analysis with two of the conjugates (gallate-TNYLFSPNGPIA and sinapate-TNYLFSPNGPIA). Results indicated that the conjugates showed higher binding with the EphB4 receptor and minimal binding to EphB2 receptor. Sinapate-TNYLFSPNGPIA showed inhibitory activity against EphB4. These studies reveal that some of the conjugates may be developed for further investigation into in vitro and in vivo studies and potential development as therapeutics.

Graphic Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s11030-023-10621-x.

Introduction

Breast cancer is among the most common and leading causes of cancer death for women in the United States [1]. Current treatment methods consist of surgical resection, chemotherapy, radiation, and hormonal therapy, all having systemic impacts [2]. There are several forms of breast cancer, including estrogen receptor (ER +), human epidermal growth factor receptor2 + ve (HER2 + ve), and triple negative (TNBC). Due to the lack of hormone receptors commonly expressed in ER + and HER2 + breast cancers, TNBC presents with more challenges for treatments as it is highly aggressive and metastatic [3, 4].

Polyphenol extracts from natural food sources such as tea, coffee, cacao are known to exhibit antioxidant and antitumor activity [5, 6]. In particular, cinnamic acid and its derivatives found in the plant Cinnamomum zeilancium [7], are also known to display anticancer activity. Previous research revealed that cinnamic acid demonstrated antiproliferative effects on both osteosarcoma and human colon adenocarcinoma cells lines [8]. Included in the subgroup of cinnamic acid derivatives are polyphenols like caffeic acid, ferulic acid, p-coumaric acid, and sinapic acid [9] that have been known to induce apoptosis in the HT-1080 human fibrosarcoma cells [10] and MDA-MB-231 breast cancer cell lines [11]. Para-coumaric acid has been shown to reduce angiogenesis and endothelial cell migration in the ECV304 cell lines [12] while sinapic acid has been shown to reduce the viability of T47D breast cancer cells by inducing apoptosis [13]. In addition, gallic acid found in several nuts and berries has demonstrated anticancer properties [14]. For example, gallic acid-grafted-chitosan-caseino-phosphopeptide nanoparticles were found to reduce the proliferation of the Caco-2 colon cancer cells [15].

In the realm of targeting, tumor targeting peptides are also gaining importance due to their specificity [16]. For example, RGD-peptide conjugates were found to display anticancer activity against integrins in non-small cell lung carcinoma and melanoma [17]. Such peptides are attractive due to their ability to target specific receptors that are often over-expressed in tumor cells [18]. Epidermal growth factor receptor (EGFR) has also been successfully targeted in breast cancer through monoclonal antibodies such as cetuximab, tyrosine kinase inhibitors, and targeted peptides including D-K₆L₉ which was found to be an EGFR-lytic peptide [19]. The tyrosine kinase (RTK) inhibitor dasatinib has been found to reduce both expression and phosphorylation levels of receptor EphA2, leading to lower rates of malignancy [20]. The RTK receptor EphB4 responds to the ephrin-B2 ligand and is typically active during early stages of development [21]. However, EphB4 provides a survival signal that prevents apoptosis and is implicated in late stages of breast and ovarian cancer, causing cytoskeletal changes, tumor angiogenesis, cell migration and metastasis [22–25]. EphB4 is also found to be overexpressed in lung tumors [26]. Furthermore, it has also been demonstrated that overexpression of EphB4 in mouse mammary epithelium led to increased angiogenesis [27, 28]. Within the same family, the EphB2 RTK has also been found to be upregulated in breast carcinoma and associated with decreased survival rates [29]. Both EphB4 and EphB2 are transmembrane receptors that auto phosphorylate upon dimerization following ligand binding, activating a signaling cascade [30, 31]. EphB2 has been found to induce invasion through coordination with MMP2 and MMP9 metalloproteases [32]. EphB4 and EphB2 receptors are consequently attractive targets for therapeutic drug design. For example, EphB4 knockdown in human and mouse prostate cancer cells led to cell death [33]. In another study, monoclonal antibodies designed to target EphB4 receptors were found to reduce vascularization and increase hypoxia in tumors of several cell lines, including SCC15, HT29, and Hey [34]. A more comprehensive list of currently approved and developing drug candidates that are known to target the EphB2 and EphB4 receptors is shown in Supplementary Information Table I.

There are several similarities between EphB4 and EphB2 receptors and some differences as well. For example, it has been reported that both EphB2 and EphB4 activation enhances stromal cell derived factor-1 (SDF-1)–induced signaling and chemotaxis that are also required for extracellular matrix–dependent endothelial cell clustering [28]. Both EphB2 and EphB4 receptor over-expression is involved to are involved the development of breast cancer and both could be potential predictive biomarkers [29, 35]. Interestingly, it has been found that the EphB4 ligand-binding domain (LBD) is highly selective toward the ephrin B2 ligand and the crystal structure of the complex has revealed that residue Leu 95 of EphB4 receptor LBD which causes a major structural rearrangement of the EphB4 receptor J-K loop is critical for EphB4’s selectivity. On the other hand, EphB2 and other Eph receptors, have a conserved arginine residue at the position corresponding to Leu-95 of EphB4 [36]. Additionally, several amino acid residues that make vital contacts with the ephrin-B2 G-H loop in the high-affinity dimerization interface of EphB2 receptor are not conserved in EphB4. For example, Ser-194 of EphB2 is conserved in other EphB receptors, but not in EphB4, which has an alanine residue present at the corresponding position. Researchers have also demonstrated that two pairs of salt bridges (Asp39 in EphB4 and Lys57 in ephrin-B2; Arg65 in EphB4 and Glu116 in ephrin-B2) are unique to EphB4, resulting in ephrinB2 recognition and conformational selection [37].

As far as the kinase domain is concerned, in general EphB structures are known to adopt the canonical bi-lobed kinase fold, though there are some differences in the hinge region of the kinase domains of different EphB receptors upon ligand binding. For instance, the well-known inhibitor, staurosporine-bound EphB4 adopts a closed conformation, with the Gly-rich loop folding tightly over the ligand, while the apoprotein structures of EphB2 are more open, with the Gly-rich loops partly disordered. Additionally, the EphB4 receptor displays a more ordered activation loop with phosphorylation on Tyr774, and marginal changes in the glycine rich region to accommodate compounds such as staurosporine [38, 39]. Moreover, although the substrate-binding surface of EphB kinases is conserved, in the EphB4 kinase domain there are a few residues around this surface which are different from EphB2 receptor. These include Ala700, Ala793 and Ser825 in EphB4 kinase, which correspond to Ser711, Gln799 and Thr831, respectively, in EphB2. These differences may play a role in varied binding interactions with designed compounds [40].

Docking simulations have been utilized to determine drug ligand interactions with targeted receptors extensively [41–43]. For example, N-9 substituted 6-(4-(propoxyphenyl) piperazin-1-yl)-9H-purine derivatives were docked with P70-S6K1 and PI3K-δ kinases to examine their binding affinity and potential to inhibit those enzymes [44]. Furthermore, the kinase domain of EphB4 receptor was recently docked with the inhibitor NVP-BHG712 (NVP), and its regioisomer (NVPiso), revealing key interactions that consisted of hydrogen bonds with residues D758 (activation loop), M696 (hinge region), and T693 (gatekeeper) in addition to salt bridges with residues K647 and E664 [45].

In addition to docking studies, molecular dynamics (MD) has also been utilized to assess the stability of receptor-ligand binding [46–49]. For instance, recently MD simulations performed by Qin and co-workers provided vital information about the protein dynamics at Eph receptor-ligand interfaces in conjunction with crystallography and NMR studies [50]. In another study, molecular dynamics, alanine scanning and docking studies were utilized to develop a model for claudin-4–ephrinA2 complex which helped elucidate the residues involved in binding interactions [51].

The peptide sequence T-N-Y-L-F-S-P-N-G-P-I-A (abbreviated TNYL) obtained from phage display techniques has been shown to selectively inhibit ephrin B2 binding to the LBD of EphB4 receptor [52]. It has been shown that not only is the sequence TNYLFSPNGPIA an inhibitor of EphB4 but it also acts as a tumor homing peptide that can be utilized to target tumors. Thus, drug delivery vehicles have been prepared by conjugating TNYLFSPNGPIA to chitosan stearate micelles that successfully targeted SKOV-3 ovarian tumor cells which over-expressed EphB4 receptors [53]. The TNYL peptide sequence contains the FSPN motif corresponding to the Ephrin B2 GH loop which is the natural ligand for EphB4 within its binding cleft. The TNYL EphB4 binding peptide has been further modified to include a carboxyl terminal RAW sequence. The resulting TNYLFSPNGPIARAW peptide is a potent antagonist of ephrin-B2 binding to EphB4 [54]. The binding interactions of TNYL-RAW peptide was compared to the binding of ephrin-B2 with EphB4 receptor. Researchers found that while there was clear deviation at the J-K loop of the EphB4 structure, there was significant overlap in the binding of the two ligands with the main difference being that ephrin-B2 binding was entropically driven while TNYL-RAW binding was enthalpically driven.

In this work, we utilized “target-hopping” approach [55] which is a methodology in which known drug candidates or compounds that have been successfully shown to act as inhibitors or agonists for a specific receptor or protein are investigated for their binding interactions with another receptor or protein with the goal of developing novel drug candidates and also exploring their potential binding mechanisms. Using this methodology, Tognolini and co-workers discovered that lithocholic acid, a known ligand for the nuclear receptor FXR and the G-protein-coupled receptor TGR5, can also act as an antagonist of the EphA2 receptor by interfering with its kinase activation, and therefore, it may have potential in tumor targeting [56]. Given these results, researchers have further gone on to study compounds with similar structural motifs as lithocholic acid, such as betulinic acid and stilbene carboxylic acid, eprhinA1-EphA2 receptor interactions to develop inhibitors [57].

Here in, we conjugated the N-terminal of TNYLFSPNGPIA and its mutants with polyphenols that are known for their inherent anticancer properties, and investigated if these polyphenol-peptide conjugates could bind to the kinase domain of EphB4 and EphB2 receptors computationally. In previous work it has been shown that certain polyphenol rich botanical extracts can interfere with EphA2–ephrinA1 system and inhibit Eph-kinase phosphorylation [58]. Additionally, in a study conducted by Noberini and co-workers, it was shown that tea polyphenols can act as antagonists of the EphA4 receptor tyrosine kinase [59]. In previous studies, polyphenols such as cinnamic acid and its derivatives have been shown to be inhibitors of protein kinases, many of which belong to the tyrosine kinase families such as EGFR, BCR-ABL, IGF, VEGF [60].

We hypothesized that by conjugating selected polyphenols with the TNYL peptide and new mutants of TNYL, the conjugates may have potential to bind to the kinase domain of the EphB4 and EphB2 receptors. We utilized the TNYL sequence as our lead sequence for conjugating with three different polyphenols to examine the binding interactions with EphB4 and EphB2 receptors. We then designed three mutated sequences of TNYL peptide: SNYLFSPNGPIA (mutation 1), TNYLFSPNGPIG (mutation 2) and TNYLFTPNGPIA (mutation 3) and examined their binding interactions using docking studies to explore the effects of the point mutations on binding with the kinase domain of the receptors. Specific peptides were conjugated to gallic acid, p-coumaric acid, and sinapic acid by coupling the N-terminal amino group of the peptides with the carboxylic group of the polyphenols (Table 1). The anticancer potential of the mutated peptides was examined using the anti-CP webserver [61]. Prior to selecting the three mutated peptides, twenty mutated systems were created by mutating residues in various positions. The top scoring three-peptide sequences were selected. COSMOS-RS was then used to examine the hydrogen bond donor and acceptor capabilities of each of the compounds and changes in hydrophobicity when the peptides were conjugated with the polyphenols. Molecular dynamics studies were carried out with the conjugates showing the highest binding affinities obtained from docking studies to determine their stability within the receptors. To gain further insight into binding interactions, both the apo form and the ATP bound receptors were examined. This is because it has been shown that Eph kinase activity depends on the phosphorylation of the conserved tyrosine residue in the activation loop, which is common in several tyrosine kinases [62]. However, the molecular interactions within the kinase domains of Eph receptors are not yet fully elucidated. Hence, we utilized MD simulations to compare interactions with apo EphB4 and EphB2 kinase domains and the respective ATP-bound kinase domains. Furthermore, pharmacokinetic predictions of the conjugates that showed highest binding were explored in order to determine bioavailability, CYP interactions, MDCK permeability, GI absorption, and the LogP values using SWISS ADME and ADMETlab2.0 webservers. Our results indicated that the optimal combination of polyphenol-peptide conjugates for targeting the apo isoform of the EphB4 receptor is sinapte conjugated to TNYLFSPNGPIA, while sinapate conjugated to SNYLFSPNGPIA (mutation 1) interacted better with apo EphB2. However, the results for the ATP bound receptors were remarkably different. From MD simulations, it was evident that a more stable and greater binding with ATP bound EphB2 receptor occurred particularly with gallate-TNYLFSPNGPIA and sinapate-SNYLFSPNGPIA, which were found to attach to the activation loop region. For the ATP bound EphB4 receptor, the coumarate-TNYLFSPNGPIA conjugate formed the most stable complex. MMGBSA calculations based on the trajectory analysis of the MD simulations revealed that for the apo EphB4 receptor, sinapate-TNYLFSPNGPIA conjugate had the most negative free energy. Overall, the polyphenol-peptide conjugates showed significantly more negative free energies compared to the neat peptide counterparts in almost all cases. For the ATP bound EphB2 receptor, Gallate-SNYLFSPNGPIA, Sinapate-SNYLFSPNGPIA, and coumarate-TNYLFSPNGPIA showed the most negative free energies. Given that there is a decrease in the free energies for the ATP bound EphB4 receptor, it is likely that the conjugates bind better to the apo form, and ATP binding may reduce some of the interactions with the binding pocket residues and may be not be as energetically favorable for ATP bound EphB4. Overall, this study demonstrates the significance of in silico characterization of the binding interactions of the designed peptide-polyphenols with EphB4 and EphB2 receptors in their apo and ATP bound forms. Our results indicated that conjugating polyphenols with the mutant (SNYLFSPNGPIA) of the known tumor targeting peptide TNYL may also allow for strong binding interactions with the kinase domain of Ephrin B4 receptor. Additionally, as a proof concept we tested the binding interactions of gallate-TNYLFSPNGPIA and sinapate-TNYLFSPNGPIA with both receptors using SPR analysis. Our results revealed that the conjugates showed higher binding with the EphB4 receptor. Therefore, further in vitro studies with the various conjugates may lead to the development of potential therapeutics.

Table 1.

Designed Peptide-polyphenol conjugates

Methods

Peptide sequence generation (ACPP, C scores)

The original peptide sequence TNYLFSPNGPIA was selected from literature for its ability to target the EphB4 receptor as mentioned in the previous section. We then incorporated several mutations into the peptide, and their properties were determined using the “AntiCP webserver for Designing and prediction of Anticancer Peptides” (ACPP). Of those the top three scoring peptides were selected. The sequences consist of TNYLFSPNGPIA (original), SNYLFSPNGPIA (mutation 1), TNYLFSPNGPIG (mutation 2) and TNYLFTPNGPIA (mutation 3), where the residue that has been mutated indicated in bold. ACPP analyzes peptide sequences for the presence of apoptotic domains with improved accuracy compared to previous anticancer peptide prediction tools, and determines properties such as SVM score, hydrophobicity, hydropathicity, hydrophilicity, pI, and molecular weight [63]. After selecting three mutations for a total of four peptide sequences, each sequence was submitted to the I-TASSER database for prediction of secondary structure, C-scores, and solvent accessibility of each residue. The I-TASSER server is used to predict 3D structures of proteins for computational studies and compares structures of known PDB files of similar proteins, producing values for topological similarity (TM-score) and RMSD with predicted ligands [64–66].

Conjugate design

The peptides and the polyphenol-peptide conjugates were designed by utilizing ChemDraw and ChemDraw 3D (19.1). The polyphenols gallic acid, sinapic acid and p-coumaric acid were attached to the N-terminal of the peptides by conjugation with the carboxyl group of the polyphenols. ChemDraw 3D was utilized in order to run the “MM2” energy minimization calculations. Then, the structures were visualized using PyMOL (2.4.0) [67] to determine hydrogen bonds and ensure no errors were found in the building process.

COSMOtherm (sigma profiles)

To better understand the H-bond donor–acceptor capabilities of the gallic-, p-coumaric, and sinapic-peptide conjugates, we examined their σ-profiles using Turbomole and COSMOtherm. A σ-profile is the probability function of finding a part of the molecule with a specific screening charge density values from − 0.03 to 0.03 where the negative values represent positive polarities and the positive values represent negative polarities [68]. The σ-profile can be divided into three regions, each corresponding to the hydrogen bond donor region which is represented by σ < − 0.0082 e/Ų, the hydrogen bond acceptor region which is represented by σ >  + 0.0082 e/Ų, and the region between − 0.0082 e/Ų and + 0.0082 e/Ų, which represents the non-polar region [69, 70].

Cleft determination (POCASA and metaPocket)

To determine surface cavities and clefts of the EphB4 and EphB2 receptors, we utilized the web server POCASA. The receptor PDB files were downloaded from RCSB website and prepared for POCASA by eliminating water molecules and ligands using PyMOL. PDB ID: 6FNL [71] was utilized for the EphB4 receptor and PDB ID: 3ZFM [38], representative of the crystal structure of the kinase domain of EphB2 receptor were used. The structures were uploaded into the server and then selected for analysis. Each cleft was filled in with markers to indicate their location and volume within the receptor. POCASA determines this volume by inserting spheres of differing radii in between atoms in the receptor, before adding a rolling probe sphere which rolls along the surface of the protein to detect the location and volume of pockets [72]. POCASA also provides output files for the parameters used and depth centers of each pocket are detected. For both receptors, the standard parameters were used, with a probe radius of 2 angstroms, an SPF of 16, a PDF of 18, an unlimited maximum number of cavities, and a grid size of 1 angstrom. To determine potential active sites for both receptors, we utilized the metaPocket web server. MetaPocket combines input from several different software, including PASS11, LIGSITE, QSiteFinder, GHECOM, POCASA, Fpocket, SURFNET, and ConCavity, then interprets the data to predict the top three most likely binding sites of the receptor [73]. This web server also provides output files for predicted binding site residues that will interact with ligands.

Docking studies

We conducted docking studies using two different software, AutoDock Vina and DockThor in order to compare the validity of the results. AutoDock Vina is a computational molecular docking tool that utilizes multithreading to improve the speed and scoring function of the study compared to other software such as AutoDock 4 [74]. Samples were prepared for docking utilizing AutoDockTools- 1.5.7 to create .pdbqt files for use in the PyMOL plugin. For the apo receptors (without ATP), the receptors, obtained from the RCSB database (PDB IDs 6FNL and 3ZFM), were prepared by deleting any ligands, water molecules, adding polar hydrogens, and adding Kollman charges. The ligands were prepared by selecting them for AutoDock Vina, which then determines the number of rotatable bonds. Grids with coordinates (35.510, 0.721, 60.063) for the EphB4 receptor and (− 0.087, 10.223, 5.088) for the EphB2 receptor were created. Eight iterations were performed for each sample with an energy range of 4. All compounds were run with the EphB4 receptor, and the five trials with the highest binding affinity were repeated with the EphB2 receptor. The grid was a 40 × 40 × 40 angstrom cube at the predicted binding site. Results were visualized utilizing PyMOL 2.4.0. For the ATP bound receptors, the receptors were saved as .pdbqt, while the ATP was saved as a ligand, and the parameters for EphB2: − 0.09, 10.22, 5.09 and EphB4: 35.51, 0.72, 60.06, for active site binding were generated through a grid box generation in AutoDockTools. Docking study with the receptor and ATP were carried out to ensure proper attachment of the ATP inside the binding pocket for the EphB2 and EphB4 receptors. Once docking was complete, the output .pdbqt file generated displayed the ATP ligand with aligned rotatable bonds. This structure was then incorporated with the receptor and saved as a single structure using the PyMol. These ATP-bound receptor structures then underwent receptor preparation again through AutoDock Tools with the addition of polar hydrogens and Kollman charges. The polyphenol conjugates and the two most optimal peptides were then docked to the ATP-bound receptors, and the grid box with the same parameters as seen in the first docking analysis were maintained. The optimal conjugates were analyzed for both the EphB2 and EphB4 receptor and eight iterations were calculated for each analysis. The highest binding affinity number from each was recorded and the interactions were visualized using PyMol.

For running docking studies using DockThor .pdb files of each of the ligands and the apo or ATP bound EphB4 and EphB2 receptors were uploaded on the server [75]. This software uses empirical scoring functions to determine protein–ligand binding affinity by examining interactions contributing to the binding free energy using MMFF94S force field. The PdbThorBox program is used to set the protein atom types and the partial charges from the MMFF94S force field. DockThor scoring functions incorporate classical van der Waals and electrostatic energy terms, optimized terms accounting for solvation, lipophilic protein–ligand interactions and an estimation of ligand torsional entropy contribution to ligand binding for better description of protein–ligand recognition [76]. Each study produced ten different docking poses and the top pose was selected based on lowest energy.

Receptor-ligand interactions (PLIP)

To determine interactions between each ligand and the EphB4 and EphB2 receptors, either ATP bound or apo complexes with the polyphenol-peptide conjugates and the neat peptides were exported from PyMOL as .pdb files. These complexes were submitted to the Protein–Ligand Interaction Profiler (PLIP) web server for analysis. PLIP detects non-covalent interactions between ligands and receptors, including hydrophobic interactions, hydrogen bonds, salt bridges, pi-stacking, and halogen bonds [77]. The data for each docked complex was exported into a spreadsheet for analysis of common residues and unique interactions.

Molecular dynamics

To assess the stability of the receptor-ligand complexes, molecular dynamics (MD) simulations were carried out. MD simulations were performed using DESMOND through the Maestro 12.5 Suite [78, 79]. For the Apo complexes, EphB4 and EphB2 receptors were prepared using the protein preparation wizard within Maestro 12.5 following the addition of hydrogens to ensure the atom identifications were in the proper format needed to run DESMOND. The resulting complexes were used to design the systems used for molecular dynamics. This procedure was then performed with the peptides TNYLFSPNGPIA (original) and SNYLFSPNGPIA (mutation 1) and their polyphenol conjugates. The systems for molecular dynamics simulations were designed in an orthorhombic box with minimized volume and in the SPC (simple point charge) water model. The systems were neutralized with the addition of Na⁺ and Cl⁻ ions. The simulations were run with an NVT class ensemble and a time step integration of 2 fs. The recording interval for the trajectory was 100 ps, and 1.2 ps was used for the energy. The simulations were run for 100 ns using the OPLS3e force field with force restraints on Cα atoms to prevent denaturation of the receptor. The systems were heated from 10 to 400 K in 300 ps, then were brought down to 300 K at 1000 ps for the remainder of the simulation. Results were analyzed using the Simulation Interactions Diagram feature in Maestro 12.5 to assess values such as RMSD, RMSF, and protein–ligand interactions. For the ATP bound receptors, similarly, the protein preparation wizard application was used to add optimized hydrogens to the receptors for proper alignment prior to starting molecular dynamics. simulation time was set at 100 ns. Once molecular dynamics was complete, the trajectory frames and output files were analyzed using the Simulation Interactions Diagram application in the Maestro Suite and the conjugates and the peptides were specifically selected when prompted for protein–ligand interactions, RMSD, and RMSF data analysis.

MMGBSA calculations

To assess the theoretical free energies of binding of the peptides and the polyphenol peptide conjugates with EphB2 and EphB4 receptors (both apo and ATP bound), we utilized the molecular mechanics generalized Born surface area (MM-GBSA) method to calculate the relative binding free energies of the trajectories [80]. The free energy of binding can be calculated as ΔG_(bind) = ΔG_(solv) + ΔE_(MM) + ΔG_(SA) where ΔG_solv is the difference in solvation energy of the ligand-bound receptor complex and the sum of the solvation energies for the free ligand and receptor. ΔE_MM is the difference in the minimized energies between receptor-ligand complex and the sum of the energies of each of the free ligands and receptor. ΔG_(SA) is the difference in surface area energies of the ligand-receptor complex and the sum of the surface area energies for each of the neat ligands and receptor. The polar effect of the free energy is assessed by a generalized Born model with an external dielectric constant of 80 and an internal dielectric constant of 1, while the non-polar energy contribution is calculated from the solvent accessible surface area (SASA) [81].

Prediction of pharmacokinetic properties

To analyze the toxicity and pharmacological properties of the conjugates, the conjugates and the peptides were analyzed using web servers SwissADME [82] and ADMETlab 2.0 [83]. These servers predict the absorption, distribution, metabolism, and excretion of drug candidates. SwissADME provides values for lipophilicity, water solubility, interactions with cytochromes P450, classification of substrate or non-substrate of permeability glycoprotein Pgp, and gastrointestinal absorption. The server utilizes set parameters for comparison to the input molecule, with the calculations performed by several freely available methods, including OpenBabel API, CHARMM, and FILTER-IT. ADMETlab 2.0 on the other hand also provides additional information regarding cell permeability, and whether the drug candidates may be hERG blockers in addition to several other aspects of the drug candidates.

Laboratory methods

Materials

The peptide TNYLFSPNGPIA was custom ordered from GenScript (Piscataway, NJ, USA). N-hydroxysuccinimide (NHS), 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDAC), 80% azelaic acid, and Dimethylformamide (DMF) were purchased from Sigma-Aldrich (St. Louis, MO, USA). 1X Dulbecco’s Phosphate Buffered Saline (PBS) was purchased from ATCC (Manassas, VA, USA). Human EphB2 and EphB4 were purchased from Sino Biological (Wayne, PA, USA). The Biotinylated Human EphB4 Protein was purchased from Acro Biosystems (Newark, Delaware, USA). Streptavidin was purchased from Millipore Sigma (Burlington, MA, USA). Gold-coated Surface Plasmon Resonance chips were purchased from Platypus Technologies (Fitchburg, WI, USA). Gallic acid was purchased from Chem-Impex International INC. (Wood Dale, IL, USA). Sinapic acid and 1,1-Mercaptoundecanoic acid were purchased from Santa Cruz Biotechnology, (Dallas, TX, USA). The ADP-Glo TM Kinase Assay was purchased from Promega Corporation. Poly (Glu, Tyr) substrate was purchased from Sigma-Aldrich.

Synthesis of conjugates

As a proof of concept, we synthesized two conjugates gallate-TNYLFSPNGPIA and sinapate-TNYLFSPNGPIA. To prepare the conjugates, we utilized established peptide coupling methods [84]. Briefly, the carboxyl group of gallic acid (0.07 M) or sinapic acid (0.07 M) was activated with 1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDAC) (0.05 M) and N-Hydroxysuccinimide (NHS) (0.05 M) in a solution of dimethylformamide (DMF) at 4 °C for one hour. Then 0.05 M of the peptide sequence was added to each reaction mixture and shaken for 48 h at 4 °C. After 48 h, the reaction mixtures were rotary evaporated and a dark yellow product was obtained. The products obtained were then recrystallized from acetone and stored at 4 °C for further analysis. The formation of the products was confirmed by ¹H NMR spectroscopy carried out using a Bruker 400 MHz NMR Spectrometer. Samples were prepared in dimethyl sulfoxide-d6 (DMSO-d6) solvent with 0.03% TMS for analysis. For the gallate-T-N-Y-L-F-S-P-N-G-P-I-A conjugate, ¹H NMR peaks were seen at δ 9.6 (s, 2H); δ 9.2 (s, 1H); δ 9.1 (s, 1H); δ 8.9 (s, 1H); δ 8.8 (s, 1H); δ 8.3 (s, 8H); δ 7.3 (s, 1H); δ 7.2 (d, 4H); δ 7.1 (s, 1H); δ 7.0 (s, 2H); δ 6.7 (s, 2H); 5.5 (s, 1H); δ 5.1 (s, 1H);δ 4.9 (t, 1H); δ 4.8 (m, 4H); δ 4.7 (t, 2H); δ 4.6 (m, 2H); δ 4.5 (m, 4H); δ 4.4 (t, 2H); δ 4.3 (d, 2H); δ 4.2 (d, 2H); δ 4.0 (s, 2H); δ 3.7 (t, 2H); δ 3.5 (m, 4H); δ 3.4 (m, 2H); δ 3.2 (d, 2H); δ 2.8 (t, 2H); δ 2.6 (s, 2H); δ 2.5 (d, 3H); δ 2.1 (m, 4H); δ 2.0 (t, 2H); δ 1.8 (d, 5H); δ 1.4 (d, 3H); δ 1.2 (m, 2H); δ 1.0 (d, 3H) δ 0.8 (d, 3H).

For the sinapate-T-N-Y-L-F-S-P-N-G-P-I-A conjugate, ¹H NMR peaks were seen at δ 9.2 (s, 1H); δ 9.1 (s, 1H); δ 8.9 (s, 1H); δ 8.3 (s, 8H); δ 7.4 (s, 1H); δ 7.3 (s, 1H); δ 7.2 (d, 4H); δ 7.0 (s, 3H); δ 6.8 (s, 2H); δ 6.7 (s, 2H); δ 6.6 (s, 2H); δ 6.5 (s, 1H); 5.5 (s, 1H); δ 5.1 (s, 1H); δ 4.9 (t, 1H); δ 4.8 (m, 5H); δ 4.7 (t, 2H); δ 4.6 (t, 1H); δ 4.5 (m, 4H); δ 4.4 (t, 2H);δ 4.3 (d, 2H); δ 4.2 (d, 2H); δ 4.0 (s, 2H); δ 3.7 (t, 2H); δ 3.6 (s, 6H); δ 3.5 (m, 4H); δ 3.4 (m, 2H); δ 3.2 (d, 2H); δ 2.8 (t, 2H); δ 2.6 (s, 2H); δ 2.5 (d, 3H); δ 2.1 (m, 4H); δ 2.0 (t, 2H); δ 1.8 (d, 5H); δ 1.4 (d, 3H); δ 1.2 (m, 2H); δ 1.0 (d, 3H); δ 0.8 (d, 3H).

Surface plasmon resonance (SPR)

Gold chips were functionalized in accordance with previously established methods [85]. The gold-coated SPR chips were cleaned for coating by soaking in ethanol and then placing them under UV light for 10 min. The chips were first coated in 200 μL of 1,1-mercaptoundecanoic acid (0.5 M) solution dissolved in ethanol. After 1 h, the 1,1-mercaptoundecanoic acid was activated with N-hydroxysuccinimide solution (0.1 M) followed by the addition of EDAC (0.1 M). The chips were allowed to sit at 4 °C for one hour in moist tissue paper. Then the activated chips were coated with EphB2 receptor by incubating with EphB2 receptor (10 μg/mL) solution in deionized water. The chips were allowed to sit at 4 °C for four hours prior to use. For coating chips with the biotinylated EphB4 receptor, the activated chip surface was first treated with streptavidin solution (7 μg/mL). Streptavidin’s high affinity for biotin is well known [86]. After refrigeration for two hours, 200 μL of a 10 μg/mL solution of biotinylated EphB4 in deionized water was added to the chip surface and refrigerated for four hours prior to use. Before running with the SPR, the chips were spotted with one drop of Cargill’s 7.21 index fluid, arranged on the SPR prism, and placed into the instrument. The optimal angle was adjusted individually for each chip. First, 1X PBS buffer solution was run to obtain the baseline value, followed by the test compounds. Chips were switched between each run and concentration to maintain consistency. We tested binding interactions with different concentrations (10 nM, 50 nM, 75 nM, and 100 nM) of each conjugate for the SPR experiments. The Gallate-TNYLFSPNGPIA and Sinapate- TNYLFSPNGPIA conjugates were analyzed with both the EphB2 and EphB4 receptors. The TNYLFSPNGPIA peptide was also analyzed as a control. Each SPR analysis was performed in triplicate. The flow-rate was kept constant at 30 μL/min.

ADP-glo kinase assay

We performed the ADP-Glo Kinase Assay to assess the kinase activity of EphB2 and EphB4 in the presence of the conjugates. Inhibition of kinases would result in a lower luminescence reflected by the assay as compared to kinase substrate controls [87]. Assays were conducted in 96-well non-binding white costar plates. DMSO was used as a vehicle control keeping all other conditions the same. All solutions were prepared in a buffer of 0.1 mg/mL BSA, 4% Tris–HCl, and 5 mM MgCl₂. After optimization of the concentration of the substrate and kinase, the assay was conducted in triplicate. To each well, 8 μL of the 5 μg/mL kinase substrate Poly (Glu, Tyr), 5 μL of EphB4 or EphB2 protein (10 nM) and 10 μL of ATP (80 μM) were added. In addition to DMSO control, we also included control wells without kinase enzyme to probe background luminescence. A separate control with no inhibitors or substrates was also prepared which contained 5 μL of either EphB4 or EphB2, 10 μL ATP, and 10 μL buffer. The plates were incubated at room temperature for 2 h in the dark under moist tissues. After this period, 25 μL of the ADP-Glo Reagent was added to each well, and the plate was once again incubated at room temperature. After 40 min, 50 μL of the Kinase Detection Reagent was added to each well and allowed to incubate at room temperature for 30 min. The luminescence was then read using a Biotek Synergy H1 Fluorescence Microplate Reader. The readings were set to Auto Gain with a read height of 4.5 and a speed of 5.00 s per well.

Results and discussion

Designing anticancer peptides

The generation of point mutations was carried out using the ACPP program by entering the initial targeting sequence and searching for a point mutation with similar properties. We initially designed twenty sequences. Based on the scores, we then selected the top three mutations as shown in Table 2 which includes sequences where the original sequence TNYLFSPNGPIA was mutated at positions 1, 12 and 6. Thus the total number of sequences was narrowed down to four sequences. All sequences shared the same charge and pI, with the most variation being in the support vector machine (SVM) score, which is the parameter used to classify sequences as anticancer peptides [88]. Based on the results, the SVM score of the peptides ranged from 0.56 to 0.61 with TNYLFTPNGPIA (mutation position 6) displaying a SVM score of 0.56 while the original peptide TNYLFSPNGPIA showed a SVM score of 0.61.

Table 2.

Peptide Sequence ACPP Results

Peptide Sequence	Mutation Position	SVM Score	Hydrophobicity	Hydropathicity	Hydrophilicity	pI
TNYLFSPNGPIA	None	0.61	0.04	− 0.04	− 0.72	5.88
SNYLFSPNGPIA	1	0.58	0.03	− 0.05	− 0.66	5.88
TNYLFSPNGPIG	12	0.61	0.03	− 0.23	− 0.67	5.88
TNYLFTPNGPIA	6	0.56	0.04	− 0.03	− 0.78	5.88

N.B. Highlighted amino acid indicates point mutation

Peptide sequence secondary structures and C-scores

Analysis of secondary structure and confidence intervals for each of the four peptide sequences was performed with I-TASSER [89]. The results are summarized in Table 3. The secondary structure of the original peptide TNYLFSPNGPIA and the sequences SNYLFSPNGPIA (mutation 1), TNYLFSPNGPIG (mutation 2), and TNYLFTPNGPIA (mutation 3) appear to be the similar, consisting of the mostly coils with two strands (CCSSCCCCCCCC). The data for C-scores indicates that the secondary structure for TNYLFTPNGPIA (mutation 3) had the highest degree of accuracy, with a C-score of − 0.56. The C-scores are given in a range of [-5, 2] and the closer to the upper limit, the stronger indication of high confidence in the structure [90].

Table 3.

Peptide Sequence I-TASSER Results

Peptide Sequence	Secondary Structure	Solvent Accessibility	C Score
TNYLFSPNGPIA	CCSSCCCCCCCC	753,434,473,448	− 0.7
SNYLFSPNGPIA	CCSSCCCCCCCC	863,434,473,448	− 0.69
TNYLFSPNGPIG	CCSSCCCCCCCC	753,434,473,548	− 0.91
TNYLFTPNGPIA	CCSSCCCCCCCC	853,433,473,448	− 0.56

Sigma profiles

To explore the behavior of polyphenol-peptide conjugates, sigma-profile plots were plotted using COSMOS-RS. The gallate, sinapate, and p-coumarate-peptide conjugates of TNYLFSPNGPIA (original), and with mutations 1, 2 and 3 in comparison with neat polyphenols are shown in Fig. 1. The sigma profiles of the neat polyphenols show the greatest variation in the nonpolar region (Fig. 1a). Furthermore, the sigma profiles showed shorter peaks in the non-polar region. The sinapic acid and p-coumaric acid showed peaks that had a slightly higher values in the non-polar region, whereas the gallic acid was found to be least hydrophobic due to the presence of the three –OH groups. In addition, gallic acid appears to have stronger peaks in both the acceptor (HBA) and the donor (HBD) region. This is likely caused by the increased number of hydroxyl groups containing lone pairs on the oxygen moiety, giving gallic acid (GA) a higher ability to form H-bonds. Sinapate showed the highest hydrophobicity due to the presence of methoxy groups in addition to the aromatic ring and the alkenyl moiety. For the conjugates, the sigma profile probability density appears within the non-polar region (Fig. 1b–e). Additionally, there are broad peaks in both the HBA (hydrogen bond acceptor) and the HBD (hydrogen bond donor) regions. This suggests the conjugates have strongly polarizable hydrogens and HBDs in their structures due to the amide groups of the peptide moieties. The H-bond donor region varies more than the acceptor region. The broader peak seen for gallic-peptide conjugate of SNYLFSPNGPIA (mutation 1) in the HBA region shows a slightly higher ability to hydrogen bond, which is likely due to the presence of -OH groups present on the gallate moiety and the –OH group of serine. For all three conjugates the majority of the binding capabilities are hydrophobic, but important contributions in the H-bond regions also co-exist. These properties are desirable for binding to the kinase domain of the receptors which have hydrophobic binding pockets in addition to polar residues [91].

Fig. 1.

Sigma profiles of the neat a polyphenols; b polyphenol-conjugated TNYLFSPNGPIA c polyphenol-conjugated SNYLFSPNGPIA (mutation 1); d polyphenol-conjugated TNYLFSPNGPIG (mutation 2); e polyphenol-conjugated TNYLFTPNGPIA (mutation 3). The surface properties of each of the corresponding materials are shown in the figures. Red is indicative of /H-bond acceptor region; Blue indicates H-bond donor region and Green is indicative of hydrophobic, non-polar region

Similar results were obtained for gallate, p-coumarate, and sinapate peptide conjugates of TNYLFSPNGPIA (original), TNYLFSPNGPIG (mutation 2), and TNYLFTPNGPIA (mutation 3). For each of the conjugates, the nonpolar region is significant, suggesting that the molecules are largely hydrophobic. Of note, among conjugates, the gallic-peptide conjugates show the lowest peak compared to the p-coumarate-peptide and sinapate-peptide conjugate in the nonpolar region. Overall, the results indicated that the peptide-polyphenol conjugates display hydrophobicity along with H-bonding capability that would allow for binding with the kinase domain of EphB4 and EphB2 receptors.

Cleft determination

To determine the binding pockets of EphB4 and EphB2 receptors, we first analyzed the known structures of the receptors using the POCASA web server. The results are shown in Fig. 2. As shown in the figure, the surface of the EphB4 receptor consists of 9 pockets, (Fig. 2a). Among these pockets, the predicted binding site was measured to have a volume of 245 A°, and a volume distribution (VD) value of 633 (Table 4). The VD value serves as a measure of the volume depth to determine the pocket which has the greatest capability of being the binding site and has been shown to be a better predictor than volume alone. The results of the EphB2 receptor (Fig. 2b) indicated the presence of 8 pockets and 1 surface cavity. The predicted binding site was shown to have a volume of 249 A° and a VD value of 606. These findings indicate similarities within the structures of the two EphB receptors.

Fig. 2.

POCASA images of a EphB4 and b EphB2 receptors. Magenta markings indicate a cavity on the surface of the receptor, with an increased number of markings representing a larger volume. MetaPocket images of c EphB4 and d EphB2 receptors. Each pink sphere represents three most likely cavities to be considered as the binding site

Table 4.

POCASA Values for EphB4 and EphB2 Receptors

EphB4				EphB2
Rank Number	Pocket Number	Volume (Å)	VD	Rank Number	Pocket Number	Volume (Å)	VD
1	67	245	633	1	192	249	606
2	307	72	185	2	185	45	380
3	16	65	182	3	337	89	251
4	352	78	181	4	16	82	199
5	311	33	79	5	305	52	140
6	167	27	66	6	73	21	115
7	198	21	54	7	85	33	76
8	315	22	52	8	253	27	64
9	93	20	45

The results of the metaPocket studies predicted three potential ligand binding sites for each receptor at the predicted active site (Fig. 2c and d). In addition, the predicted pockets obtained from metaPocket for both the EphB4 and EphB2 receptors were identified by the POCASA server, indicating agreement across multiple analyses of the surface structure of each receptor.

To identify the active site, the metaPocket server compiled a list of predicted residues involved in receptor–ligand interactions, which is summarized in Table 5. The list of residues for EphB4 consisted of 42 residues, including the residues R744, A700, I621, T693, D702, and S703. The analysis of EphB2 active site residues from metaPocket yielded a list of 43 residues, including residues F667, E670, D764, S763, R750, S709, and S706. The results from metaPocket consisted of residues capable of forming hydrogen bonds, hydrophobic interactions, and pi-stacking interactions. This further indicates that the active site contains various residues within the juxtamembrane region and kinase domain capable of multiple interactions and was located in the pocket identified by both metaPocket and POCASA [92].

Table 5.

metaPocket Residues for EphB4 and EphB2 Receptors

EphB4		EphB2
Active Site Residues (metaPOCKET)	Residues Shared with PLIP	Active Site Residues (metaPOCKET)	Residues Shared with PLIP
F626 Q657 I782 F797 P783 G760 R739 S663 F661 T787 R659 D740 H738 F759 Y736 V737 G622 M668 I677 A645 G699 E694 N698 E697 G627 E628 T648 E660	E625 K781 D758 R744 S757 A700 I621 T693 L747 D702 S703 F695 M696 R631	S669 F765 I697 L668 M674 I652 E700 V635 T654 N704 S637 F710 R745 E634 Q713 V743 L655 D746 G705 V626 Y742	F632 D666 F667 E670 D764 T699 I683 S763 K653 L753 A651 N751 M702 F701 G633 R750 C636 S709 H744 S706 E703 I 627

Molecular docking studies

To analyze the binding affinities and interactions between the designed peptides and the peptide polyphenol conjugates with EphB4 and EphB2 receptors (both ATP bound and apo forms), molecular docking studies were performed using two different docking programs. In previous work it was shown that using multiple docking software can improve the fidelity of results and allows for specific scoring and obtaining the best protein–ligand-binding models [93]. We, therefore, compared the binding affinities obtained using Dockthor and Autodock Vina 1.2.0. Results of the binding affinities for each of the peptides and conjugates for the apo form using both software are summarized in Table 6 and the corresponding results for the ATP bound receptors with the conjugates are shown in Table S2 (Supplementary information). The resulting images of the docking studies for the ATP bound and the apo receptors obtained using autodock vina with the TNYLFSPNGPIA sequence with EphB4 and EphB2 receptors are shown in Fig. 3 while that some of the optimal polyphenol-peptide conjugates with EphB4 and EphB2 receptors are shown in Fig. 4.

Table 6.

Binding Affinities of peptides and peptide polyphenol conjugates for EphB4 and EphB2 Receptors (unphosphorylated, APO form)

Peptide Sequence conjugated	Polyphenol		Polyphenol		Polyphenol
	Gallic Acid Binding Affinity (Kcal/mol)		p-Coumaric Acid Binding Affinity (Kcal/mol)		Sinapic Acid Binding Affinity (Kcal/mol)		Neat Peptide (unconjugated) Binding Affinity (Kcal/mol)
	Autodock vina	Dock Thor	Autodock Vina	DockThor	Autodock Vina	DockThor	Autodock Vina	DockThor
EphB4
TNYLFSPNGPIA (original)	− 8.9 − 8.4		− 8.5 − 9.4		− 9.6 − 10.1		− 8.4 − 8.5
SNYLFSPNGPIA (mut1)	− 9.8 − 9.4		− 9.1 − 8.6		− 9.7 − 9.2		− 8.9 − 8.4
TNYLFSPNGPIG (mut2)	− 9.1 − 9.2		− 9 − 9.2		− 8.7 − 9.1		− 8.6 − 8.5
TNYLFTPNGPIA (mut3)	− 8.9 − 9.3		− 8 − 9.0		− 7.7 − 8.8		− 9.1 − 8.9
None	− 5.8 − 6.2		− 6.1 − 6.4		− 6.1 − 7.2		− −
EphB2
TNYLFSPNGPIA (original)	− 7.6 − 8.7		− 7.5 − 8.6		− 7.4 − 5.6		− 7.1 − 7.8
SNYLFSPNGPIA (mut1)	− 7.6 − 9.2		− 7.7 − 7.9		− 7.7 − 8.9		− 7.3 − 6.8
TNYLFSPNGPIG (mut2)	− 7.5 − 7.9		− 6.9 − 7.7		− 7.9 − 7.9		− 7.5 − 7.8
TNYLFTPNGPIA (mut3)	− 7.1 − 8.5		− 6.8 − 8.5		− 6.5 − 5.7		− 7.0 − 8.3
None	− 5.7 − 6.3		− 5.8 − 6.2		− 5.5 − 6.7		-

Fig. 3.

TNYLFSPNGPIA docked with a ATP bound EphB4 and b Apo EphB4; c ATP bound EphB2 and d Apo EphB2. The red ball and stick structure indicates ATP (Figs. 3a and 3c)

Fig. 4.

Polyphenol-peptide conjugates docked with EphB4 (a through h) and EphB2 (i through p). a gallate-TNYLFSPNGPIA b sinapate-TNYLFSPNGPIA c gallate-SNYLFSPNGPIA and d sinapate-SNYLFSPNGPIA e gallate-TNYLFSPNGPIA f sinapate-TNYLFSPNGPIA; g gallate-SNYLFSPNGPIA; h sinapate-SNYLFSPNGPIA. i gallate- TNYLFSPNGPIA; (j) sinapate-TNYLFSPNGPIA; k gallate-SNYLFSPNGPIA; l sinapate-SNYLFSPNGPIA; m gallate-TNYLFSPNGPIA; n sinapate-TNYLFSPNGPIA; o Gallate-SNYLFSPNGPIA; p Sinapate-SNYLFSPNGPIA. Blue Ribbon structures represents Apo form of EphB4 receptor and golden ribbon structures represent ATP bound EphB4 receptor. Green Ribbon structures represents Apo form of EphB2 receptor and purple ribbon structures represent ATP bound EphB2 receptor. ATP is represented by ball and stick model in red, while conjugates are represented in green for ATP bound receptors

As shown, from the docking results of the apo receptor obtained from Autodock Vina, gallate-, p-coumarate-, and sinapate-SNYLFSPNGPIA (mutation 1) conjugates were found to have higher binding affinities for EphB4 receptor. Similar trends were observed when the mutated peptide mutation 2 was conjugated with the polyphenols. Additionally, all conjugates showed higher binding affinities compared to unconjugated polyphenols. Comparable tendencies were observed for DockThor; however, in the case of DockThor, the sinapate-TNYLFSPNGPIA conjugate was also found to have high affinity. The slight differences between DockThor and autodock vina results could be attributed to the fact that the properties of the binding pocket and the target molecules could affect the outcome of the docking software. In previous work, it has been shown that while DockThor is a robust docking software capable of providing accurate binding modes for large peptides, however, the performance of DockThor decreases as the number of residues in the sequence and the degree of flexibility increases [94]. This may contribute to the differences observed. Thus, for the apo isoform, the highest binding affinity for the EphB4 receptor was seen for gallate-SNYLFSPNGPIA conjugate at -9.8 kcal/mol using autodock Vina, and sinapate- TNYLFSPNGPIA with DockThor at -10.1 kcal/mol for the EphB4 receptor. To further elucidate these docking studies results we carried out MMGBSA studies of the best-docked poses to estimate the binding affinities, which are discussed later in the manuscript. In the case of the ATP bound EphB4 receptor, the results showed comparable trends as that of the apo receptor as per autodock vina results with the gallate-SNYLFSPNGPIA conjugate showing the highest binding affinity at -8.7 kcal/mol. However, the para-coumarate-TNYLFSPNGPIA conjugate showing highest binding affinity with DockThor at -10.8 kcal/mol, followed by the mutation 1 peptide (SNYLFSPNGPIA) at -10.7 kcal/mol. Interestingly, the binding affinities observed for the conjugates were not as significantly increased compared to the original peptide and, in fact, showed a minor decrease upon conjugation with p-coumarate and p-sinapate according to autodock results for the ATP bound receptor. This is likely because in the presence of ATP, the active site chain adopts a different conformation, which in some cases may lead to slight decrease in binding affinity due to competitive binding with the ATP binding site. These slight conformational changes may have an effect on receptor interactions with each of the conjugates and the peptides. As shown in Fig. 3a, the original peptide was found to make key H-bond interactions with Met 696 (hinge region) along with hydrophobic interactions with Arg 744 and Asn 745 (activation loop), indicating that the peptide is attached to the catalytic domain and likely may compete with ATP binding [95]. The regions within the connecting loops are responsible for interactions with ATP in the active site, while the C-lobe interacts with the peptide substrate. In the case of the original peptide, the binding appears to span the catalytic site as well as the C-lobe region of the receptor. For the apo receptor (Fig. 3b) the peptide was found to share some similar interactions with Arg 744 and Asn 745 as that of the ATP bound receptor, however, additional interactions are seen with A623, R706 and K547, which were not seen for the ATP bound receptor.

In a recent study, receptor-based pharmacophore models indicated that compounds such as GK03503 with aromatic systems containing electron donating methoxy groups displayed inhibitory effects toward EphB4 receptor [96]. Thus, the interactions may be similar to that of the functional group of sinapate moiety. In another study, potential competitive inhibitors of the EphB4 receptor generated by an anchor-based tailoring library approach showed that the presence of phenyl ring was critical for binding to EphB4 receptor [97]. These results corroborate with our findings which show relatively higher binding affinity with select polyphenol-peptide conjugates utilized in the current study and emphasize the importance of aromatic ring systems in interacting with the EphB4 receptor active site which contains Phe and Tyr residues. The sequences TNYLFSPNGPIA and SNYLFSPNGPIA (mutation 1) differ at position one with a threonine to serine point mutation. In one study, researchers found that mutating the Thr301 residue of P-450 (16α) to Ser increased hydroxylase activity toward the ligand progesterone, indicating that this particular substitution was capable of altering the binding affinity and interactions of receptor and ligand [98]. In another study, the mutation of Thr to Ser in a proteasome led to a significant decrease in the K_m value for the enzyme, indicating a faster and stronger binding of the substrate to the enzyme [99]. These findings agree with the results of the current study where the point mutation of Thr to Ser resulted in a slightly higher binding affinity to the receptor mostly likely because of increased H-bonding interactions with receptor.

The difference between the sequence TNYLFSPNGPIG (mutation 2) and the other sequences included in the study was the hydropathicity score, which has been shown to influence protein aggregation [100]. This sequence, TNYLFSPNGPIG was determined to have a hydropathicity of − 0.23 compared to the sequence TNYLFSPNGPIA (− 0.04) and SNYLFSPNGPIA scoring at (− 0.05) (Table 1).

This indicates that TNYLFSPNGPIG is more hydrophilic than SNYLFSPNGPIA or TNYLFSPNGPIA [101] resulting in a weaker binding affinity due to its lesser ability to interact with a hydrophobic pocket. In addition, the point mutation of Ala to Gly could have influenced the interaction of the sequence TNYLFSPNGPIG with the EphB4 receptor due to replacement of methyl group of Ala with hydrogen of Gly. The sequence TNYLFTPNGPIA (mutation 3) resulted in a binding affinity of -9.1 kcal/mol with the EphB4 apo receptor and -8.1 kcal/mol with the ATP bound receptor. This sequence was formed by mutating TNYLFSPNGPIA at position 6, replacing Ser with Thr. At position 6, Thr is next to Phe (at position 5) and Pro (at position 7). This change overall increases the relative hydrophobicity and rigidity of the sequence due to the methyl group of Thr. Thus, the mutated sequence TNYLFTPNGPIA (mutation 3) may result in increasing binding interactions of the receptor-ligand complex due to greater interactions with the hydrophobic pocket of the receptor. However, it was seen that upon conjugating with the polyphenols, the binding affinity decreased. This is likely due to changes in the local environment of the peptide upon conjugation with the polyphenols that may result in steric hindrance and additional intramolecular C-H–-π interactions between the peptide and polyphenol components that may hinder binding interactions with the receptor [102]. As shown in Fig. 4, the polyphenol-peptide conjugates were also found to bind to the active site of the receptor. Both EphB4 and EphB2 structures adopt the bi-lobed kinase fold. The main differences occur in the hinge region and the activation loop upon binding to the conjugates. Upon binding to the conjugates, all of the apo isoform of the EphB4 receptors and two of the ATP bound EphB4 receptors (coumarate- TNYLFSPNGPIA, and gallate- SNYLFSPNGPIA) adopt a closed conformation, with the Gly-rich loop folding firmly. The other two ATP bound receptor complexes with sinapate-TNYLFSPNGPIA and sinapate-SNYLFSPNGPIA are relatively more open, with the Gly-rich loops partially disordered. As for the EphB2 binding interactions, the conjugates appear to bind to the area encompassing the juxtamembrane and kinase domain interface. There was a slight increase in binding affinity for the EphB2 receptors for the polyphenol conjugates for the apo receptor, though it was lower than that observed for EphB4 receptors in most cases. We also observed that the values observed were marginally lower for the ATP bound receptor with the highest binding affinity seen for Sinapate-SNYLFSPNGPIA (-7.2 kcal/mol) for EphB2 according to AutoDock results and SNYLFSPNGPIA (-10.2 kcal/mol) according to DockThor results.

As controls, the neat polyphenols were also analyzed for their ability to bind with the EphB4 receptor. The resulting binding affinities were found to be significantly lower compared to the conjugates and the neat peptide sequences, indicative that the peptide sequence enhances interactions with the receptor. This is likely due to the fact that with the neat polyphenols only limited interactions occur with the receptor. Overall, the results of the docking studies with the polyphenols indicated that the incorporation of a polyphenol conjugated to the peptide sequences increased EphB4 receptor binding affinity in most cases, except mutation 3. Given that SNYLFSPNGPIA and TNYLFSPNGPIA conjugates showed higher binding affinities, we further examined the residues involved in binding. We also compared interactions with the neat peptides SNYLFSPNGPIA and TNYLFSPNGPIA using PLIP (protein ligand interaction profiler). Results obtained are shown in Supplementary Information Tables S3 and S4 for the Apo and the ATP bound receptors, respectively. Almost all interactions were primarily H-bonds or hydrophobic interactions with a few π-stacking interactions and salt bridges observed. One of the interesting findings is that while the Apo EphB4 showed a larger number of interactions with most of the conjugates and the peptides, the ATP bound EphB2 showed significantly higher interactions with the conjugates and the peptides. This implies that in the presence of the ATP, the number of interactions is further strengthened with the EphB2 receptor kinase domain. TNYLFSPNGPIA formed 16 hydrogen bonds with apo EphB4 compared to only 10 hydrogen bonds with apo EphB2. For the ATP bound EphB4 receptor, TNYLFSPNGPIA formed 12 H-bonds and 20 H-bonds with the ATP bound EphB2 receptor. Hydrogen bonds thus played a key role in strengthening the interactions. In a previous study involving Eph kinase inhibitors, researchers concluded that the stabilization from hydrogen bond formation was key to inhibit the function of the EphB4 receptor kinase [103]. The TNYLFSPNGPIA peptide also formed 9 hydrophobic interactions with apo EphB4, compared to 5 with apo EphB2, and 8 hydrophobic interactions with ATP bound EphB4. Thirteen hydrophobic interactions were seen with ATP bound EphB2 indicating influence of higher hydrophobic interactions and H-bonding interactions.

The conjugation of polyphenols to TNYLFSPNGPIA influenced the interactions with both forms of the EphB4 receptor. With the apo EphB4, the gallate-TNYLFSPNGPIA conjugate formed 19 H-bonds. H-bonds were seen with D702, D740, and N826, which were not shared with the other conjugates. It also was the only conjugate to show hydrophobic interaction with F695 of the apo form of EphB4, and shared the hydrophobic interaction with residue R819 only with the sequence TNYLFSPNGPIA. For the ATP bound EphB4, the gallate-TNYLFSPNGPIA formed 11 H bonds and 9 hydrophobic interactions. While new H-bonds were observed compared to the apo receptor, the H-bonds to D702 and A700 were conserved. The decrease in the overall number of interactions of the gallate-TNYLFSPNGPIA with the ATP bound receptor compared to the apo receptor is likely due to the competitive behavior of the conjugate in the presence of ATP.

For the apo EphB4, the sinapate-TNYLFSPNGPIA conjugate formed 20 H-bonds including H-bonds with E664, M696, and R819. A total of 18 hydrophobic interactions were observed, including those with V629 and L707. In a previous study, hydrogen bond interactions of EphB4 inhibitors with the M696 residue was found to be of importance when targeting the kinase domain within the active site of the receptor [35]. Thus, the M696 H-bond interaction could indicate stronger inhibition of the sinapate-TNYLFSPNGPIA conjugate. In comparison, the ATP bound EphB4 receptor showed the formation of 11 H-bonds and 9 hydrophobic interactions once again indicating that in the presence of ATP the number of interactions are lowered for the EphB4. The p-coumarate-TNYLFSPNGPIA formed 18 H-bonds and 14 hydrophobic interactions with and was the only conjugate to form hydrogen bonds with residues S757 and G817 as well as hydrophobic interactions with M696 and K781 with the apo EphB4 receptor. For the ATP bound EphB4, p-coumarate-TNYLFSPNGPIA formed 12 H-bonds and 7 hydrophobic interactions. The H-bonds with D702 and N744 seen for the peptide by itself were conserved.

The polyphenol-TNYLFSPNGPIA conjugates were also analyzed with the apo EphB2 and ATP bound EphB2 receptor. For the apo EphB2 receptor, all of the conjugates and the sequence TNYLFSPNGPIA showed hydrophobic interactions with L753 which is a residue within the activation loop and hydrogen bonds with S706 and S709. However, in the case of the ATP bound EphB2 receptor, the L753 interaction was seen only for TNYLFSPNGPIA and its conjugates with sinapate and coumarate. Likewise, the S709 interaction was only seen for the gallate-TNYLFSPNGPIA, while the S706 H-bond interactions were shared by all of the conjugates indicating the importance of the S706 H-bond in all of the interactions.

We then analyzed the interactions of polyphenol-SNYLFSPNGPIA (mutation 1) conjugates as well as that of SNYLFSPNGPIA peptide with the Apo and ATP bound EphB4 receptor as those conjugates showed highest binding affinity. Residues involved in H-bonds commonly seen in gallate, p-coumarate, and sinapate-SNYLFSPNGPIA conjugates as well as the SNYLFSPNGPIA sequence include A700, S703, R706, and R744 for the Apo EphB4. Of particular importance is the R744 residue present in the catalytic loop. These interactions were also found between the sequence TNYLFSPNGPIA and Apo EphB4. Notably, R744 forms several hydrogens bonds with each of the constructs. Hydrogen bonds that were not found with the TNYLFSPNGPIA (original) sequence but found in the gallate-SNYLFSPNGPIA, p-coumarate-SNYLFSPNGPIA, and sinapate-SNYLFSPNGPIA conjugates include residues T693, D702, and R819, indicating a tighter binding. The common hydrophobic interactions across these models are with I621 and W786, which were also shared with hydrophobic interactions seen for the TNYLFSPNGPIA peptide and SNYLFSPNGPIA peptide. A salt bridge forms only when the sinapate-SNYLFSPNGPIA docks with apo EphB4, with residue R706. The formation of salt bridges with the EphB4 receptor was shown to stabilize interactions with its natural ligand ephrin-B2 [37]. Thus salt bridge formation may further enhance stabilization with the polyphenol-peptide conjugate. Additional interactions formed between sinapate-SNYLFSPNGPIA and the apo EphB4 receptor that are not common with the other constructs include a hydrogen bond with R631, and hydrophobic interactions with residues M696 and A700. This construct also formed a H-bond with N745 (also part of the catalytic activation loop) and a hydrophobic interaction with both A645, also shared with the sequence TNYLFSPNGPIA and the sequence SNYLFSPNGPIA. Additionally, it formed hydrophobic interactions with V629 and F695 that were only shared with the sequence SNYLFSPNGPIA.

For the ATP bound EphB4 receptor, as before there was a decrease in number of interactions in the active site due to competition with ATP. The crucial H-bond with M696 was seen for the Gallate- SNYLFSPNGPIA and p-coumarate-SNYLFSPNGPIA. The R 744 H-bond interaction was observed in all conjugates except the p-coumarate-SNYLFSPNGPIA conjugate while N745 H-bond was seen for Gallate-SNYLFSPNGPIA; SNYLFSPNGPIA and p-coumarate-SNYLFSPNGPIA. The F-695 hydrophobic interaction was only seen for sinapate-SNYLFSPNGPIA and p-coumarate- SNYLFSPNGPIA. On average the conjugates and the SNYLFSPNGPIA peptide displayed between seven to fifteen H-bonds with the ATP bound EphB4 with the highest number of H-bonds (15) seen for the gallate-SNYLFSPNGPIA. This corroborates with the docking study which also showed higher binding affinity with gallate-SNYLFSPNGPIA. The number of hydrophobic interactions was found to be between five and nine, with the highest number of hydrophobic interactions seen for sinapate-SNYLFSPNGPIA.

Each of these peptides and polyphenol-conjugated peptides showed different interactions with the EphB2 receptor. These variances are explained by the difference in binding site residues for the EphB2 receptor compared to EphB4. For the Apo EphB2, new pi-stacking interactions were seen with the sinapate-SNYLFSPNGPIA conjugate with the residue F632. Pi-stacking interactions have previously been shown to increase and stabilize binding of enzymes with drugs [104]. The shared interactions across all three polyphenol peptide conjugates and the sequences TNYLFSPNGPIA and SNYLFSPNGPIA were H-bonds with residues S706, R750, and S763. In addition, all of the polyphenol-peptide conjugates formed a hydrophobic interaction with the residue E670. The gallate-SNYLFSPNGPIA conjugate formed H-bonds to residues C636 and E703 which were not shared by any of the other constructs indicating interactions with the juxtamembrane–kinase domain interface. Among the unique interactions for the sinapate-SNYLFSPNGPIA conjugate was a H-bond to R825 in addition to hydrophobic interactions with T699 and F701. This conjugate also formed the only pi-stacking interaction with residue F632 and a salt bridge with R712. The p-coumarate-SNYLFSPNGPIA formed distinct interactions with apo EphB2, consisting of a hydrogen bond with H744, and hydrophobic interactions with D666 and I683. This conjugate also formed H-bonds with F632 and M702, as well as hydrophobic interactions with I627. Additionally, it formed a H-bond with N751 which was shared only with the neat SNYLFSPNGPIA sequence.

Interestingly, gallate-SNYLFSPNGPIA shared 15 H-bonds with EphB4 and EphB2 (ATP bound), compared to 12 H-bonds seen for the SNYLFSPNGPIA for ATP bound EphB2 and 20 H-bonds seen with TNYLFSPNGPIA for ATP bound EphB2. Thus the original peptide appears to have strong H-bond interactions with EphB2 receptor despite the presence of ATP. Interestingly, however, the number of hydrophobic interactions were found to be seven for gallate-SNYLFSPNGPIA and eight for SNYLFSPNGPIA peptide by itself with ATP bound EphB2. The sinapate-SNYLFSPNGPIA conjugate, however, showed significantly higher hydrophobic interactions with ATP bound EphB2 which can be attributed to the hydrophobic sinapate moiety. The coumarate-SNYLFSPNGPIA also showed a large number of hydrophobic interactions (fourteen). These results indicate that both hydrophobic and H-bond interactions played a major role in binding with the EphB2 ATP bound receptor. Interactions are seen in the juxtamembrane-kinase domain interface with hydrophobic residues such as V626, I637 and F632 and V635 [77] as well as the kinase domain residues such as D764, F 765, W 792 and A 797.

Molecular dynamics

We first compared the MD simulations of peptides TNYLFSPNGPIA and SNYLFSPNGPIA, upon complexation with the Apo and the ATP bound receptors. The results are shown in Fig. 5. The sequence TNYLFSPNGPIA complex with apo EphB4 showed a stable root mean square deviation (RMSD) for both the receptor and ligand, while the RMSD for the Cα chain stabilizes in the range of 0.42 A°, the complex remains stable between 1.5 A° and 1.6 A° for the EphB4 apo receptor (Fig. 5a). The ATP bound receptor with the peptide, however, showed higher RMSD values and numerous deviations particularly for the first 50 ns of the simulation, after which it was found to stabilize (Fig. 5b). The RMSD value after 65 ns remained in range of 12 A° to 12.5 A° for Cα, while that of the TNYLFSPNGPIA-complex leveled off at 17.5 A°. These results indicate that in the presence of ATP, the TNYLFSPNGPIA peptide appears to meander within the binding pocket for the first half of the simulation, after which it appears to stabilize. For the last twenty-five nanoseconds, the RMSD value remains steady. However, the relatively high RMSD value is indicative of a likely change in conformation of the receptor and the movement of the peptide within the binding pocket. We then compared the interactions of SNYLFSPNGPIA peptide. As seen in Fig. 5c, the apo form once again shows a much lower RMSD with fairly low deviations throughout the simulation. However, for the ATP bound EphB4 receptor (Fig. 5d) significant deviations are observed during the first 30 ns and then stable binding is observed. The RMSD value for the SNYLFSPNGPIA bound receptor was found to be lower (11 A°) than that of Cα which stabilized around 15 A°, indicative that the complex formed with SNYLFSPNGPIA was overall showing less deviation.

Fig. 5.

RMSD plots comparing EphB4 and EphB2 receptor binding with TNYLFSPNGPIA and SNYLFSPNGPIA. a TNYLFSPNGPIA complex with apo EphB4; b TNYLFSPNGPIA complex with ATP bound EphB4; c SNYLFSPNGPIA complex with apo EphB4 d SNYLFSPNGPIA complex with ATP bound EphB4. e TNYLFSPNGPIA complex with apo EphB2; f TNYLFSPNGPIA complex with ATP bound EphB2; g SNYLFSPNGPIA complex with apo EphB2 h SNYLFSPNGPIA complex with ATP bound EphB2

We then compared the stability of the complexes with the EphB2 receptor. As seen in Fig. 5e, TNYLFSPNGPIA forms a relatively stable complex with EphB2 apo form. Minor deviations are observed in the first 40 ns, after which it remains fairly stable in the range of 1.3 to 1.4 A°. The difference between the Cα and the receptor bound TNYLFSPNGPIA is 1A°, indicating high stability. Figure 5f shows that for the ATP bound EphB2 receptor, the RMSDs appear to increase for the first 70 ns, after which it levels off. The difference between the complex and Cα is within 4A°. Overall it appears that the deviations were comparatively lower for the EphB2 ATP bound receptor compared to the EphB4 ATP bound receptor complex with the peptide TNYLFSPNGPIA. The Apo receptor complex with SNYLFSPNGPIA (Fig. 5g) once again shows the formation of a stable complex except for a jump between 40 and 50 ns, after which the complex remains stable at 1.5A°. Figure 5h shows the receptor-SNYLFSPNGPIA complex with deviations throughout the 100 ns run, indicating that the peptide constantly fluctuates within the binding pocket of the ATP bound EphB2 receptor and is not very stable. To further assess these deviations, we examined the trajectories at three different time points (0 ns, 50 ns and 100 ns) for the ATP bound complexes with the two peptides (Supplementary Information Figure S1). The TNYLFSPNGPIA molecule makes critical contacts in the hinge region including Met 696. However, over the course of the simulation, a conformation change is observed for the receptor, particularly at the N-terminal lobe, and the C-helix region. Moreover, at 50 ns, the ATP molecule moves away; however, by 100 ns it is seen making contacts with the catalytic loop region. For the SNYLFSPNGPIA bound EphB4 complex, once again changes appear in the N-lobe region where the beta-sheets appear to undergo a conformation change. Furthermore, while initially the ATP appears to be mostly away from the receptor, making almost no contact with the receptor, the peptide makes contact with critical residues such as Asn 745 and Thr 693. Over the course of the simulation however, the ATP and the peptide are partially attached to the catalytic loop, though changes in conformation of the peptide occur, where it completely folds up. For the EphB2 receptor complex with TNYLFSPNGPIA, the peptide initially spans a large part of the C-lobe region while the C terminal region is attached partially to the catalytic loop region. However, over the course of the simulation, it moves entirely into the C-lobe region, while the ATP remains attached to the catalytic loop. Furthermore, a change in conformation in the C-helix region is observed where it opens up toward the end of the simulation. For the SNYLFSPNGPIA complexed with EphB2 receptor, the ligand spreads across the C-lobe and the catalytic loop with the ATP remaining at the catalytic loop. However the ligand folds up at 50 ns, and once again begins to spread out at 100 ns. Additionally, conformation changes are also seen for the receptor in the C-lobe region. These changes account for an increase in deviations and consequently relatively higher RMSD values observed for the ATP bound complexes particularly for the TNYLFSPNGPIA bound to EphB4 and EphB2 receptors.

For the polyphenol-peptide complexes, results are shown in Fig. 6 for the EphB4 receptors. As seen in the Fig. 6a–f, both the TNYLFSPNGPIA and SNYLFSPNGPIA conjugates with gallate, coumarate and sinapate were found to be highly stable with very little deviations throughout the simulation with the Apo form of the EphB4 receptor, with all three conjugates showing RMSDs in the range of 1.3A° to 1.6A for the receptor bound conjugates and slightly higher values seen for the SNYLFSPNGPIA conjugates. The coumarate-TNYLFSPNGPIA conjugate complex showed relatively more deviations in the 40 to 60 ns however, but after that stabilized at 1.2 A°. Likewise, the sinapate-SNYLFSPNGPIA conjugate complex showed deviations in the range of 15 to 25 ns after which it stabilized. Overall all, complexes were found to be highly stable for the apo EphB4 complex, with the highest stability attained by gallate-TNYLFSPNGPIA conjugate.

Fig. 6.

Comparison of RMSD plots of polyphenol peptide conjugates complexes with EphB4 receptors. a through f indicate ApoEphB4 receptors with a gallate-TNYLFSPNGPIA; b coumarate-TNYLFSPNGPIA; c sinapate-TNYLFSPNGPIA; dgallate- SNYLFSPNGPIA; e coumarate-S NYLFSPNGPIA; f sinapate- SNYLFSPNGPIA. g through J indicate ATP bound receptors. g complexes with polyphenols conjugated with TNYLFSPNGPIA; h corresponding Cα; i complexes with polyphenols conjugated with SNYLFSPNGPIA; J corresponding Cα. Mut1 = SNYLFSPNGPIA; Pep = TNYLFSPNGPIA

We next examined the ATP bound EphB4 receptor complexes with these peptide-polyphenol conjugates (Fig. 6g–j). As seen in Fig. 6g, the RMSD values ranged from 3 to 11 A° on average, with the most stable complex observed for the coumarate-TNYLFSPNGPIA complex. Its RMSD value increased from 3A° to 5.8 A°, and it remained stable at that value for the majority of the simulation. The sinapate- TNYLFSPNGPIA conjugate was found to be stable up to 60 ns after which it showed deviations and finally stabilized at a RMSD value of 9A° by the end of the simulation. The gallate-TNYLFSPNGPIA conjugate showed stability throughout most of the simulation, except an increase in deviation observed between 70 to 80 ns before a stable conformation was again obtained for the rest of the simulation. The Cα RMSDS overall, were found to be stable in the range of 2.5 to 4A° with the Cα for the gallate complex showing slightly more deviations. In general, however, the polyphenol-TNYLFSPNGPIA complexes were all found to form relatively more stable complexes compared to the peptide TNYLFSPNGPIA alone which indicates that the conjugates are better suited to be binding to the binding pocket. The SNYLFSPNGPIA-polyphenol conjugate complexes with ATP bound EphB4 (Fig. 6i) showed that the gallate-SNYLFSPNGPIA complex formed the most stable complex, though deviations were observed in the 40 to 60 ns range. Overall, the complex remained stable with a RMSD value of 6.1 A° after 70 ns. The sinapate-SNYLFSPNGPIA complex showed major deviations (upto 12 A°) for the first 20 ns, after which it was found to be very stable for the rest of the simulation. This indicates that initially the sinapate-SNYLFSPNGPIA was moving around within the binding pocket, but later was tightly bound to the binding pocket. Interestingly the gallate-SNYLFSPNGPIA showed continuous increase throughout the simulation, which is indicative that it didn’t form a very stable complex. The corresponding Cα also showed a similar trend, while the sinapate-SNYLFSPNGPIA and the coumarate-SNYLFSPNGPIA Cα RMSDs were remained between 2 to 2.5A°.

For the Apo EphB2 receptor Fig. 7a–f, the most stable complexes were formed for the gallate- SNYLFSPNGPIA and sinapate-SNYLFSPNGPIA with the least amount of deviations, while gallate- TNYLFSPNGPIA was found to display the most amount of deviation throughout the simulation. On average, the difference between the Cα RMSD and the complexes were found to be around 1A°, which is considered stable for all complexes. We next examined the ATP bound EphB2 receptors with the various polyphenol peptide conjugates. As can be seen in Fig. 7g–l, the most stable complex was formed with the gallate-TNYLFSPNGPIA conjugate, while the least stable one was with the sinapate-TYLFSPNGPIA conjugate. In this case least stable case, continuous deviations were observed for the first 60 ns after which the RMSD value continued to increase for the rest of the simulation, indicating that the conjugate was not stable within the binding pocket. In the case of coumarate-TNYLFSPNGPIA and sinapate-TNYLFSPNGPIA complexes (Fig. 7h and i), particularly for sinapate-TNYLFSPNGPIA, the RMSD appears to continue to increase even toward the end of the simulation, showing a significant difference in the RMSD value. Thus, these two complexes appear not to attain stability during the course of the simulation with the ATP bound EphB2 receptor, in comparison to the gallate-TNYLFSPNGPIA conjugate. To assess these results, we examined the trajectories of these complexes (data not shown). It was seen that in the case of coumarate-TNYLFSPNGPIA conjugate, initially the peptide spreads across the hinge region and the catalytic loop region where the ATP is also bound region. However, by the end of the simulation, conformation changes are observed in the N-lobe region and the activation loop region, accounting for the increase in RMSD value. For the sinapate-TNYLFSPNGPIA conjugate complex, the conjugate initially binds across the C-lobe and the activation loop region, however, by the end of the simulation the conjugate appears to completely move away from the receptor, while the receptor itself appears to open up and show major conformation change in C-lobe region. These results further account for the instability of these two complexes with ATP-bound EphB2 receptor.

Fig. 7.

Comparison of RMSD plots of polyphenol peptide conjugates complexes with EphB2 receptors. a through f indicate Apo EphB2 receptors with a gallate-TNYLFSPNGPIA; b coumarate-TNYLFSPNGPIA; c sinapate-TNYLFSPNGPIA; d gallate- SNYLFSPNGPIA; e coumarate-S NYLFSPNGPIA; f sinapate- SNYLFSPNGPIA. g through l indicate ATP bound receptors. g complexes with gallate-TNYLFSPNGPIA; h coumarate-TNYLFSPNGPIA; i sinapate-TNYLFSPNGPIA; J gallate- SNYLFSPNGPIA; k coumarate-S NYLFSPNGPIA; l sinapate- SNYLFSPNGPIA

Interestingly, however, the corresponding sinapte-SNYLFSPNGPIA conjugate was found to attain stability within the binding pocket after the first 30 ns with a RMSD value of 6 A°. The coumarate-SNYLFSPNGPIA was found to show minor deviations throughout the simulation. Thus, overall the RMSD results indicate that specific conjugates bind to the receptors and especially that the apo receptors have more stable conformation. Still most of the ATP bound receptors stabilize after an initial deviation, likely due to competition with ATP in the binding pocket.

To further elucidate these results, the receptor RMSFs (root mean square fluctuations) were analyzed as shown in Supplementary Information Fig. S2. As shown in the figures, for the ATP bound EphB2 receptor complex with SNYLFSPNGPIA and its conjugates, the highest fluctuations were observed for residues Phe 632, Arg 664, Ser 706, Tyr 827 and Gln 799. The unconjugated peptide showed strong fluctuations with Ser 706. The higher fluctuations in these regions are indicative that interactions are occurring primarily in the catalytic domain within the activation loop region, with some in the C-lobe residues. For the TNYLFSPNGPIA-polyphenol conjugates, highest fluctuations were observed for the sinapate-TNYLFSPNGPIA with residues including Gln 663, Val 695, Thr 719 and Ile 798. These residues are also within the catalytic domain; however, they are distinctly different from that observed for the SNYLFSPNGPIA conjugates implicating different modes of binding with the binding pocket. For the ATP bound EphB4 receptor, the highest fluctuations were seen upon binding to SNYLFSPNGPIA peptide, specifically with residues Arg 656, Gly 710, His 738, Lys 781, Phe 794 and Asn 685. The conjugates showed fluctuations at the same residues. Notably SNYLFSPNGPIA conjugate showed high fluctuations with residue Gly 710 and Gly 818 possibly due to higher H-bonding interactions with those residues. The TNYLFSPNGPIA alone once again showed greater amount of fluctuation. The main residues involved were Gly 638, Ser 686, Lys 781, Gly 710 and Gln 827.

The protein–ligand (P-L) interaction profiles further corroborated these findings. Figure 8 illustrates the protein–ligand contact profiles obtained, and a comparison of interactions between ATP bound receptors and apo receptors with TNYLFSPNGPIA and some of the conjugates. Although in almost all of the cases, higher interaction fractions were obtained for the apo receptors, the total number of amino acid residues involved was higher in the case of the ATP bound receptors. Figure 8a and b show a comparison of the interactions for apo and ATP-bound EphB4 receptors respectively, upon complexation with TNYLFSPNGPIA. H-bond interactions were predominant for both Apo and ATP-bound isoforms including Arg 744, Ile 621, Asp 702, Ala 700 and Asn 745. Interestingly, for the apo receptor, Asn displayed primarily H-bond interactions with little water bridges. However, for the ATP bound EphB4 receptor, this residue formed significantly higher water bridges compared to hydrogen bonds. Met 696 formed a strong H-bond with TNYLFSPNGPIA with the ATP-bound receptor, however, this interaction was significantly diminished in the case of the apo receptor. Slightly more hydrophobic interactions were seen in the case of the ATP bound EphB4 receptor compared to the apo counterpart, though Phe 695 interactions was involved in both cases. Similar trends were observed for the gallate-TNYLFSPNGPIA conjugate complex (Fig. 8c and d), although stronger interactions were observed with Ala 700, Asp 702, Asn 698, Glu 812 and Ser 703 for the ATP-bound EphB4 receptor. The interactions with Met 696 were reduced compared to the TNYLFSPNGPIA alone. The P–L interaction profiles for gallate-TNYLFSPNGPIA conjugate with EphB2 receptors (Fig. 8e and f) showed a similar trend demonstrating a higher number of residues of the ATP-bound receptors being involved in interactions compared to the apo isoform. Both the apo and the ATP bound forms, showed interactions with Asp 764, Met 702, Ser 706. The strong H-bond seen with Glu 624 for the ATP-bound receptor; was not seen for the apo receptor; however, Lys 653, which formed a strong H-bond in the case of the apo receptor, formed a relatively weak H-bond but showed both hydrophobic interaction and water bridge formation in the case of the ATP bound receptor. Figure 8g and h show the coumarate- TNYLFSPNGPIA conjugate with the apo and ATP bound forms of the EphB4 receptor. As can be seen, Ala 623 formed a strong H-bond as well as displayed hydrophobic interactions, with both receptors. Interactions with Ile 621, Asp 702, Glu 812 and Arg 744 were conserved in both cases, but, interactions with Thr 847, Gln 851, His 850 found with the ATP bound receptor are not seen in the case of the apo EphB4 with coumarin-TNYLFSPNGPIA conjugate.

Fig. 8.

Comparison of Protein Ligand contacts with TNYLFSPNGPIA and some of its conjugate complexes. a TNYLFSPNGPIA with apo form EphB4; b TNYLFSPNGPIA with ATP bound EphB4; c Gallate-TNYLFSPNGPIA with apo form of EphB4; d Gallate-TNYLFSPNGPIA with ATP bound EphB4; e Gallate-TNYLFSPNGPIA with apo form of EphB2 receptor; f Gallate-TNYLFSPNGPIA with ATP bound EphB2; g Coumarate-TNYLFSPNGPIA with apo form of EphB4 receptor; f Coumarate-TNYLFSPNGPIA with ATP bound EphB4

The p-coumarate-TNYLFSPNGPIA conjugate with apo EphB2 (data not shown) exhibited protein–ligand contacts with Asp 708, Arg 750, Glu 670, Met 702, and Lys 653 for both apo and ATP-bound receptors, while the sinapate-TNYLFSPNGPIA conjugate showed protein–ligand contacts dominated by water bridges and hydrophobic interactions for both apo and ATP bound forms. Of note, this structure showed the highest proportion of hydrophobic interactions compared to other conjugates. The residues with the largest interaction fractions were Glu 625, Ser 703, Arg 740, Asp 758, and Ala 700. All of the other PL contacts with TNYLFSPNGPIA and its conjugates (data not shown) showed similar results, indicating that higher numbers of residues were involved for the ATP bound receptors with some overlaps in the identity of the residues involved. Based on the residues involved, it is clear that even in the presence of ATP, the conjugates are binding to the catalytic binding domain, though more hydrophobic interactions are involved. The binding is likely to be competitive, as the overall number of interactions with the receptors are increased in the presence of ATP.

The PL contacts with mutation I (SNYLFSPNGPIA) and its polyphenol conjugates were also examined (Fig. 9). Similar to TNYLFSPNGPIA, the SNYLFSPNGPIA sequence also showed interactions with a significantly higher number of residues with the ATP bound form of the EphB4 receptor (Fig. 9c) versus the apo receptor (Fig. 9a) (44 vs 29) though the overall interaction fraction was reduced. Notably, the Ile 621 interaction was reduced in the case of the ATP bound EphB4 receptor (Fig. 9b). Common residues found in both cases included Arg 744, Ser 603, Asp 702, Met 696, Ala 700, Asn 745. The Glu 812 H-bond interaction seen prominently in the case of the apo form is reduced in the ATP bound form. A similar trend was observed for the SNYLFSPNGPIA peptide EphB2 receptors (Fig. 9b and d), with a higher number of residues being involved for the ATP bound receptor. Interactions with Asp 746 and Arg 745 were common to both forms of EphB2 receptors upon interacting with SNYLFSPNGPIA, but, the prominent H-bond with Asp 764 found in the apo form is significantly reduced for the ATP bound EphB2 receptor. For the gallate-SNYLFSPNGPIA conjugate (Fig. 9e and g), once again a higher number of residues was observed for the EphB4 ATP-bound receptor (fifty-five) versus thirty-nine for the apo receptor. Furthermore, Arg 744, Asn 745, Glu 812, Ala 700 were commonly found interacting for both the ATP bound and apo EphB4 receptors. For the EphB2 receptor with gallate-SNYLFSPNGPIA, both forms (Fig. 9f and h) showed about the same amount of interaction fraction, though more hydrophobic interactions were involved with Ala 651, Val 635, Leu 753, Phe 701, Ile 27 in both cases.

Fig. 9.

Comparison of Protein Ligand contacts with SNYLFSPNGPIA and some of its conjugate complexes. a SNYLFSPNGPIA with apo form EphB4; b SNYLFSPNGPIA with apo form EphB2; c SNYLFSPNGPIA with ATP bound form of EphB4 receptor. d SNYLFSPNGPIA with ATP bound form of EphB2 receptor e Gallate-SNYLFSPNGPIA with apo form of EphB4; f Gallate-SNYLFSPNGPIA with apo form of EphB2 receptor; g Gallate-SNYLFSPNGPIA with ATP bound form of EphB4 receptor h Gallate-SNYLFSPNGPIA with ATP bound EphB2 receptor; i Sinapate-SNYLFSPNGPIA with apo from of EphB4 receptor; j Sinapate-SNYLFSPNGPIA with apo form of the EphB2 receptor; k Sinapate-SNYLFSPNGPIA with ATP bound form of EphB4 receptor; l Sinapate-SNYLFSPNGPIA with ATP bound form of EphB2 receptor

Interestingly, Asp 764 also shows ionic interactions with SNYLFSPNGPIA in addition to H-bonds and water bridges in the ATP bound form, but no ionic interactions are seen in the apo form. These subtle changes are likely due to the positioning of the gallate-SNYLFSPNGPIA conjugate in the binding pocket of the receptor when ATP is already present. The interactions between sinapate-SNYLFSPNGPIA and EphB4 ATP bound receptor once again showed that a higher number of residues are involved. Although several of the residue interactions are maintained in both the Apo and the ATP bound EphB4 (Fig. 9i and k), it is worth noting that new contacts are observed with Gly 622, Phe 626, Ile 646 and Ser 758 between sinapate-SNYLFSPNGPIA and the ATP bound EphB4. The ATP bound EphB2 receptor (Fig. 9l) likewise also showed involvement of more residues from the binding pocket interacting with sinapate-SNYLFSPNGPIA, with prominent interactions being found with Arg 750, Asp 746, Leu 753, Arg 801. Interestingly the Asp 708 water bridge interaction seen for the apo form (Fig. 9j) is significantly less with the ATP bound form. The coumarate-SNYLFSPNGPIA (data not shown) showed similar trends, with higher interactions being seen for the ATP bound receptors (both EphB2 and EphB4). Some of the common interactions of coumarate-SNYLFSPNGPIA with both conformations of EphB4 included Ile 621, Ala 623 and Asn 745. The interactions with Arg 819 and Arg 744, which are prominent in the apo form, are significantly reduced in the ATP bound form. Common interactions with coumarate-SNYLFSPNGPIA for the ATP bound and apo EphB2 receptors included Asp 764, Met 702, Ser 706, Arg 750, Gly 633 and Ile 627. Interactions with residues such as Gln 624, Gln 626, Phe 701, Met 674 and Val 626 are not seen in the apo form, though they appear when coumarate-SNYLFSPNGPIA interacts with the ATP bound EphB2 receptor.

MMGBSA binding free energy of MD trajectories

The mechanics generalized Born model and solvent accessibility (MM/GBSA) calculation of the original peptide TNYLFSPNGPIA, the mutated peptide SNYLFSPNGPIA and their respective polyphenol conjugates was performed (Table 7). In addition to docking studies, the (ΔGbind) of protein–ligand complexes gives a good estimate of the binding affinities, on the basis of their binding free energies for the entire simulation [105]. To further elucidate the binding free energies during the simulation process, MMGBSA-binding energy estimation of trajectory snapshots were performed. The binding free energy was found to be highest (− 167.11 kcal/mol) for the ATP bound EphB2 receptor with coumarate-TNYLFSPNGPIA. The sinapate-SNYLFSPNGPIA displayed the highest ΔG bind amongst the apo receptors for apo EphB4. The apo EphB2 receptor showed relatively weaker binding interactions comparatively. Among the ATP bound EphB4 receptors, sinapate-TNYLFSPNGPIA showed highest ΔG bind at − 81.36 kcal/mol. Overall, the high ΔG bind values seen for the ATP bound EphB2 receptor may be attributed to the interactions occurring between the ATP moiety and the conjugates as well as the individual TNYLFSPNGPIA and the SNYLFSPNGPIA peptides. This is also confirmed by the high-coulomb energies observed (in the range between −109.3 to − 93.9 kcal/mol). For the ATP bound EphB4 receptors, the peptides and the polyphenol conjugates appear to move away from the ATP molecule in most cases either toward to the N lobe or C lobe during the course of the simulation. Therefore, the ΔG-binding energies observed is relatively lower. As an example, the results obtained by comparing the trajectory snapshots acquired at the end of the 100 ns runs for the SNYLFSPNGPIA-polyphenol conjugates with EphB2 and EphB4 (ATP bound) are seen in Fig. 10. The results show that the conjugates fold up and attach tightly to the binding pocket in very close proximity to ATP, favoring interactions in the case of EphB2 receptor. This is particularly evident for the sinpate-SNYLFSPNGPIA and the gallate-SNYLFSPNGPIA, which have higher ΔG bind. For the coumarate-SNYLFSPNGPIA although close to ATP, a major part of the conjugate appears outside of the binding pocket, despite being attached to the G loop region partially. It is also apparent from these images that the sinapate-SNYLFSPNGPIA and gallate-SNYLFSPNGPIA are competing for the ATP-binding pocket in the activation loop region and may potentially function as competitive inhibitors.

Table 7.

MMGBSA Average binding free energy based on trajectories

Ligand	EphB4 (apo) ΔGbind kcal/mol	With EphB2 (apo) ΔGbind kcal/mol)	EphB4 (ATP bound) ΔGbind (kcal/mol)	EphB2 (ATP bound) ΔGbind kcal/mol)
TNYLFSPNGPIA (original)	− 54.75	− 40.30	− 56.82	− 146.27
SNYLFSPNGPIA (mut1)	− 56.56	− 42.61	− 46.06	− 153.38
Gallate-SNYLFSPNGPIA	− 75.24	− 59.49	− 60.70	− 165.96
Gallate-TNYLFSPNGPIA	− 62.62	− 67.37	− 47.59	− 150.84
Sinapate-SNYLFSPNGPIA	− 73.98	− 94.1	− 69.27	− 163.478
Sinapate-TNYLFSPNGPIA	− 97.10	− 64.56	− 81.36	− 146.79
Coumarate-SNYLFSPNGPIA	− 55.59	− 59.02	− 53.54	− 143.99
Coumarate-TNYLFSPNGPIA	− 66.88	− 76.87	− 45.03	− 167.11

Fig. 10.

Comparison of trajectory snapshots obtained at the end of 100 ns simulations. Top row: ATP bound EphB4 complexed with a coumarate-SNYLFSPNGPIA; c sinapate-SNYLFSPNGPIA; e gallate-SNYLFSPNGPIA. Bottom row: ATP bound EphB2 receptors. b coumarate-SNYLFSPNGPIA; d sinapate-SNYLFSPNGPIA; f gallate-SNYLFSPNGPIA. Green ball and stick = conjugate; red = ATP at the bottom. Top cyan = ATP; pink indicates conjugate

The binding interactions may strengthen or weaken the contribution of the residues involved in binding. Furthermore, it may result in interactions with the ATP binding pocket residues that are either energetically favorable or unfavorable. In the case of corresponding ATP bound EphB4 receptors, the conjugates do not fold up. Instead they span throughout the entire catalytic domain (as seen in the case of the gallate-SNYLFSPNGPIA), though parts of it are in close proximity to ATP. The coumarate conjugate appears to occupy the N-lobe, while the sinapate conjugate appears to be only partially bound to the binding pocket. For the ATP bound EphB4 receptor bound to the TNYLFSPNGPIA conjugates, we observed lower ΔG bind compared to the apo isoform, and this may be attributed to the fact that the presence of ATP may weaken the binding interactions.

Pharmacokinetic prediction (SwissADME and ADMETlab 2.0)

Pharmacokinetics provides information on the absorption, distribution, metabolism, and excretion of molecules in relation to the physiological context of the human body and is essential to determine drug interactions. These factors help to determine the toxicity and metabolites of a particular drug as well as its ability to function in a specific physiological system [106, 107]. ADME properties of designed drugs have been often utilized to determine the potential of the compounds as oral drug candidates [108]. ADME studies were conducted for the peptide sequences as well as the polyphenols conjugated to TNYLFSPNGPIA and SNYLFSPNGPIA. The results are shown in Fig. 11 and the corresponding data is summarized in Tables 8 and 9. Each of the peptide sequences showed a bioavailability value of 0.17, formed a Pgp substrate, and were not CYP inhibitors. The greatest variation among the peptide sequences occurred with respect to their calculated iLOGP values. The original sequence (TNYLFSPNGPIA) had an iLOGP of 5.33 while for mutation 1 (SNYLFSPNGPIA) it was found to be 2.14. When the peptide sequence TNYLFSPNGPIA was conjugated with the polyphenols, all of the structures showed a bioavailability of 0.11, formed Pgp substrates, and exhibited low GI absorption. In addition, the iLOGP for each of the structures increased upon conjugation to the peptide sequence.

Fig. 11.

SwissADME properties of polyphenol-peptide conjugates of a Gallate b, sinapate- c, p-coumarate conjugates of TNYLFSPNGPIGA; d Gallate e, sinapate- f, p-coumarate- conjugates of SNYLFSPNGPIGA. The red shaded area indicates the ideal range for each of the six properties shown in Table 8

Table 8.

SwissADME Results

Name of Compound	iLOGP	GI Absorption	Pgp Substrate	CYP Inhibitor	Bioavailability Score
TNYLFSPNGPIA	5.33	Low	Yes	No	0.1
SNYLFSPNGPIA	2.14	Low	Yes	No	0.17
Gallate-TNYLFSPNGPIA	1.18	Low	Yes	No	0.11
Sinapate-TNYLFSPNGPIA	3.43	Low	Yes	No	0.11
p-Coumarate-TNYLFSPNGPIA	4.04	Low	Yes	No	0.11
Gallate-SNYLFSPNGPIA	2.87	Low	Yes	No	0.11
Sinapate- SNYLFSPNGPIA	4.24	Low	Yes	No	0.11
p-Coumarate- SNYLFSPNGPIA	1.44	Low	Yes	No	0.11

Table 9.

ADMElab 2.0 Results

Name of Compound	hERG Blocker	MDCK Permeability	MCE-18	Pgp-Inhibtor
Sinapate- SNYLFSPNGPIA	no	1.7E-05	166.9	0.0
Sinapate-TNYLFSPNGPIA	no	1.04E-05	163.924	0.0
p-Coumarate- SNYLFSPNGPIA	no	2.50E-05	157.921	0
p-Coumarate-TNYLFSPNGPIA	no	1.06E-05	157.921	0
Gallate-TNYLFSPNGPIA	no	1.5e-05	167.396	0
Gallate-SNYLFSPNGPIA	no	2.18E-05	164.404	0
SNYLFSPNGPIA	no	0.000109	139.652	0.001
TNYLFSPNGPIA	no	5.35E-05	139.652	0

The SwissADME data were also collected for the gallate, p-coumarate-, and sinapate- SNYLFSPNGPIA conjugates (Fig. 11d–f). Of note, the bioavailability score for all three compounds was also 0.11. This value is the same as the bioavailability of the polyphenol-TNYLFSPNGPIA conjugates. Oral bioavailability is a property measured through factors such as intestinal membrane permeability, hepatic first-pass metabolism, and gastrointestinal tract dissolution, and it has been shown that predicting bioavailability using simple molecular properties is highly variable [109, 110]. While the relatively low values may restrict oral administration, they may be administered by encapsulation in nanoscale drug delivery vehicles. In addition, it has been demonstrated that a bioavailability score of 0.10 or higher in rats was sufficient for a molecule to move forward in the process of drug development [111]. Thus, the polyphenol conjugates of both TNYLFSPNGPIA and SNYLFSPNGPIA may be developed as candidates for drug development.

Lipophilicity is measured through the n-octanol/water partition coefficient, LOGP. The implicit log P method uses implicit solvation models to calculate the solvation free energy through a combination of generalized Born and solvent accessible surface area methodologies and has been shown to exhibit strong predictive power when compared to similar approaches and control molecules [112, 113]. In a recent study, several potential drugs for the treatment of SARS-CoV-2 were analyzed with the use of the SwissADME database, reporting iLOGP as a parameter for determining if the molecules adhered to Lipinski’s rule of five [114]. The iLOGP for the p-coumarate- SNYLFSPNGPIA conjugate was found to be 1.44, which was the lowest of the three structures analyzed. The sinapate-SNYLFSPNGPIA showed an iLOGP of 4.24, making it the most lipophilic of the three structures. All of these structures report a LOGP of less than five, which adheres to Lipinski’s rule of five and suggests that they could be developed as potential drug candidates [115]. These results are consistent with the iLOGP values of the polyphenol-peptide conjugates with the sequence TNYLFSPNGPIA, which are all less than five and adhere to Lipinski’s rule of five. These findings suggest both sequences TNYLFSPNGPIA and SNYLFSPNGPIA (mutation 1) are viable drug candidates.

All of the structures exhibited low GI absorption and were not inhibitors of CYP. In general, CYP isozymes have an important role in the metabolism of drugs, and the evaluation of CYP inhibitors when designing drug candidates is essential for assessing potential drug-drug interactions that could have adverse effects in the body [116]. It has been shown that efflux proteins including Pgp and MRP2 on the apical membrane of the intestine can block the absorption of lipophilic compounds and limit their gastrointestinal and oral bioavailability [117]. All of these drug candidates formed Pgp substrates. Pgp has been found on cancerous cells within humans and is encoded by a multidrug resistance gene MDR1, indicating that forming a Pgp substrate that inhibits the normal functioning of Pgp can help to bypass drug resistance in cancerous cells [118]. The ability of these compounds to form Pgp substrate would indicate interactions with Pgp on cells with upregulation of the receptor, targeting those cancerous cells more directly, but it can also have negative impacts on the absorption of the compounds.

To further evaluate the pharmacological properties of the designed conjugates, we also utilized the ADMETlab 2.0 server to gain insights. The results are shown in Table 9. As seen in the table, all of the conjugates were found to be MDCK cell membrane permeable and none were found to be hERG blockers, which indicates that the molecules display no cardiotoxicity [119]. All of the conjugates were also found to be acceptable by the Pfizer rule and had relatively high MCE-18 values. Thus overall, these molecules may be considered for further laboratory analysis for developing potential modulators of the receptors studied here in.

SPR analysis

As a proof of concept, we explored the binding interactions with EphB2 and EphB4 receptors using SPR analysis for two of the synthesized conjugates. The sensograms obtained at a concentration range between 10 to 100 nM for the gallate-TNYLFSPNGPIA and sinapate-TNYLFSPNGPIA conjugates with the receptors are shown in Fig. 12. In previous work, SPR analysis has demonstrated that it is a vital tool that can aid in the measurement binding interactions of proteins, small molecules, lipids and drugs [120]. As can be seen in the figure, in the concentration range that was studied, the highest binding was seen for the sinapate-T-N-Y-L-F-S-P-N-G-P-I-A conjugates with the EphB4 receptor. In comparison gallate-T-N-Y-L-F-S-P-N-G-P-I-A conjugates showed lower binding as indicated by the response units. However, both gallate and sinapate conjugates did show an expected trend where the binding increased with concentration. In the case of EphB2 receptor, the binding was also found to be higher for the sinapate-T-N-Y-L-F-S-P-N-G-P-I-A conjugate, though overall it was lower compared to that obtained for EphB4 receptor. The gallate conjugate once again showed much slower and less binding. We then calculated the KD values for the sinapate-T-N-Y-L-F-S-P-N-G-P-I-A conjugates by inputting the data obtained into graphpad prism. Because the reflectivity was low for the gallate conjugates, no KD values were calculated. Interestingly, SPR analysis for the sinapate conjugates generated slow reduction in affinity (dissociation) at higher concentrations and KD was found to be essentially the same for both the EphB4 and EphB2 receptors at ~ 82.1 nM. We also compared the sensograms of the neat T-N-Y-L-F-S-P-N-G-P-I-A peptide with the receptors (Supplementary Information Fig. S3). As seen in the figure, for the EphB4 receptor, the peptide did not disassociate from the receptor at the higher concentrations, however, very little binding was observed at the 10 nM concentration. The T-N-Y-L-F-S-P-N-G-P-I-A peptide showed poor binding to EphB2 receptor. Comparatively, the sinapate- T-N-Y-L-F-S-P-N-G-P-I-A conjugate showed relatively higher binding with EphB2 receptor, indicating that conjugation with sinapate may enhance binding with EphB2 receptor.

Fig. 12.

Surface Plasmon Resonance (SPR) determination of binding affinity of sinapate-TNYLFSPNGPIGA and gallate- TNYLFSPNGPIGA with immobilized EphB2 and EphB4 receptor. a EphB2 receptor with sinapate-TNYLFSPNGPIGA; b EphB4 receptor with sinapate-TNYLFSPNGPIGA; c EphB2 receptor with gallate-TNYLFSPNGPIGA; d EphB4 receptor with gallate-TNYLFSPNGPIGA. Increasing concentrations of the conjugates (in nM) showed increases in binding to immobilized receptors

To further assess if kinase activity of the receptors was affected, we carried out ADP-Glo assay. The effect of the two conjugates, gallate-T-N-Y-L-F-S-P-N-G-P-I-A and sinapate-T-N-Y-L-F-S-P-N-G-P-I-A with EphB4 and EphB2 receptors were examined [121]. The activity was determined by quantification of the ADP generated by the kinase using the ADP-Glo assay. As can be seen in Fig. 13, for EphB4 receptor 42% inhibition was observed at 100 nM concentration of the sinpate-T-N-Y-L-F-S-P-N-G-P-I-A, while the peptide alone showed an inhibition of ~ 14% and the gallate-T-N-Y-L-F-S-P-N-G-P-I-A displayed an inhibition of 12.2%. The conjugates or the peptide, however, did not show kinase inhibitory activity with the EphB2 receptor. These results indicate that the sinapate-T-N-Y-L-F-S-P-N-G-P-I-A conjugate likely may act as a kinase inhibitor for the EphB4 receptor. While this was a proof of concept experimental study, future work will involve exploring all of the conjugates and examining affects at the cellular level.

Fig. 13.

Kinase activity of the peptide and the two conjugates by ADP-Glo kinase assay are shown at varying concentrations of the peptide and the conjugates. Percent inhibition is shown and was compared with respect to DMSO which was considered to have no inhibition. Error bars indicate standard error mean (n = 3)

Conclusions

In this work, we have designed new polyphenol-peptide conjugates for investigating interactions with EphB4 and EphB2 receptor kinase domains. Starting with the known antagonist peptide sequence, TNYLFSPNGPIA, which inhibits Ephrin B2 binding to EphB4 in the G-H loop region, we have designed three mutants of this peptide by creating point mutations at three separate positions and explored their interactions with the kinase domain of the EphB4 and EphB2 receptor through docking. To enhance targeting, optimal peptides SNYLFSPNGPIA and TNYLFSPNGPIA were conjugated to three polyphenols which are known for their ability to bind to kinases as well as for their antitumor properties. Our results indicated that the binding affinities were increased for most of the conjugates. Furthermore, we examined the stability of receptor-ligand complexes with these conjugates and neat peptides with both the Apo and the ATP bound EphB4 and EphB2 receptors. Molecular dynamics studies indicated that sinapate-SNYLFSPNGPIA, gallate-SNYLFSPNGPIA and coumarate-TNYLFSPNGPIA showed strong binding to the active site. The Apo EphB4 receptor showed relatively high binding and formation of stable complexes in most cases, while binding was slightly reduced in the presence of ATP. These findings suggest that targeting of the receptors can be improved with conjugation of the peptide sequences to these polyphenols. All of the peptides studied were considered anticancer peptides according to ACPP analysis. The findings from metaPocket and POCASA studies confirmed that the polyphenol-peptide conjugates were binding to the predicted active site of both the EphB4 and EphB2 kinase domains of the receptors. Overall, there were a larger number of protein–ligand contacts found with the ATP bound EphB2, indicating a strong association. Results also showed that ligand flexibility was conserved across the simulations. These structures were shown to have lipophilic properties and did not form CYP inhibitors. Proof of concept laboratory studies were carried out, which confirmed that the conjugates successfully can bind to the EphB4 receptor (specifically the sinapate conjugate). Minimal binding is observed for the EphB2 receptor. Furthermore, sinapate-T-N-Y-L-F-S-P-N-G-P-I-A showed higher kinase activity inhibition compared to the peptide alone and the gallate conjugate. Further laboratory studies with the other conjugates with the mutated peptide will need to be carried out to fully assess the inhibitory activities followed by in vitro cell studies. It is also to be noted that EphB4 receptors, like some of the other Eph receptors also show tumor inhibitory activity in addition to tumorigenesis. Therefore, the conjugates will need to be carefully evaluated when conducting in vitro and in vivo studies. The current results suggest that the sinpate conjugate may have potential for future lab studies and therapeutic assessment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conception, design of study: IAB; Acquisition of computational data: SMM, RMH, RED and DJL. MEM carried out synthesis, purification of conjugates, SPR analysis and the ADP-Glo assays. Analysis, writing and interpretation: IAB, SMM, RMH.

Data availability

Associated data are included in the supplementary information. The manuscript will be shared on faculty website.

Code availability

NA.

Declarations

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

NA.

Consent to participate

NA.

Consent for publication

All authors give full consent for publication.