Identification of potent and novel inhibitors against RAC1: a Rho family GTPase

Abstract

Head and Neck Squamous Cell Carcinoma (HNSCC) is one of the most common form of cancer worldwide. It has high incidence and mortality rate making it one of the top causes of cancer related deaths. Tremendous efforts have being made towards treatment of HNSCC but still the overall survival rate hasn’t improved much. Unregulated activation of Rho GTPase Ras-related C3 botulinum toxin substrate 1 or Rac1 has been reported in various tumor such as HNSCC, breast cancer, pancreatic cancer, etc. Rac1 is significant in activation and regulation of multiple signaling pathways and it’s aberrant activation leads to uncontrolled proliferation, invasion and metastasis which contributes to the hallmarks of cancer. Therefore for treating proliferative disorders such as cancer, inhibition of Rac1 could be a viable approach. Rho GTPases were earlier considered “undruggable” due to their picomolar binding affinity for their guanine nucleotides. In addition presence of high micromolar concentrations of GDP (> 30 μm) and GTP (> 300 μm) in the cell, led to unsuccessful attempts in identification of potent or selective nucleotide competitive GTPase inhibitors. Therefore we identified small molecule inhibitors that target the GEF binding site of the Rho GTPase instead of nucleotide binding site by performing high throughput screening, molecular dynamics simulations, free energy calculations and protein-ligand interaction studies. As a result of this study, we identified four potential inhibitors against RAC1. This study provides a significant in-depth understanding of the Rho GTPases and can prove beneficial in the development of potential therapeutics against HNSCC.

Introduction

The Rho-GTPase family is also known as “molecular switch” due to its ability to transform between the active GTP bound conformation and inactive GDP bound conformation. The activation and deactivation of the “molecular switch” is controlled by guanine nucleotide exchange factors (GEFs) and GTPase activating protein (GAPs), respectively (Bosco et al. 2009). Ras-related C3 botulinum toxin substrate 1 (Rac1) belongs to the Rho family GTPases that are activated by GEFs which consist of two families’ members: the classical Dbl-homology (DH) domain-containing protein family and the dedicator of cytokinesis proteins also known as Dock proteins. The GEF family comprises more than 80 members, with at least 20 GEFs implicated in directly activating Rac1. All Dbl-family members contain a conserved DH domain followed by a tandem pleckstrin-homology (PH) domain. The DH domain is responsible for establishing maximum interface with the GTPase, catalysing nucleotide exchange and dictating GTPase specificity. The PH domain on the other hand plays regulatory role but its exact functions remain unclear, a few example of Dbl family GEFs are -Tiam, TrioN, VAV2, etc. (Baumeister et al. 2006; Chhatriwala et al. 2007). The Dock proteins are atypical GEFs due to the absence of ‘Dbl’ domain. The presence of highly conserved Dock homology region (DHR1 or DHR2) domain mainly characterized the Dock proteins. The DHR1 is the membrane associated domain that binds to membrane phosphoinositide and the downstream DHR2 domain is the catalytic domain that is responsible for GDP/GTP nucleotide exchange (GEF) activity (Fort and Blangy 2017). Dock GEFs form an 11 member family, sub grouped into 4 categories- DOCK A(contains Dock1,Dock2 and Dock5), DOCK B(contains Dock3 and Dock4), DOCK C (contains Dock6,Dock7 and Dock8) and DOCK D(contains Dock9,Dock10 and Dock11) (Laurin and Côté 2014).

Rac1 functions in cytoskeleton modulation, normal cellular activities, growth signalling, cell cycle regulation pathways, cell-cell adhesions and contact inhibition (Bosco et al. 2009; Marei and Malliri 2017). Recent studies by Skvortsov et al. have reported the increased expression of RAC1 in tumor cells of HNSCC showing limited response to radiotherapy (Madhukar and Subbarao 2021; Skvortsov and Duda 2014). Mendoza-Catalan et al., 2012 have associated the nuclear expression of Rac1 with carcinogenesis and malignancies (Mendoza-Catalán et al. 2012). Data presented by GDC portal-NCI also shows the significance of RAC1 alteration in HNSCC (Fig. 1). Several studies associate increased expression of Rac1 with the limited response rate and enhanced tumors recurrence risks in HNSCC patients (Arnst et al. 2017; Saci et al. 2011; Skvortsov and Duda 2014). (Fig. 2) shows normal distribution of individual samples across the TCGA datasets of multiple RNA-seq analyses visualized with box plots for Rac1. Preclinical data supporting various combination therapies like radiation/cisplatin + Rac1 inhibitor are reported by several groups (Bosco et al. 2010; Skvortsov and Duda 2014). These studies illustrated that the Rac1 inhibitor blocks survival and progression in HNSCC tumor cells that exhibit enhanced Rac1 expression. Inhibition of RAC1 is found significant in breast cancer and ovarian cancer. RAC1 holds a lot of scope and its inhibition should therefore, be explored in details.

Fig. 1.

GDC data portal- NCI shows distribution of RAC1 alteration in different cancers, TCGA-HNSC indicates the TCGA data on head and neck cancer. BLCA – Bladder Urothelial Carcinoma; BRCA– Breast Invasive Carcinoma; CESC – Cervical Squamous Cell Carcinoma and Endocervical Adenocarcinoma; CHOL – Cholangiocarcinoma; COAD - Colon Adenocarcinoma; ESCA - Esophageal Carcinoma; GBM – Glioblastoma Multiforme; KIRC – Kidney Renal Clear Cell Carci- noma; LGG – Brain Lower Grade Glioma; LIHC – Liver Hepatocellular Carcinoma; LUSC – Lungs Squamous Cell Carcinoma; PRAD – Prostate Adenocarcinoma; READ – Rectum Adenocarcinoma; SARC – Sarcoma; SKCM – Skin Cutaneous Melanoma; STAD – Stomach Adenocarcinoma; THYM – Thymoma; UCS – Uterine Carcinosarcoma

Fig. 2.

The plot from the human protein atlas shows normal distribution of individual samples across the TCGA datasets of multiple RNA-seq analyses visualized with box plots for Rac1. Points are displayed as outliers if they are above or below 1.5 times the interquartile range

Rho GTPases were earlier considered “undruggable” due to their picomolar binding affinity for their guanine nucleotides. In addition presence of high micromolar concentrations of GDP (> 30 μm) and GTP (> 300 μm) in the cell, led to unsuccessful attempts in identification of potent or selective nucleotide competitive GTPase inhibitors (Gray et al. 2020; Maldonado and Dharmawardhane 2018). To tackle this problem we targeted the GEF binding site of the Rho GTPase instead of nucleotide binding site. This enables the inhibitors to block the activation of GTPase thereby inhibiting downstream signalling pathway.

In a similar attempt, the insilico study by Yuan Gao et al. identified small molecule NSC23766 that fitted into the surface groove of Rac1 critical for GEF interaction (Gao et al. 2004). On experimental evaluation NSC23766 was found to reduced growth and invasion in several cancer types, including prostate, breast, gastric, chronic myelogenous leukemia (Maldonado and Dharmawardhane 2018). However, the off target effects of NSC23766 in mouse platelets, as well as the high IC₅₀ (~ 50 µM) renders it ineffective for pharmacological use. Since then a few studies focusing on developing derivatives of NSC23766 such as EHop-016 were carried out and they reported relatively high effective concentrations, and the moderate bioavailability of EHop-016. This creates a huge need for identification of new and potent therapeutics against RAC1. Till date, there are no FDA approved specific inhibitors of activation of RAC1 protein therefore this study could help in development of such inhibitor.

Materials and methods

The crystal structure of human WT RAC1 bound to GDP was taken from PDB (PDB ID: 5N6O). The protein was prepared using the Protein Preparation Wizard by Schrödinger (Schrödinger Release 2018-3: Maestro, Build 12, Schrödinger, LLC, New York, NY, 2018) (Madhavi et al. 2013). During this process bond orders and formal charges were assigned, and hydrogens were added. Optimization of hydrogen bonding network including the reorientation of thiol and hydroxyl groups, sampling Asn, Gln and His side chains, and the prediction of the protonation states of His, Asp and Glu (Sastry et al. 2013), followed by brief energy minimization was performed. The generation of grid was performed to specify the active site of receptor for ligand docking jobs. Since Rho GTPases activate on binding with GEFs, the RAC1 residues (residue no. 53 to 72) involved in GEF binding were included in grid generation. Mutation studies of Asp38, Asn39, Gln61, Tyr64, or Arg66/Leu67 into Ala have shown that both Switch I (Asp38, Asn39) and Switch II (Gln61, Tyr64, Arg66/Leu67, Leu70/Ser71) have their implication in both GEF binding and catalysis (Gao et al. 2001). Therefore, the grid was generated by selecting the residues essential for GEF binding i.e., ASP38 ASN39 LEU53 GLY54 LEU55 TRP56 ASP57 THR58 ALA59 GLY60 GLN61 GLU62 ASP63 TYR64 ASP65 ARG66 LEU67 ARG68 PRO69 LEU70 SER71 TYR72, using Grid Generation of Maestro. The coordinates of grid center were: -17.63, 3.51,15.09, of the inner box were- 10, 10, 10 (Å) and of the outer box were- 30, 30, 30 (Å). The size of the grid box was set so as to dock ligands with size up to 20 Å. The robust and chemically correct structures of the compounds from these libraries were prepared using LigPrep of Maestro (Schrödinger Release 2016-2: Maestro, version 10.6, Schrödinger, LLC, New York, NY, 2016). Finally, the docking study was performed using Ligand Docking in Glide module of Schrodinger. The compounds were screened using Glide HTVS, SP and XP docking methodologies. The 10% of top-ranked compounds from HTVS output were taken as input for SP and 10% of top-ranked compounds from SP output were taken as input for XP (Friesner et al. 2006). To evaluate the docking results, the top 20 compounds from Glide XP output were further docked by using GOLD SUITE (GOLD Suite 5.2) which uses the genetic algorithm for docking. The top-ranked compounds after XP run were analyzed by X-Score (v1.2.1 software package) (Wang et al. 2002) and LigPlot (Wallace et al. 1995) for further evaluation of docking results. To further evaluate and determine the stability of the protein-ligand complexes formed by the top ligands selected after running Glide XP, MD simulation was performed using Gromacs (version 5.1.2). MD simulations were carried out for 100 nanoseconds (50,000,000 steps). For parameterisation of protein GROMOS96 43a2 force field was utilized. The topology file of ligand was generated by PRODRG server. The protein was placed in the centre of the cubic box defined during solvation with 10 Å distance from the centre. The SPC/E water molecules solvated the system and ions were added to neutralise the overall charge of the system. The number of SPC/E water molecule added for LAS 52,449,110 system were 17,236, for RAC1-7309-0137 system were 17,230, for RAC1-K292-1664 system were 17,241 and for RAC1-4333-0420 system were 17,234. In Gromacs the charges on a protein are set beforehand and held constant throughout the simulation, which means that it uses constant protonation state because all species in the system are treated in their predominant state at that particular pH (the pKa values of titratable groups are given and they are taken as representative of the system at equilibrium). For energy minimization which is a crucial step while preparing the system for MD simulation, following parameters were kept- integrator = steep, emtol = 1000.0, emstep = 0.01 and nsteps = 50,000. The steepest-descent energy-minimization algorithm was used for 50,000 steps, with periodic boundaries and a given tolerance on the change in energy. Long range electrostatic interactions were calculated using Particle mess Ewald (PME) keeping default parameters. The equilibration of the system was carried out in two phases. NVT (constant number of particles, volume and temperature) equilibration was performed for 100 picoseconds (ps) to stabilize the temperature of the system and NPT (constant number of particle, pressure and temperature) equilibration was performed for 100 picoseconds (ps) for stabilization of pressure (Abraham et al. 2015). The temperature was maintained at 300 K and the pressure was maintained at 1 bar using Berendsen barostat with pressure relaxation time of 1.0 ps. Constraints were imposed using default LINCS method. Finally, production run was performed for 100 ns at a time step of 2 fs and co-ordinates were saved at every 10 ps. The final estimation in the study was the prediction of binding free energy of the protein-ligand complex which was done by using g_mmpbsa tool designed for GROMACS interface (Baker et al. 2001; Kumari et al. 2014).

Results and discussion

Virtual screening

This study aims to find out potent and novel inhibitors of RAC1 using structure-based drug design approach. The screening ascertains that the top selected compounds i.e., LAS 52449110 from Asinex PPI library; compounds 7309-0137, K292-1664 and 4333-0420 from ChemDiv PPI library occupy the same binding pocket. The electrostatic potential was calculated for the surface visualization of the binding site and of the selected compounds using the Linearized Poisson–Boltzman equation at 300 K (Fig. 3). The complementarity of the binding pocket with ligands could be seen. The major portion of the binding pocket has either neutral or slight negative charge and the selected ligand were mostly neutral and sit in the binding pocket according to their surface potential complementarity.

Fig. 3.

Depiction of the electrostatic potentials. (A) Shows the electrostatic potential on the surface of the average conformation of the RAC1 GEF binding pocket, and (B) Shows electrostatic potential of the different ligands binding in the GEF binding pocket ( LAS 52449110 is shown in cyan; 7309-0137 is shown in yellow, K292-1664 is shown in blue and 4333-0420 in red)

In the 3D structure of RAC1 GEF-binding domain from PDB (PDB ID: 5n6o) the pocked formed by residues selected for grid generation was used as binding site for this study. To identify inhibitors whose conformations would fit into this binding pocket of RAC1, a 3D database search was performed. The databases we used were freely available ChemDiv (www.chemdiv.com/), Asinex (http://www.asinex.com) and Enamine (https://enamine.net). Different possible confirmations were generated for all the compounds using LigPrep (Maestro). The prepared small molecules were docked into the defined binding pocket of RAC1. For docking the compounds HTVS, SP and XP docking methodologies of Ligand Docking in Glide module of Schrodinger were used. The range of gscore for top 10 compounds from Enamine PPI library was -6.77 kcal/mol to -6.41 kcal/mol. The gscores from Asinex PPI library ranged from  -7.34 kcal/mol to -6.40 kcal/mol. The highest gscores were observed in case of ChemDiv PPI library. The gscores for ChemDiv PPI library ranged from -7.84 kcal/mol to -6.99 kcal/mol for top 10 compounds. Results of docking performed using Glide module from Schrodinger were evaluated using GOLD software and post docking analysis was performed using X-score and Ligplot software. X-score scored the binding of protein and highest scoring conformation of the selected top ligand for both Glide and GOLD output. The analysis results of selected compounds are summarized Table 1. Since the overall scores of compounds from Enamine PPI library were quite low as compared to overall scores of compounds from other libraries therefore, they were not included in further analyses of the study. The selected top compounds from Asinex PPI and ChemDiv PPI library were next analysed using Ligplot for their hydrophobic and hydrogen bond interactions with the protein (Table 2; Fig. 4).

Table 1.

Comparison of docking and post-docking results for top ranking potential inhibitors

Compounds	Glide score (kcal/mol)	Glide Xscore (kcal/mol)	Gold Fitness Score	Gold Xscore (kcal/mol)
Enamine
Z30720891	-6.77	-7.74	48.06	-7.96
Z235345683	-6.76	-8.02	52.20	-8.10
Z108564022	-6.72	-8.37	41.65	-8.75
Z127592722	-6.71	-7.65	47.88	-7.55
Z71395449	-6.72	-8.25	54.65	-8.69
Z1023111714	-7.06	-7.46	46.66	-7.50
Z1144606021	-7.00	-8.01	49.04	-7.82
Z108862406	-6.50	-7.72	45.57	-7.29
Z88310789	-6.46	-7.67	50.42	-7.59
Z1023973788	-6.41	-7.94	48.40	-8.17
Asinex
LAS 52449110	-7.34	-9.58	56.64	-9.04
BDC 26811615	-6.98	-8.29	49.61	-7.37
BDD 30136591	-6.93	-7.70	44.47	-7.29
LAS 53847930	-6.69	-7.78	51.72	-7.56
BDE 30137089	-6.64	-7.04	41.65	-6.53
BDC 30331036	-6.51	-7.26	46.10	-7.06
BDE 30161783	-6.90	-7.44	51.30	-7.17
BDC 30334111	-6.42	-7.25	44.00	-6.04
BDD 21879677	-6.40	-8.03	53.57	-8.03
ChemDiv PPI
7309 − 0137	-7.84	-8.22	52.17	-7.58
K292-1664	-7.58	-7.67	67.51	-7.91
F797-0107	-7.30	-7.75	49.64	-7.31
4333 − 0420	-7.26	-8.35	50.06	-8.50
D436-0438	-7.23	-8.80	57.27	-9.05
D321-0870	-7.20	-7.64	49.20	-7.21
D436-0451	-7.08	-8.92	57.62	-9.00
D436-0305	-7.06	-8.58	52.28	-8.76
D436-0137	-6.99	-8.17	48.53	-8.33

Table 2.

Ligplot results of selected compounds

Compounds Name	No. of Hydrogen Bond Interaction	No. of Hydrophobic interaction	Donor- Acceptor residues involved in HB formation
LAS 52449110	3	50	Donor Acceptor Distance Ligand ASN39         2.94 Ligand ASP38          2.84 Ligand ASP38          3.05
7309 − 0137	2	66	Donor Acceptor Distance PHE37    Ligand           3.09 Ligand THR35           2.74
K292-1664	1	44	Donor Acceptor Distance Ligand LEU67            2.93
4333 − 0420	2	38	Donor Acceptor Distance ALA59     Ligand           2.93 Ligand ASP57           2.76

Fig. 4.

Interaction profile of selected ligands with RAC1, generated using Ligplot; (A) LAS 52449110, (B) 7309-0137, (C) K292-1664 and (D) 4333-0420. Ligand in violet, hydrogen bonds between ligand and protein are represented by green colour dashed line, red radial spokes represents residues forming hydrophobic interactions and hydrogen bond forming residues are represented in golden colour bonds. Hydrogen bond lengths are labelled in Å

Finally, based on the scores from different analysis one compound from Asinex PPI library and three compounds from ChemDiv PPI library were selected for further validations (Fig. 5). LAS 52449110 was the compound from Asinex PPI library, it was found to bind the RAC1 with the gscore of -7.34 kcal/mol, which was higher than the other molecules from same library. Due to its bulky structure, the compound covers the binding pocket more liberally (Fig. 6), thus forming 3 hydrogen bond and 50 non-hydrogen bond interactions which was highest among all the four selected compounds. The next selected compounds were the three top compounds from ChemDiv PPI library. Compound 7309-0137, K292-1664 and 4333-0420 were the three of the four best molecules from ChemDiv PPI. They bind to RAC1 with the gscores -7.84 kcal/mol, -7.58 kcal/mol and  -7.26 kcal/mol, respectively. All the three molecules binds to RAC1 with HB and non-HB interactions. Compound 7309-0137 forms 2 hydrogen bond and 66 non-hydrogen bond interactions, compounds K292-1664 forms 1 hydrogen bond and 44 non-hydrogen bond interactions and compound 4333-0420 forms 2 hydrogen bond and 38 non-hydrogen bond interactions, within the binding pocket. These four selected compounds were finally analysed using MD simulation for 100 ns using GROMACS.

Fig. 5.

2D structures of selected compounds

Fig. 6.

Interaction of selected ligands with RAC1 (golden) important GEF binding site residues are highlighted in pink and are labelled, (A) LAS 52449110 is shown in cyan, (B) K292-1664 is shown in blue, (C) 4333-0420 in red and (D) 7309-0137 is shown in yellow,

On comparing the molecular docking of selected compounds with molecular docking results of EHop-016 and NSC23766 with Rac1 as reported by Montalvo-Ortiz BL et al. it was observed that similar to EHop-016 and NSC23766, the selected compounds bind to Rac1 into the cleft formed by same residues. Unlike NSC23766 which is stretched over the surface of Rac1, EHop-016 and the four compound identified in this study appears to favor a bent conformation that binds to a deeper binding pocket. Similar to EHop-016, in its energetically most favorable conformation, the binding of LAS 52449110 is strengthened by hydrogen bonding interactions with residues Asp-38 and Asn-39. Similarly, as EHop-016 the identified compound 4333-0420 also has a close interaction with Trp-56, which has been shown to be critical for binding of Rac to its GEFs. From the comparative analysis of molecular docking results generated in this study and molecular docking results of EHop-016 and NSC23766 reported by Montalvo-Ortiz BL et al. it is reasonable to postulate that LAS 52449110, 7309-0137, K292-1664 and 4333-0420 also interferes with binding of Rac1 with its GEFs via binding to the three-way intersection site involving the switch I, switch II, and β loops of the effector region of Rac that interacts with the DH domain of Rac GEFs (Montalvo-Ortiz et al. 2012) .

From the interaction analysis we can see that the protein and selected ligands form large number of hydrophobic interactions which are formed due to the narrow distance between non-polar amino acid side chains of the protein and lipophilic groups on the ligand. The hydrophobic interactions are not directional but for ligands with large lipophilic groups as present in the selected compounds, the hydrophobic interactions contribute significantly to the binding affinity .

Molecular dynamics

To determine the stability of the complex formed between selected ligands (LAS 52449110, 7309-0137, K292-1664 and 4333-0420) and RAC1 as well as to compare it with the apoprotein, a molecular dynamic simulation of 100 ns was performed, for simulation movies, see the supplementary information. To measure the average deviation of the backbone atoms of protein-ligand complex suggesting the stability of the complex, RMSD was calculated. The average RMSD values of LAS 52449110, 7309-0137, K292-1664, 4333-0420 and apoprotein were 0.33 nm, 0.28 nm, 0.34 nm, 0.26 and 0.30 nm, respectively. In case of 7309-0137 and 4333-0420, the protein-ligand was observed to be quite stable since the beginning of the simulation till the end. Towards the final 20ns of the run the deviation of apoprotein as well as complexes of protein with LAS 52449110 and K292-1664 get stabilized and the overall deviation was well within the acceptable limited (2Å). From the overall RMSD trend, it was observed that the complex of 7309-0137 and 4333-0420 with RAC1 was observed to be more stable than all the other complexes taken into account including the apoprotein. Also, throughout the run the average RMSD values of all the complexes lies within 0.2 nm, which suggests that there wasn’t a significant deviation of complexes from the reference structure (Fig. 7(A)). Similar results were reported by Chunwen Zheng et al. when they simulated RAC1-NSC23766 complex for 10ns on three different structure generated by Autodock4, Autodock Vina and HDOCK using PDB ID: 5N6O as template. They observed that there was no significant fluctuations and the average RMSD values of all the complexes lies within 0.15 nm (Zheng et al. 2021).

Fig. 7.

MD simulation data analysis plots for RAC1. (A) RMSD plot. (B) Backbone atomic fluctuations (RMSF) plot. (C) Plot of Rg evaluation over time. (D) Number of hydrogen bond averaged over time

The RMSF analysis shows no drastic fluctuation in the binding site region in which the drug is accommodated. The main regions exhibiting fluctuations are the loop regions and the fluctuations of protein in bound state have same pattern as in unbound state (apoprotein) as seen from the Fig. 7(B). In case of 7309-0137 and 4333-0420, the overall fluctuations were less as compared to the apoprotein, specifically in case of 4333-0420 it was observed that the fluctuations were significantly less than others in the high fluctuating loop regions and the complex is not only more stable than apoprotein but it was more stable than the other complexes too. On comparing RMSF of compounds identified through this study and the RMSF analysis of Rac1-NSC23766 complex as reported by Chunwen Zheng et al. it was observed that similar to results of present study the complex of Rac1-NSC23766 follow the same pattern of fluctuation as the apoprotein as well as have same fluctuating residues as observed in the present study and the peaks of fluctuating residues lies within same range as observed in the present study (Zheng et al. 2021). The RMSD and RMSF fluctuation analysis is capable to reflect the conformational changes occurring in the protein active site upon ligand binding. The results from both these analysis imply that the bound state of protein with the ligand is more stable in its conformation and have less deviation specifically in case of 4333-0420.

From Rg analysis it was observed that for the selected as well as reference (apoprotein), the difference in Rg values fluctuate within 1–2Å (Fig. 7(C)). For all the four molecules the structures were stable on the commencement of the simulation and their size and stability remained intact throughout 100 ns run. This indicates that the protein compactness do not varies much on binding with these ligands, thus suggesting that protein stability is retained on forming complex with the selected compounds.

Finally, the analysis of H-bonds was done to analyse the specificity of protein-ligand interactions. Also, to further explore the interactions between the selected compounds and RAC1 protein, the occupancies of hydrogen bonds were calculated (Fig. 8). The stability of docked complexes was validated by computing HB, paired with 0.35 nm donor and acceptor. The result indicates that selected compounds were able to make stable complexes through reasonable numbers of hydrogen bonds during the 100 ns MD simulation process (Fig. 7(D)).

Fig. 8.

Pair-wise hydrogen bond occupancy involved in interactions between selected compounds against RAC1. LAS 52449110 is shown in blue; 7309-0137 is shown in yellow, K292-1664 is shown in red and 4333-0420 in green

Binding free energy

For calculating binding free energies, the criterion of polar and apolar solvation was used. In this study, the binding energies related to the binding of selected four compounds i.e., LAS 52449110, 7309-0137, K292-1664 and 4333-0420 during the 100 ns MD simulations were calculated (Table 3). The final average binding energy calculations show that all four selected compounds have strong binding affinity for RAC1. In order to identify the hotspot residues of protein-ligand binding, the contribution of every residue was explored (Fig. 9). This analysis of free binding energy per residue was done to find out whether the selected residues favour the complex formation against the cluster of monomer waters or vice-versa. The negative energy contributions suggest that the residues favour complex formation whereas positive energy contributions suggest that the residue is not contributing in stabilizing the protein-ligand complex rather it is stabilizing monomers-water formation such as protein-water and ligand-water (Woods et al. 2014). The range of energy contribution of residues considered while defining the binding site in this study was between 2.44 kcal/mol to -7.79 kcal/mol. Residues ASP38 LEU55 TRP56 ASP63 ARG66 PRO69 were the major negative energy contributors favouring stability of protein-ligand complex. It was observed that the range of energy contribution of residues in case of 7309-0137 was between 5 kcal/mol to -7 kcal/mol. Residues ASP57 THR58 ALA59 GLY60 GLN61 and GLU62 were the major contributor in binding of 7309-0137 with RAC1. Among these residues, THR58 ALA59 GLY60 GLN61 and GLU62 contributed  -2.25 kcal/mol, -2.09 kcal/mol, -3.37 kcal/mol, -2.46 kcal/mol and  -1.27 kcal/mol, respectively. These residues contributed in formation of protein-ligand complex and were defined during grid generations. The other major contributors of negative energy favoring protein-ligand complex formation were GLU31 TYR32 ILE33 PRO34 and VAL36, these were not explicitly defined during grid formation but formed part of the defined 20Å grid box. In molecule 4333-0420, residues ASP38 ASN39 GLY54 LEU55 ASP57 THR58 ALA59 GLY60 GLU62 ASP63 TYR64 ARG66 PRO69 LEU70 were the major negative energy contributors. These make 14 out of 22 residues of the defined binding site, suggesting that the ligand was involved in stable interactions with the binding site. Similarly, in K292-1664 also majority of the residues contributed towards stability of protein-ligand complex. In all the selected compounds, more than 90 residues out of 176 residues of the binding domain of RAC1 contributed towards protein-ligand complex formation suggesting that the majority of residues were contributing to stabilizing the protein-ligand complex.

Table 3.

Comparison of MM-PBSA results for top ranking potential inhibitors

	LAS 52449110	7309-0137	K292-1664	4333-0420
Van der Waals energy (kcal/mol)	-25.81	-24.66	-37.96	-51.29
Electrostatic energy (kcal/mol)	-2.87	-2.75	-11.22	-13.28
Polar Solvation energy (kcal/mol)	9.83	11.79	28.74	34.10
SASA energy (kcal/mol)	-2.33	-1.82	-3.11	-3.83
Binding energy (kcal/mol)	-21.18	-17.44	-23.55	-34.30

Fig. 9.

Energy contribution of residues to binding of (A) LAS 52449110, (B) 7309-0137 (C) K292-1664 and (D) 4333-0420 with RAC1

ADME Analysis

The drug-likeliness of the selected compounds was studied using SwissADME tool from Swiss Institute of Bioinformatics. The screened results of ADME were summarized in Table 4, revealing various descriptors categorised under properties such as physicochemical properties, lipophilicity, water solubility, pharmacokinetics, drug-likeliness and medicinal chemistry. It can be seen that the selected compounds show violations because of their molecular weight fall outside the optimal range. LAS 52449110 from Asinex PPI library is the bulkiest of all four selected compound and due to its higher weight a few violations can be seen in the SwissADME data. The TPSA (topological polar surface area) of three of the four molecule was observed to be more than 140 Å which typically limits membrane permeability. The compounds that can adopt both membrane-permeable (low PSA) and water-soluble (high PSA) conformations have capability to expand beyond the borders of traditional drug space (Matsson and Kihlberg 2017). Most of the selected compounds have low lipophilicity and moderate water solubility allowing them a scope outside the traditional drug space, at higher molecular weight. The BBB permeant property measured through this analysis gives account of whether or not the drug is capable of crossing the blood brain barrier or BBB. The BBB is a network of blood vessels and cells that filters blood flowing to the brain and protect brain from harmful substances. It is essential for the drugs designed for treating brain tumors or tumors metastasized into brain to cross BBB. Since in HNSCC, the most common sites of distant metastases are the lungs (70%), the liver (42%), and the bones (15%) (Zbären and Lehmann 1987), therefore, compounds inability to cross BBB is acceptable and even desirable. The presence of gastrointestinal absorption or GI absorption is desired for anticancer compounds. From the ADME analysis, it is observed that though the level of absorption varies but all the identified compounds get absorbed in GI. The overall ADME analysis suggested that the most druggable compound among all four molecules was 4333-0420, it had most ADME properties well within the defined range and is not an inhibitor of any Cytochrome P450 isoenzymes. These predicted pharmacokinetic properties provide additional support to the identified molecules for further validation as therapeutics in HNSCC treatment.

Table 4.

Comparison of ADME results for top ranking potential inhibitors

Molecule	LAS 52449110	7309-0137	K292-1664	4333-0420
MW	778.96	455.49	564.7	346.36
#Heavy atoms	57	32	37	24
#Aromatic heavy atoms	21	17	15	6
Fraction Csp3	0.43	0.19	0.42	0.27
#Rotatable bonds	10	11	11	4
#H-bond acceptors	6	7	8	5
#H-bond donors	7	4	3	1
MR	238.92	117.25	143.07	98.06
TPSA	174.96	171.74	223.37	124.45
iLOGP	4.44	2.56	2	2.03
XLOGP3	4.67	0.86	2.69	0.37
WLOGP	0.32	1.19	3.53	-0.52
MLOGP	-2.34	0.68	1.22	0.94
Silicos-IT Log P	4.78	1.17	2.77	0.76
Consensus Log P	2.37	1.29	2.44	0.72
ESOL Log S	-7.22	-2.87	-4.61	-2.14
ESOL Solubility (mg/ml)	4.65E-05	6.10E-01	1.39E-02	2.50E + 00
ESOL Solubility (mol/l)	5.97E-08	1.34E-03	2.46E-05	7.22E-03
ESOL Class	Poorly soluble	Soluble	Moderately soluble	Soluble
Ali Log S	-8.07	-4.05	-7.03	-2.55
Ali Solubility (mg/ml)	6.61E-06	4.06E-02	5.23E-05	9.79E-01
Ali Solubility (mol/l)	8.48E-09	8.91E-05	9.26E-08	2.83E-03
Ali Class	Poorly soluble	Moderately soluble	Poorly soluble	Soluble
Silicos-IT LogSw	-12.56	-5.1	-5.82	-2.94
Silicos-IT Solubility (mg/ml)	2.13E-10	3.64E-03	8.61E-04	4.00E-01
Silicos-IT Solubility (mol/l)	2.74E-13	8.00E-06	1.52E-06	1.15E-03
Silicos-IT class	Insoluble	Moderately soluble	Moderately soluble	Soluble
GI absorption	Low	Low	Low	High
BBB permeant	No	No	No	No
Pgp substrate	Yes	Yes	Yes	No
CYP1A2 inhibitor	No	No	No	No
CYP2C19 inhibitor	No	No	No	No
CYP2C9 inhibitor	No	No	No	No
CYP2D6 inhibitor	No	No	No	No
CYP3A4 inhibitor	Yes	No	Yes	No
log Kp (cm/s)	-7.74	-8.47	-7.83	-8.15
Lipinski #violations	3	0	1	0
Ghose #violations	3	0	2	1
Veber #violations	1	2	2	0
Egan #violations	1	1	1	0
Muegge #violations	3	1	1	0
Bioavailability Score	0.17	0.11	0.11	0.55
PAINS #alerts	0	0	0	0
Brenk #alerts	0	0	0	0
Leadlikeness #violations	3	2	2	0
Synthetic Accessibility	6.78	3.86	4.63	4.38

Conclusions

In this study, Rho family GTPase RAC1 was studies and compounds were identified for its selective inhibition. Different computational approaches were employed to screen and validate various compounds against RAC1. Various repositories of chemical compounds targeting protein-protein inhibition such as Asinex PPI, ChemDiv PPI and Enamine PPI were used to perform structure based virtual screening. The evaluation of the compounds from several huge databases, allowed the identification of four potent inhibitors of RAC1. Three of the selected four compounds were from the ChemDiv PPI and the remaining one was from the Asinex PPI. The compounds were identified such that they are capable of inhibiting activation of RAC1 protein by binding to the GEF binding and catalytic site of the protein. The interaction energies of the compounds were calculated using both in the virtual screening and molecular dynamics approaches. From the docking studies it was observed that the selected compounds sits in the defined binding pocket and the MD simulation analysis helped in observing greater conformations of hydrogen bonds between the selected compounds and RAC1. Moreover, the stability of the complex is confirmed by the existence of these interactions throughout the simulation. The result of this study were also compared with the previous theoretical and experimental studies of EHop-016 and NSC23766 and it was observed that the compounds LAS 52449110, 7309-0137, K292-1664 and 4333-0420 identified in this study sit in the same binding cleft of RAC1 where EHop-016 and NSC23766 were experimentally validated to inhibit binding of RAC1 with GEFs. The compounds selected from this study interact with residues of the cleft forming hydrogen and non-hydrogen bond interactions significant in GEFs binding to RAC1. This implies that the selected compound could be potential inhibitors of RAC1. This study used various computational approaches to predict the inhibitory effects of four new compounds against RAC1. This could provide an initial thrust to future development and studies for the discovery of new RAC1 inhibitors.

Funding

This work was supported by DST-Purse and Drugs and Pharmaceuticals Research Programme by Department of Science and Technology under Sanction Order [number VI-D&P/546/2016-17/TDT]; and first author was supported by CSIR fellowship under File No. [09/263(1132)/2017-EMR-I] to pursue Ph.D.

Data Availability

Not applicable.

Code Availability

Not applicable.

Declarations

Conflict of interest

The authors report no conflict of interest.

Ethics approval

Not applicable.