In silico identification and characterization of small-molecule inhibitors specific to RhoG/Rac1 signaling pathway

Abstract Rho family GTPases serve as molecular switches in numerous cellular processes, and their overexpression is involved in disease conditions. RhoG is one of the less explored Rho GTPases with significant sequential and structural homology with Rac1. Experimental mutations in RhoG (i.e., RhoGG12V and RhoGQ61L) are shown to dysregulate cell migration. Thus, targeting upstream activators of RhoG, such as guanine nucleotide exchange factors (GEFs), maybe an important strategy for inhibiting RhoG activation. In the current study, we have modelled the 3D structure of RhoG with greater accuracy as confirmed through PROCHECK, ProSA, and Verify3D. Our results indicate that 90.4% of residues are in the Ramachandran plots favoured region, with the Z-score of –6.46, and 87.96% of residues had an averaged 3D–1D score ≥0.2. Further, we have evaluated and binding dynamics of ten Rac1 inhibitors to investigate their potential to inhibit RhoG by targeting GEFs binding grooves. To this end, the binding energy of the docked complexes of the wild-type (WT) RhoG and its mutant proteins with inhibitor molecules was calculated using the MM/PBSA method. Our results from docking studies showed that macrolide1 binds efficiently with the GEF site of WT RhoG and the mutants mentioned above. However, an extensive analysis using MD simulations (200 ns) showed that the Rac1 based inhibitor, EHop-016, and NSC23766 might bind with greater affinity to GEF sites of mutants and WT RhoG. Thus, the results from the study indicate that Rac1 inhibitors have the potential for use as therapeutics in conditions involving dysregulation of RhoG. Communicated by Ramaswamy H. Sarma


Introduction
Small monomeric G-proteins of Rho family GTPases function as molecular switches that control different signalling processes that regulate cytoskeleton reorganization, gene transcription, cellular motility, cell cycle progression, and other cellular functions (Etienne-Manneville & Hall, 2002). Rho GTPases can be activated through interaction with guanine nucleotide exchange factors (GEFs), catalyzing the GTP/GDP exchange. At the same time, the inactivation of Rho proteins is mediated by GTPase activating proteins (GAPs) (Kumawat et al., 2017). Reports suggest that 80 GEFs and 70 GAPs are known to date, and their interaction with Rho proteins is associated with activation and inactivation of distinct downstream signalling pathways by serving as core nodes for integrating and propagating signals from the extracellular stimulus (Parri & Chiarugi, 2010). RhoG (Ras homolog growth-related) is a member of the Rac (Ras-related C3 botulinum toxin substrate) subfamily, with close sequence homology with Rac1 and Cdc42 (Cell division control protein 42 homologs). Similar to other Rho GTPases, RhoG is also shown to be involved in multiple cellular processes such as cell adhesion, migration, phagocytosis, neurite cone growth, T cell gene transcription, and neutrophil NADPH oxidase activation (Goicoechea et al., 2017). Besides, RhoG and Rac1 overexpression or hyperactivation are also shown to facilitate cancer progression through invadopodium formation in glioblastoma and breast cancer models (Goicoechea et al., 2017;Kwiatkowska et al., 2012). Therefore, targeting the binding of GEFs to RhoG/Rac1 proteins is a viable strategy to inhibit the activity of these proteins and regulate cancer invasion.
Previous research has shown that mutations in RhoG at the 12th (RhoGG12V) and 61st (RhoGQ61L) positions result in a constitutively active phenotype (Prieto-S anchez & Bustelo, 2003). In this study, dysregulation of HEK-293, NIH3T3 migration is caused by a constitutively active mutation in RhoG, which is analogous to the phenotypically defective migration caused by Rho protein overexpression or hyperactivation (Hauser et al., 1994;Prieto-S anchez & Bustelo, 2003;Wennerberg et al., 2002). The current information is available in various databases, including HGMD (http://www.hgmd.cf. ac.uk/ac/index.php) and NCBI (https://www.ncbi.nlm.nih.gov), point towards the possible presence of mutations in RhoG that could be present in the human genome and could be associated with incidences of cancers. Also, in our study, mutational analysis of RhoG mutants (RhoGG12V & RhoGQ61L) by computational tools correlates with cancer incidence. Therefore, it is essential to investigate compounds that inhibit WT RhoG and its mutants with a possible association with pathological conditions. Recently, some RhoG specific inhibitors were screened through in silico approaches, but they are not validated in experimental systems (Dasari et al., 2017). Previously, In vitro and in vivo studies showed that different chemical inhibitors such as NSC23766, EHop-016, EHT1864 bind with GEFs binding grooves in Rac proteins and inactivate the downstream cellular pathways (Bouquier et al., 2009). The GEF binding grooves of RhoG shares sequence and structural homology with the GEF binding grooves of Rac1 (Meller et al., 2008). Thus, we hypothesized that experimentally validated inhibitors of Rac1-GEF may also function as RhoG inhibitors via targeting GEF binding domains. In the present study, in addition to modelling the 3D structure of RhoG with greater accuracy, we have performed the virtual screening of ten known Rac-GEFs inhibitors, namely, migrastatin, iso-migrastatin, macroketone, EHop-016, isoalvaxanthone, NSC23766, macrolide1, macro-lide3, and macrolactam for their possible implication in targeting RhoG-GEFs interaction. The MD simulation and MM/ PBSA studies of the docked complex showed that NSC23766 and EHop-016 could bind with GEF grooves of WT RhoG and its mutants with higher binding energy. Thus, the results of this study indicate that the specific inhibitors of Rac1 have the potential to target the RhoG molecule, and they could make potential candidates for pathological conditions such as cancers associated with RhoG hyperactivation.

Materials and methods
2.1. FASTA sequence retrieval, sequence alignment, prediction of secondary structure, homology modelling of human RhoG and refinement of the model in YASARA server The FASTA sequence of Human RhoG (Uniport Id: P84095) was retrieved from the UniProt database (https://www. uniprot.org) (Apweiler et al., 2004), and the PSI-PRED 4.0 (http://bioinf.cs.ucl.ac.uk/psipred/) was used to predict the secondary structure of RhoG (Jones, 1999) in order to compare and validate the structure obtained from homology modelling. Since the crystal and NMR structure of RhoG is not available in the Protein Database (PDB) (Berman et al., 2003), the homology modelling of RhoG was performed using Modeller 9.23 (Eswar et al., 2006). In order to perform homology modelling, the template was searched and selected by performing Protein BLAST (BLAST-P) with Protein Data Bank (PDB) as the target database (Altschul et al., 2005). The selection of the best template was performed based on identity, query coverage percentage and E-value score. ClustalW tool (http://www.genome.jp/tools/clustalw/) is used to perform the sequence alignment of RhoG with its template (Larkin et al., 2007). Further, out of 20 models generated through modeller, the best model was selected by assessing the lowest modeller objective function such as DOPE, molpdf and GA341 score. Also, the quality of the generated model was checked by aligning the generated best model and template used for modelling. Furthermore, the stereochemical stability of the predicted structure was studied using SAVES v5.0 (https://servicesn.mbi.ucla.edu/ SAVES/), and the visualizations were performed using PyMOL 1.3 (Schrodinger, 2010). Further, the energy minimization of the best modelled RhoG protein was performed by submitting a generated pdb file to the YASARA server (Krieger et al., 2009). Before energy minimization, the server performed several clean-up procedures, including predicting the optimal amino acid rotamer using the SCWRL algorithm and then tweaking dihedral angles of rotamers using the YASARA2 force field. The hydrogen-bonding network is then optimised, and the solute is embedded in the solvent shell. The side-chain pKas are predicted, and protonation states are fine-tuned at pH 7.4. After that, the steepest descent minimization is run to remove any remaining clashes, followed by a simulated annealing minimization with atom velocities scaled down by 0.9 every ten steps to get a stable local energy minimum. The minimization process ends when energy improves less than 0.05 kJ mol À1 per atom during 200 steps.

Structural validation, in silico pathogenicity, protein stability, and conservation analysis of modelled human RhoG proteins
The quality of the 3D modelled structure of human RhoG protein was analyzed using PROCHECK (Ramachandran plot), Protein Structure Analysis (ProSA), and Verify3D. The stereochemical stability of the modelled protein was checked in the Ramachandran plot by measuring the w (Psi) and u (Phi) angle potential energy of the backbone against the residues of amino acids in the protein structure (Laskowski et al., 1993). Additionally, Verify3D was used to evaluate the consistency of the RhoG atomic model against its template (Bowie et al., 1991;L€ uthy et al., 1992), and ProSA considers defects in 3D modelled proteins (Wiederstein & Sippl, 2007), where Z-score measures the overall performance of the model, a negative Z-score value indicates a better-quality model (Pontius et al., 1996). Some of the frequently used pathogenicity predictors like FATHMM (Functional Analysis through Hidden Markov Models) (Shihab et al., 2013), SIFT (Ng & Henikoff, 2003), SNAP (Johnson et al., 2008), PhD-SNP (Capriotti et al., 2006), and PolyPhen-2 (Adzhubei et al., 2010), were engaged onto the mutants RhoG followed by I-Mutant suite (http://gpcr2.biocomp.unibo.it/cgi/ predictors/I-Mutant3.0/I-Mutant3.0.cgi) analysis to know the shift in protein stability due to mutation (if any). FATHMM produces the output score on the scale of zero to less than zero, where a neutral mutation receives zero scores while a pathogenic mutation receives less than zeros scores. This helps to discriminate between neutral and cancer in the occurring mutations. SIFTs sequence alignment method was used to predict mutational pathogenicity and produce output as 1 for neutral mutations and 0 for highly deleterious mutations. SNAP segregates non-neutral and neutral mutations through a neural network (NN) based on general polymorphisms exercising secondary structures, practical effects, conservation knowledge, and solvent approachability information from different sources. PhD-SNP computation is built on single sequence Support Vector Machines (SVMs) to demarcate neutral and disease-linked mutations. PolyPhen2 predicts structural and functional consequences on possible amino acid substitution using straightforward physical and comparative deliberation. I-Mutant Suite is also an SVMs based platform that accepts input as a protein sequence or structure (if available) and aids in predicting protein stability change after a particular amino acid substitution. Moreover, the Consurf server (Glaser et al., 2003) analysed the conserved and functionally relevant regions in a protein. The input to this server was taken as a FASTA sequence of RhoG, and multiple sequence alignment was developed for it. The output from this server contains information about every amino acid sequence in the form of a conservation score, which directly correlates to their conserved nature.

Molecular dynamics simulation (MDS) of WT RhoG and mutant RhoG
GROMOS96 43a1 force field in the GROMACS v2018.2 package was used on a LINUX based workstation to carry out Molecular Dynamics simulations for the WT and mutant RhoG (Van Der Spoel et al., 2005;van Gunsteren et al., 1996). The conversion of pdb to the gro file, which is a readable file format of GROMACS, was produced using the pdb2gmx option, and PRODRG 2.5 server was used for the generation of ligand topology, localized in a cubical box through SPC/E water molecules, and kept at a boundary distance 1.0 nm from the corner of the box (Sch€ uttelkopf & Van Aalten, 2004;Van Aalten et al., 1996). The system was further neutralized with the 'genion'option, then energy minimization for 50,000 stages with the highest force equal to 1000.0 kJ mol À1 nm À1 applied. The algorithm employed here to constrain the bond lengths was 'steepest descent minimization, while the 'Particle Mesh Ewald' method was used to measure the electrostatic interaction. The entire system was further equilibrated via position-restrained dynamics simulation (NVT and NPT) at 300 K for 100 ps. Lastly, this well-equilibrated system was exposed to an MD simulation run for 200 ns at a temperature of 300 K and a pressure of 1 bar. Various GROMACS options like rms, rmsf, gyrate, h-bond, and sasa were used to perform analysis of RMSD, RMSF, Radius of gyration, hydrogen bond, and solvent accessible surface area (SASA), respectively. All the graphs were plotted using the Grace program.
2.4. Molecular docking of Rac based inhibitor with WT RhoG and their mutants RhoG G12V and RhoG Q61L receptor The structure of ligands such as migrastatin, iso-migrastatin, macroketone, EHop-016, isoalvaxanthone, and NSC23766 was retrieved from PubChem (Kim et al., 2019), and the remaining structures such as macrolide1, macrolide3, macrolactam were built-in and optimized in ACD chem sketch software (Hunter, 1997). Further, energy minimization of ligands was performed with the help of the Avogadro 1.0.1 tool (Hanwell et al., 2012

Molecular dynamic simulation of proteinligand complex
The Molecular dynamic study was conducted to analyze the protein-ligand complexes flexibility and stability in the appropriate hydration medium, i.e., WT RhoG and mutant RhoG G12V RhoG Q61L with all ten complexed ligands as described in the previous section. The protein complex was solvated, and counterions were added as explained above, and the energy minimization step was performed using the steepest descent algorithm for 50000 iteration steps with a maximum force up to 1000 kJ mol À1 nm À1 . Further, the conjugate gradient algorithm with the same parameter is used to achieve the energy minima of the system. In addition to this, the system was further equilibrated with NVT and NPT ensemble and LINCS algorithm used for covalent bond position restrained. The NVT equilibration step was accomplished with a constant number of particles, volume, and temperature, for 1000 ps of each step of 2 fs. Similarly, the Berendsen barostat pressure coupling method was used for NPT equilibration with a constant number of particles, pressure, and temperature at 300 K for 1000 ps of each step of 2 fs. The 12 Å radius cut-offs were used to measure shortrange interactions, such as Lennard-Jones and Coulomb interactions. Particle Mesh Ewald (PME) method with Fourier grid spacing of 1.6 Å was used for long-range electrostatics (Harvey & De Fabritiis, 2009). The system constant temperature was maintained using V-rescale, a modified version of the Berendsen temperature coupling method (Bussi et al., 2007). Next, the MD production run was performed for 200 ns with each step of 2fs. Lastly, the analysis of the stability and flexibility of WT RhoG and mutant RhoG G12V , RhoG Q61L with EHop-016 and NSC23766 complex was performed in RMSD, RMSF, radius of gyration, hydrogen bond, and conformational flexibility analysis of complexes using 20000 frames of MD trajectory. The hydrogen bond analysis was performed using grid space of 20x20x20, and with Radius, set the cut-off to 0.35 nm. Furthermore, the principal component analysis (PCA) and free energy landscape (FEL) analysis were performed using a python script. The free energy of receptors was expressed in kJ mol À1 units.

Free energy calculation of complex by MMPBSA method
In the GROMACS module, the g_mmpbsa tool was employed to calculate the binding energy of biomolecular associations like protein-ligand complexes accomplished by the Molecular Mechanic/Poisson-Boltzmann Surface Area (MMPBSA) method using trajectories derived from molecular dynamic (MD) simulation (Kumari et al., 2014). Additionally, it also calculates residue-wise contribution to the total binding energy, which will give insight into critical contributing residues to the protein-ligand association. The stable and well-equilibrated trajectory of WT RhoG and mutant RhoG G12V , RhoG Q61L with all ten ligand complexes was selected to calculate free binding energy. Generally, the binding free energy, DG bind , was determined by the following equation from the free energy of the receptor-ligand complex (G rlc ) with respect to the unbound receptor (G rec ) and ligand (G lig ): MMPBSA allows the estimation of binding free energy of ligand and desolvated protein. The binding energy is the sum of average potential energy in the vacuum, polar solvation energy, and non-polar solvation energy. In this study, frames of every 10 ps between 190 ns and 200 ns were chosen to estimate the binding free energy, which was expressed in units of kJ mol À1 .

Density functional theory (DFT) calculations
The 3D structure of EHop-016 and NSC23766 extracted from 200 ns snapshot used for Density Functional Theory (DFT) calculation with the help of Gaussian 09 W using the DFT/ Becke's three-parameter hybrid functional combined with Lee-Yang-Parr correlation functional (DFT/RB3LYP) method with 6-311 G (d,p) basis set (Becke, 1992;Schlegel, 1982). DFT calculation output was used to study the molecular electrostatic potential (MEP) and applied to determine the nucleophilic and electrophilic group within the ligand by finding the highest occupied and lowest unoccupied molecular orbital (HOMO and LUMO). The visualization of these compounds' MEP and HOMO, LUMO state was performed with the help of Gauss view 5.0.8.

Secondary structure prediction & homology modelling of human RhoG
The primary amino acid sequence of RhoG in FASTA format was retrieved from the UniProtKB, consisted of 191 amino acid residues and secondary structure prediction with PSI-PRED 4.0 showed that this protein consists of 6a-helices & 7b-sheets with a higher confidence level of prediction ( Figure 1(a)). Subsequently, the 3D structure of RhoG protein was generated as described in the methodology. According to the results obtained in the BLAST, the protein Rac1-GDP complexed with Arfaptin p21 (PDB ID: 1I4D) was selected as a template with 72% sequence identity, E-value (1e-99), and 100% query coverage (Figure 1(b)). Twenty models were produced using Modeller 9.23 software, and the best model had DOPE score, mol pdf, and GA341 value was found to be À20514.02, 1262.11 and 1.00, respectively. As shown in Figure 1(c), the N-and C-terminal domains of the generated 3D model of RhoG represented the starting and ending residues, respectively. Additionally, the 3D structure of RhoG comprised of a-helices and b-strands with putative GEF sites containing a group of amino acids such as Lys 5 , 3.2. The generated 3D model of RhoG is consistent, stereochemical stable, and has better quality as predicted using multiple computational tools The structural validation of the generated modelled RhoG protein was accomplished by multiple computational tools such as PROCHECK, ProSA, and Verify3D. The Ramachandran plot was used to examine the stereochemical stability of a modelled RhoG, which was performed by providing the PDB file of RhoG to the SAVES v5.0 server (https://servicesn. mbi.ucla.edu/SAVES/). Previous literature had demonstrated that the model with 90% residue in the most favoured region and the additional allowed region is regarded as a stereochemical stable model (Laskowski et al., 1993). Accordingly, as shown in Figure 2(a), the Ramachandran plot of model RhoG revealed that 99% of amino acid residues remained in the most favoured region and the additional allowed region confirming the stereochemical stability of the 3D model. The 3D model of RhoG protein was further validated using ProSA, which provides insight into the error in a modelled structure and the overall energy interaction value, as measured from the Z-score  (Pontius et al., 1996;Wiederstein & Sippl, 2007). As shown in Figure 2(b), the generated RhoG model had a Z-score of -6.46, suggesting that most of the amino acid residues of modelled RhoG come under the negative region.
Furthermore, the consistency of the model was evaluated using the Verify3D tool. As shown in Figure 2(c), 88% of residues had an averaged 3D-1D score !0.2, indicating the better quality of the generated model. Ramachandran plot signifies the 3D model's stereochemical stability with a central area is depicted in red color, additionally enabled region in the black, permitted area in the light yellow and white field reflects disallowed zone. (b) ProSA plot of Human RhoG showed that the overall quality model examined by Z-score (dark spot) is -6.46. The Z-score protein falls in the dark blue region (NMR) spectrum and light blue area (X-ray), for all existing crystal structures with comparable residue numbers an abscissa. (c) Verify3D profile gives insight into the compatibility of atomic models of Human RhoG, indicating that 87.96% of residues had an averaged 3D-1D score !0.2.

Conservation analysis of RhoG and in silico pathogenic and stability prediction of mutations
In the process of evolution, functionally important residues are known to be conserved in various groups of proteins, and any mutation in the conserved residues is known to interfere in the functionality of the protein (Nygaard et al., 2016). Supplementary material Figure1 shows a conservation chart depicting the information of all the conserved residues in RhoG protein. Existing reports suggest that specific amino acid substitution at the 12th and 61st position in RhoG protein results in constitutive activation of RhoG and promotes cell migration (Meller et al., 2008;Prieto-S anchez & Bustelo, 2003). Looking at the conservation score of 12th and 61st position in RhoG obtained from the Consurf server, the latter seemed to be more conserved than the former and may impact the structure and function of RhoG protein.
Thus, to understand the effect of the point mutation in a more conserved (61 st position) Vs. Less conserved position (12th position), computational prediction tools were employed. The fact that point mutations in proteins result in a change in protein structure (Schaefer & Rost, 2012), but how exactly it affects the functionality of RhoG is still not clear. So, multiple in silico tools based on different algorithms were used to predict the effects of the point mutation in RhoG protein, and their results in terms of stability and pathogenicity are shown in Table 1. The FATHMM tool predicted a less than zero score for each of the substitutions of valine and leucine at the 12th and 61st position, respectively, which may lead to the cancerous nature of RhoG. As mentioned in the methodology, FATHMM gives zero scores for neutral and less than zero scores for pathogenic mutation.
Similarly, the SIFT and PolyPhen-2 tools also anticipated the change in protein function due to both amino acid substitutions. Further, SNAP and PhD-SNP output based on NN and SVMs demonstrated that both the substitutions have deleterious effects and could give rise to disease conditions. On the stability front, I-Mutant Suite (based on SVMs) calculates a decrease in the stability due to the valine substitution at the 12th position, and the same decrement in the stability was observed due to leucine substitution at the 61st position.

MD Simulation analysis of WT RhoG protein and its mutant RhoG G12V & RhoG Q61L
To inspect the noted behaviour of WT RhoG protein and its mutant RhoG G12V and RhoG Q61L were annotated by a molecular dynamics simulation for 200 ns. At the end of the simulation, the RMSD for WT and both mutant proteins were obtained via 'rms' function against the energy minimized structure, and as shown in Figure 3(a), RMSD value of $0.5 nm and $0.39 nm was obtained for WT and G12V mutant RhoG, respectively. Interestingly, the RMSD value for Q61L mutant RhoG came out to be around $0.40 nm compared to the other two counterparts. The greater RMSD value of the WT RhoG indicates its prominent destabilizing nature with respect to the two mutants.
Next, the 'rmsf' function of GROMACS was used to investigate the Root mean square fluctuation of all the three RhoG proteins analysed above. As shown in Figure 3(b), the highest fluctuation was observed in the WT protein followed by G12V and Q61L mutant proteins at various residue positions, e.g., $15, $62, $130 and $135. Next, to further understand the effect of the mutation on the structural aspect of the protein, the packaging of atoms in the vicinity of the central axis was analysed through the Radius of gyration (Rg). The GROMACS function 'gyrate' finds a structure's compactness (depicted by lesser radii value) compared to a less compact structure (depicted by larger radii value). As shown in Figure  3(c), a lesser radii value of $1.54 nm in G12V indicated it to have a more compact structure than Q61L ($1.57 nm). However, the WT protein has larger radii compared to the mutant proteins. Next, the 'hbond' function in GROMACS was used to calculate the intermolecular and intramolecular hydrogen bonds. As shown in Figure 3(d), the intermolecular bonding for all the proteins (WT and mutants) decreased as the simulation progressed. The intramolecular hydrogen bonding (Figure 3(e)) increased during the simulation for all the proteins, indicating the change in mutant proteins stability. Further, the SASA calculation was performed by the 'sasa' function to find the solvent molecule accessibility on the protein surface. As shown in Figure 3(f), the SASA value decreased from Q61L to G12V mutant proteins and finally to the WT RhoG protein. Thus, the above observations demonstrate a structural impact on the protein structure due to the mutations.

Molecular docking of the ligands with the WT RhoG and its mutant proteins
Previous studies have shown that mutation in the RhoG protein leading to its constitutive activation was associated with a dysregulation in cell migration. Such dysregulation in cell migration and functions are shown to be linked with a variety of cancers and inflammatory diseases (Hauser et al., 1994;Prieto-S anchez & Bustelo, 2003). Therefore, we next searched for potential inhibitors of WT RhoG and their constitutively active mutants RhoG G12V and RhoG Q61L . To this end, screening of lead compounds for both WT and mutant RhoG e.g. RhoGG12V, RhoG Q61L , was accomplished by performing site-specific docking of Rac1 based inhibitors using Autodock Vina 1.1. The Rac1 based inhibitor such as migrastatin and its analogues iso-migrastatin, macrolide1, macro-lide3, macroketone, macrolactam, and NSC23766, EHop-016, ITX3, and other compounds such as isoalvaxantone (Gao et al., 2004;Shan et al., 2005;Wang et al., 2010) were docked to the GEF site of both WT and mutant RhoG proteins. The docking results showed that all ligands fit into the GEFs grooves of WT RhoG, RhoGG12V & RhoGQ61L (supplementary material Table 2). As demonstrated in Table 2, the binding affinity of all ten ligands for WT RhoG ranged from -8.2 kcal mol À1 to 6.4 kcal mol À1 . Among all the known Rac1 inhibitors, macrolide1 showed a maximum binding affinity with WT and mutants. Furthermore, macrolide1 was found to be hydrophobic and binds to the hydrophobic pocket of WT RhoG with two hydrogen bonds (Phe37, Ala59). Similarly, RhoG mutants such as RhoGG12V and RhoGQ61L could form three hydrogen bonds (Asp57, Ala59 & Leu67) and two hydrogen bonds (Ala59 & Leu67), respectively. The 3D & 2 D interaction profile diagram and its surface representation of the macrolide1 with WT and mutant RhoG protein complexes are shown in supplementary material Figure 2(a,c,e) and Figure 2(b,d,f), respectively. Previous reports have shown that the accuracy of free energy prediction for most computational docking approaches, including AutoDock, is roughly 2-3 kcal mol À1 standard error (Cosconati et al., 2010). Therefore, next, we performed MMPBSA analysis of MD trajectories of all docked complexes. The docking results showed that EHop-016 and NSC23766 bind to GEFs grooves of WT RhoG and its mutants (Figures 4 and 5), and their binding energies with WT RhoG and its mutant were more compared to macrolide1. The binding affinity of docked ligands EHop-016 and NSC23766 with their polar interaction with WT RhoG and its mutant RhoG is mentioned in Table 2. To further understand the interaction dynamics of EHop-016 and NSC23766 with RhoG proteins, we performed molecular docking analysis. As shown in Figure 4(a,c,e), the EHop-016 formed two hydrogen bonds (Ala 57 & Ala 59 ) with WT RhoG and RhoG Q61L while RhoG G12V formed two hydrogen bonds with amino acids Asp 38 & Ser 41 . Moreover, EHop-016 binds to the hydrophobic pocket of WT RhoG and its mutant complex shown in Figure  4(b,d,f). Similarly, as shown in Figure 5(a,c,e), NSC23766 binds with two hydrogen bonds (Asp 38 & Asn 39 ) with WT RhoG and three polar contacts (two with Asn 39 & one with Ala 57 ) with RhoG G12V , while the mutant RhoGQ61L bind with NSC23766 by forming single polar contact (Leu 67 ). Furthermore, NSC23766 is fit into the hydrophobic grooves of WT RhoG and its mutant, as shown in Figure 5(b,d,f). Furthermore, molecular docking studies were also performed with Rac1, and the polar interaction between ligand and Rac1 are represented in Table 3. Supplementary material Table 3 shows the detailed interaction profile and the binding energy of various complexes. The docking of all ten ligands with Rac1 suggests that the binding energy and binding pattern are analogous to RhoG. The detailed analysis of hydrogen bonding between EHop-016 and NSC23766 with WT RhoG and its mutants is described in Table 4.

Density functional theory calculation of EHop-016 & NSC23766 compound
Next, to understand the protein-ligand interactions, we performed the DFT modelling as described in the methodology to estimate the molecular and electronic behaviour of EHop-016 and NSC23766. As shown in Figure 6(a,c), the distribution of electron-rich and electron-poor regions of EHop-016 and NSC23766 was determined by the electrostatic potential. The DFT modelling shown in Figure 6(a,c) represents the distribution of electron-rich and electron-poor regions of EHop-016 and NSC23766, respectively, determined by the electrostatic potential. Additionally, the calculation of frontier molecular orbitals (FMOs) contributes significantly in predicting the electrical and optical properties of a molecule using HOMO (highest occupied molecular orbital) and LUMO (lowest unoccupied molecular orbital) analysis. The propensity of a molecule to transfer an electron is described as HOMO, whereas the ability to obtain an electron is known as LUMO.
As depicted in Figure 6(b), the illustrative manifestation of isodensity plots of the energy transition for EHop-016 shows that in the ground state (HOMO), the density is distributed throughout the molecule, mainly at ethyl carbazole moiety and morpholinopropyl ring. In contrast, in LUMO, the charge density is mainly condensed throughout the molecule. In the case of NSC23766, the charge is evenly distributed at the methyl pyrimidine and methyl quinoline ring in HOMO, whereas in LUMO, charge density is condensed methyl pyrimidine and diethylamino-pentane moiety. The methyl pyrimidine group also lost the density from HOMO to LUMO, indicating its role in biological properties (Figure 6(d)). The interacting moieties of ligands signifying the charge transfer and participation in the bond formation of active site residues of receptors show that simulated charge density lies statistically close to the experimental parameters. As per the literature, an increase in the energy gap between HOMO and LUMO of ligands suggest an enhanced binding with a receptor (Anitha et al., 2020;Singh et al., 2020). As shown in Figure 6(b,d), the HOMO and LUMO energy gap for EHop-016 and NSC23766 were 1.22 eV and 0.74 eV, respectively.

Stability and functionality analysis of the MD trajectories of WT RhoG and mutant complexed with EHop-016 and NSC23766
Next, to examine the flexibility and stability of the proteinligand complexes, a molecular dynamics study was performed on the WT RhoG, RhoG G12V , or RhoG Q61L complexes with EHop-016 and NSC23766. The quality of 200 ns simulation of the protein-ligand complex was examined via temperature (T), pressure (P), density (D), potential energy (E_p), and total energy (E) plots (supplementary material Figures 3  and 4). The temperature (300 K), pressure (within the range of -100 to þ100 bar), and average density (1000 kg m À3 ) of the system was nearly constant during the entire simulation. Moreover, the potential and total energy of the system were consistent for all complexes. Thus, the plot analysis indicated that the protein-ligand complex shows consistent behaviour in the system, and simulation is of acceptable quality.  The functionality of the protein-ligand complex is solely determined by the folding and backbone movement within the system. Thus, following the molecular dynamics analysis, the stability of the protein complexes was further analysed by performing RMSD, RMSF and Rg analysis. The RMSD analysis provided insight into the structural deviation in Ca-backbone and the protein's side chain while performing MD simulation of 200 ns. Thus, the time-dependent RMSD graph against the energy minimized structure was plotted for the Ca-backbone of WT RhoG, RhoG G12V & RhoG Q61L complex with EHop-016 and NSC23766. The RMSD plot shown in Figure 7(a), indicates that the WT RhoG, & RhoG G12V have RMSD of $0.4 nm & $0.45 nm respectively, while the RMSD value for RhoG Q61L is $0.5 nm, complexed with EHop-016, suggest that WT RhoG is more stable compared to mutant protein. Similarly, the RMSD value of WT RhoG, RhoG G12V & RhoG Q61L complexed with NSC23766 was ($0.4 nm), ($0.5 nm) ($0.35) respectively, suggest that mutant RhoG Q61L is more stable than WT RhoG and mutant RhoG G12V as shown in Figure 8(a). All the studied protein-drug complexes initially demonstrated higher RMSD fluctuation, but they attained stability after 175 ns and remained stable till 200 ns. It was also observed that WT RhoG, RhoG G12V , and RhoG Q61L complexes have a moderate difference in their backbone RMSD value. The overall RMSD result implies that the binding of EHop-016 and NSC23766 in the active site of WT RhoG, RhoG G12V , and RhoG Q61L is stable.
Next, RMSF analysis was performed to understand deviation concerning the active site residues with respect to the reference structure used in the MD simulation studies. The RMSF graph was plotted for each residue of the above protein and drug (EHop-016 & NSC23766) complexes using 20000 frames. As shown in the Figure 7(b), the RMSF plot indicated that the interacting active site residues, Lys 5 , Thr 35 , Val 36 , Asp 38 , Asn 39 , Ser 41 , Gln 43 , Asn 52 , Trp 56 , Asp 57 , Ala 59 , Gly 60 , Gln 61 , Leu 67 & Leu 70 showed lesser fluctuation in WT RhoG-EHop-016 compared to RhoG G12V -EHop-016 complex and RhoG Q61L -EHop-016. The active site residues of all the complexes showed moderate RMSF deviation. This indicates that the binding of EHop-016 does not show significant changes in the position of active site residues of the protein complexes and further confirmed the stability of the same. Similarly, the lesser RMSF fluctuation is observed in the WT RhoG-NSC23766 complex compared to RhoG G12V -NSC23766 and RhoG Q61L -NSC237766 complex. Thus, the RMSF analysis indicates that the binding of EHop-016 and NSC23766 does not show significant changes in the reference position of active site residues of the protein complexes and further confirmed the stability of the same. Moreover, protein compactness during the MD simulation was also studied using the Radius of gyration (Rg) analysis. The Rg Graph was plotted for protein complexes with EHop-016 using 200 ns trajectories as represented in Figure 7(c). The Rg plot indicates that the Rg value for WT RhoG ($1.7 nm), RhoG Q61L ($1.65 nm) and RhoG G12V ($1.65 nm) confirming that the mutant protein is more compact than WT, but at the end of the simulation, all complexes did not show a significant change in the Rg value of the above complex. Moreover, Rg analysis of proteins complexed with NSC23766 indicated that the Rg value for WT RhoG ($1.65 nm), RhoG G12V ($1.65 nm) and RhoG Q61L ($1.60 nm), confirming that there is no significant change in Rg value indicated that all complexes are closely packed as represented in Figure 8(c). Thus, the Rg analysis confirmed that all the protein-drug complexes are compact during the entire simulation run of 200 ns.
The interaction of EHop-016 and NSC23766 with WT RhoG and their mutants RhoG G12V and RhoG Q61L was further analysed using hydrogen bond analysis of 200 ns MD trajectories using 20000 frames. As shown in Figure 7(d), there is no significant change in the average number of intermolecular hydrogen bond formation between EHop-016 and WT RhoG, RhoG G12V, RhoG Q61L Also, similar results were observed for the NSC23766 complex as shown in Figure 8(d), but EHop-016 formed a minimum of three hydrogen bonds while NSC23766 binds to their respective receptor by forming two hydrogen bonds. Furthermore, the intramolecular hydrogen bond formation was analyzed, and results suggest that the average number of intramolecular hydrogen bonds formed per timeframe by WT RhoG, RhoG G12V and RhoG Q61L was 135, 134, 128 respectively, as shown in Figure 7(e), while with NSC23766 complex, the number of intramolecular hydrogen bonds formed by WT RhoG, RhoG G12V and RhoG Q61L was 136, 134, 139 respectively, as shown in Figure 8(e). Next, we analyzed the intermolecular hydrogen bond formed between water solvent and wild-type and mutant receptor complexes with EHop-016 and NSC23766. The results suggest that the EHop-016 complex with WT RhoG, RhoG G12V and RhoG Q61L formed 331, 320 and 331 intermolecular hydrogen bonds, respectively (Figure 7(f)). Similar to EHop-016, NSC23766 complexes show a similar trend of intermolecular hydrogen bonds formed between the receptors and water moiety shown in Figure 8(f). Additionally, the time-dependent RMSD graph was plotted for the Ca-backbone of WT RhoG -complexed with all ten ligands mentioned above in the molecular docking section. As shown in Figure 9, the RMSD for WT RhoG are within the range of $ 0.2 -$ 0.45 nm, and most of them attain stability after 150 ns.

EHop-016 and NSC23766 shows higher binding energy with mutant receptor than WT RhoG
Our molecular docking results showed no significant differences in the binding affinities of the ten ligand with the receptor protein. Previous reports have also shown that the accuracy of free energy prediction for most computational docking approaches, including AutoDock, is roughly 2-3 kcal mol À1 standard error. Therefore, to calculate the binding energy accurately, we performed the gromos_mmpbsa (g_mmpbsa) analysis of all ten ligands complexed with the WT receptor. The MMPBSA analysis suggests that EHop-016 and NSC23766 have higher binding energy than macrolide1, as shown in Table 5. Next, to gain further insight into the total intermolecular interactions between mutant receptor RhoG G12V and RhoG Q61L and ligand (EHop-016 and NSC23766), the binding free energy (DG bind ) of these complexes was calculated, and results suggest that the mutant receptor RhoG Q61L , RhoG G12V complexed with EHop-016 and NSC23766 showed higher binding energy compared to WT RhoG complex (Table 6). This implies that the EHop-016 and NSC23766 strongly bind with the mutant receptor compared to the WT receptor. Moreover, the individual contribution of receptor and ligand Ehop-016 in the total binding energy is more in the mutant complex compared to WT complexes, as shown in Figure 10(a). Further, we analyzed each amino acids individual negative binding energy contribution in WT and mutant  receptors. As shown in Figure 10(b), the amino acids such as Pro34, Phe37, Trp56, Tyr64 were significant negative contributors in binding energy with WT RhoG complexed with EHop-016, while in RhoG G12V mutant, the Trp56, Tyr64, Leu67, Arg68 (Figure 10(c)) and Val7, Phe37, Trp56, Thr58, Tyr64, Leu67 & Leu70 are the major contributor in RhoG Q61L mutants ( Figure 10(d)). Moreover, as shown in Figure 11(a), the individual contribution of receptor and ligands NSC23766 in the total binding energy was more in the mutant complex compared to WT complexes. Next, we analyzed the individual contribution of WT and mutant receptor complex with NSC23766, and we found that Phe37, Asn39, Trp56, Tyr64 and Leu67 are the significant contributor in WT RhoG ( Figure  11(b)), while Asp38, Asn39, Trp56, Tyr64 are the main contributors of the mutant RhoG G12V (Figure 11(c)). Also, as shown in Figure 11(d), the RhoG Q61L amino acid residues such as Val36, Phe37, Trp56, Leu67, Leu70 and Ser71 significantly contribute to the total binding energy. The overall result suggests that the EHop-016 and NSC23766 strongly bind with mutant receptors compared to WT receptors. The PCA tells about the important conformational changes that help in deciding the protein function (Amir et al., 2021). Thus, PCA was performed to investigate the significant conformational shifts in the structures of the proteinligand complex as described in the methods (Singh et al., 2020). The first two PCs were taken into consideration to separate the random protein motions in our simulation interpretation. Figure 12(a,b) represent the projection of significant PCs of WT RhoG, RhoG G12V and RhoG Q61L complexes in 2D space (EHop-016 and NSC23766, respectively). The distribution of lines within the graph correlates with the conformational changes in the three complexes each of EHop-016 and NSC23766, where the innate mobility of WT RhoG is higher than both the mutants. Similarly, the free energy landscape of the WT RhoG, RhoG G12V and RhoG Q61L complexes with EHop-016 and NSC23766 was also extracted from the 2D representation of the PCA plots ( Figure 12(c-h)). The obtained Gibbs free energy range is related to the stability of protein complexes. This range increases from 0 to 15.9 kJ mol À1 for WT RhoG, 0 to 16.6 kJ mol À1 for RhoG Q61L , and 0 to 16.8 kJ mol À1 for RhoG G12V (in EHop-016 complexes), whereas in NSC23766 complexes range increases from 0 to 14.8 kJ mol À1 for WT RhoG, 0 to 15.3 kJ mol À1 for RhoG G12V , and 0 to 16.3 kJ mol À1 for RhoG Q61L . The stable state represented by blue coverage in the plot decreases subsequently from WT RhoG, RhoG G12V , RhoG Q61L . The free energy landscape (FEL) analysis carried out on PCA is a simple demonstration of the protein conformational space concerning energy and time. FEL results infer that the EHop-016-WT RhoG complex has a larger and more positive difference in Gibbs free energy, which relates to its better stability to denaturation.

Discussion
Rho signalling pathways regulate a diverse cellular function in cancer and inflammatory conditions, including actin cytoskeleton rearrangement, cell adhesion, invasion, apoptosis, vesicular trafficking, and transcriptional regulation (Gao et al., 2004). Thus, inhibitors targeting specific elements such as GEFs of the Rho signalling pathway could help to develop therapeutic drugs targeting these molecules. Rho GEFs activate more than one Rho GTPase, which in turn binds with a variety of effectors (Dipankar et al., 2021). Thus, Rho GEFs are ideal targets for modulating the strength of response to a specific upstream signal. RhoG is placed upstream to Rac1 and regulates the signalling events in many pathological conditions. RhoG also shows an overall and GEF binding domain sequence and structural similarity to Rac1 GTPases. In the present study, we have generated the 3D structure of the WT and two experimentally demonstrated RhoG mutants with improved accuracy compared to previously reported models. Additionally, from among various Rac-1 inhibitors,    we found that the EHop-016 and NSC23766 have better binding energy towards the GEF site of WT as well as mutant RhoG G12V and RhoG Q61L proteins compared to macrolide1 and other tested inhibitors.
A previous study demonstrated that point mutation in different conserved sites such as G12V and Q61L in RhoG proteins (RhoG G12V and RhoG Q61L ) resulted in constitutively active mutation and an ON state in RhoG protein (Prieto-  Bustelo, 2003). This study has used multiple in silico tools and servers based on different algorithms to predict the desired mutations effect on pathogenicity and stability. Our results suggest that such mutations could lead to instability in protein structure and its dysfunction. Moreover, to understand the protein's structural changes by constitutively active mutations, we have performed a molecular dynamics simulation study to validate the outcomes of in silico predictions in all the protein variants. The rise of structural perturbations in the Q61L variant of RhoG protein over 200 ns of simulation reflects its conserved location over the G12V variant. Similarly, an increase in the stability and a decrease in the deviation and fluctuation of the Q61L variant was observed in the Rg, RMSD, and RMSF graphs of MDs. These results align with the previous findings by Hubbard et al., where such behaviour in the Q61L variant was attributed to the loss of hydrogen bonding in the protein structure (Hubbard & Haider, 2010). Moreover, the higher RMSD value for WT RhoG and G12V mutant specifies that the overall topology of the structure has changed. The observed RMSD of 0.6 Å is predominantly accredited to huge deviations of residues at the termini and loop. Even though the simulation was for 200 ns, the divergence from the starting structure for the mutant structure after 100 ns was sufficient to indicate that both the mutant structures significantly adopted a denatured conformation. The simulation revealed that the Valine side chain has a more immediate and significant effect in disrupting the structure by affecting the folding property and stability (Futatsugi & Tsuda, 2001;Zhao et al., 2015). Similarly, higher RMSF values indicate greater flexibility during the MD simulation. The RMSF results around the residues Thr25, Ile80, and Pro139 showed a rapid increase within the first 50 ns of the simulation, indicating that the larger isoleucine side chain may have a greater structural impact compared to Threonine and Proline side chain (Zhao et al., 2015).

S anchez &
To date, unlike Ras, there has been no constitutively active RhoG mutation reported in human cancer. However, as predicted by computational studies, constitutively active mutations (e.g., RhoG G12V and RhoG Q61L ) proposed in the experimental models may be linked with cancers and other disease conditions. In other words, it is possible that, like other Rho GTPases, RhoG upregulation or overexpression may be linked to tumorigenic properties (Gao et al., 2004). Conceivably, as shown in our analysis, RhoG signaling can be modulated by targeting the steps that involve its activation by GEFs. Therefore, it is essential to look for inhibitors that could work on both WT and Mutant variants of RhoG specific GEF. In support, Rac1-based inhibitors such as NSC23766 and EHop-016 were shown to inhibit cancer cell metastasis both in vitro and in vivo (Gao et al., 2004;Montalvo-Ortiz et al., 2012).
In order to screen the best inhibitor of WT RhoG, RhoG G12V and RhoG Q61L mutant, we have performed molecular docking of all established and proven Rac1 inhibitors such as migrastatin and its structural analogues such as isomigrastatin, macrolide1, macrolide3, macroketone and macrolactam, including other compounds such as NSC23766, its derivative EHop-016, and isoalvaxanthone (Gao et al., 2004;Montalvo-Ortiz et al., 2012;P erez & Danishefsky, 2007;L. Wang et al., 2010). Our result suggests that, among all docked compounds, macrolide1 is the lead compound with maximum binding energy -8.2, -8.1, -8.1 kcal mol À1 for WT RhoG, RhoG G12V and RhoGQ61L, respectively. Moreover, EHop-016 has a binding affinity of -6.9, -7.3, -7.0 kcal mol À1 for WT RhoG, RhoG G12V and RhoG Q61L , respectively. While NSC23766 showed binding affinity of -6.8, -6.8 & -6.7 kcal mol À1 for WT RhoG, RhoG G12V and RhoG Q61L, respectively. Also, previous reports suggest that the accuracy of free energy prediction by AutoDock is roughly 2-3 kcal mol À1 standard error (Cosconati et al., 2010). Therefore, to calculate accurate binding energy, we have performed the g_mmpbsa analysis of all ten ligands and then ranked them according to their binding energy. As presented in our findings, the re-ranking of these ligands suggests that the EHop-016 and NSC23766 showed higher binding energy than macrolide1 and all other tested inhibitors. This difference in molecular docking and the g_mmpbsa may result because protein flexibility is not considered in the standard molecular docking procedure. In contrast, in the g_mmpbsa method, the flexibility and different conformation of the protein is taken into account to calculate binding energy. Thus, the latter is considered more accurate and reliable than the molecular docking procedure (Alonso et al., 2006;Davis & Baker, 2009;Morris et al., 2009). Importantly, our docking analysis showed that the EHop-016 and NSC23766 bind to the cleft formed by residues Phe 37 , Asp 57 , Ala 59 , Asn 39 , Trp 56 , Thr 58 , Tyr 64 , Arg 68 , Leu 67 , Leu 70 , Ser 71 in both WT and mutant receptors. Moreover, our data also suggest that the primary interactions between EHop-016 and NSC23766 with these receptors are hydrophobic. Our results coincide with the previous finding by Gao et al., which shows NSC23766 is hydrophobic and binds to hydrophobic clefts formed in Rac1 (Gao et al., 2004). Moreover, EHop-016 binds deeper to GEFs grooves of WT and mutant receptor RhoG than NSC23766. Thus, our findings align with previous reports suggesting that EHop-016 binds deeper to Rac1 than NSC23766 (Montalvo-Ortiz et al., 2012).
Furthermore, the binding energy of EHop-16 is complexed with WT and mutant receptors. To understand why EHop-016 shows higher binding energy, we have analyzed the MMPBSA results. The analysis showed that the high binding energy is due to a higher magnitude of van der Waals interaction than other interactions such as electrostatic, polar and non-polar energy. The molecular docking result suggests that amino acid residues such as Asn 39 , Trp 56 , Thr 58 , Tyr 64 , Leu 70 , Ser 71 of WT RhoG, while Thr 35 , Asn 39 , Tyr 40 , Trp56, Asp 57 , Thr 58 , Arg 68 , Ser 71 residues of RhoGG12V and RhoGQ61L residues such as Asn 39 , Trp 56 , Thr 58 , Tyr 64 , Leu 70 , Ser 71 are involved van der Waals interaction with EHop-016. Further, the mmpbsa analysis suggests that the contribution energy of Trp 56 in all complexes is higher and thus plays a vital role in increasing the magnitude of van der Waals interaction. Thus, Trp 56 might be essential residues for the binding of EHop-016 with wild and mutant receptors. Similarly, Trp 56 is also an essential contributor to the NSC23766 complex WT and mutant receptors. Our findings correlated with the previous study, which suggests that Trp 56 of Rac1 is an essential determinant in the binding with the NSC23766 (Gao et al., 2004). Apart from Trp 56 , Phe 37 in WT RhoG, Tyr 64 in RhoG G12V , Phe 37 and Leu 67 in RhoG Q61L are the significant contributors to the total van der Waals interaction energy. So, overall results suggest that the involvement of different residues with the difference in magnitude of contribution energy results in higher van der Waals interaction and hence total binding energy.
Moreover, several lines of evidence from the previous studies suggest that the binding energy of the receptor-ligand complex calculated using mmpbsa method is more precise and sometimes correlates with experimental results (Kumari et al., 2014;Wang et al., 2019). The previous report suggested that the EHop-016 binds with Rac1 and inhibits breast cancer metastasis at a concentration of 1 lM, while NSC23766 showed optimal inhibition efficiency of 50-100 mM (Montalvo-Ortiz et al., 2012;P erez & Danishefsky, 2007). Moreover, ITX3 is the only reported inhibitor of RhoG/Rac1 signalling pathways that inhibit these proteins at the concentration of 100 micromolar (Bouquier et al., 2009). Also, in our study, the binding energy of ITX3 is less than EHop-016 and NSC23766 with WT and mutant receptors. Our binding energy data is similar to these findings that the EHop-016 binds more efficiently than NSC23766 with WT RhoG and its mutant receptor than ITX3. Although the inhibition efficiency varies from cell to cell, from our study, we could speculate that, like Rac1, the EHop-016 and NSC23766 inhibit RhoG by blocking the GEFs interaction with similar inhibition efficiency.
For suitable ligand binding and stable interactions, it is essential to look at the ligands' electronic and molecular properties. Using a frontier orbital analysis, the molecular interactions of the ligands with the protein were assessed, and the findings confirmed that the inhibitors are chemically reactive. The identified inhibitors manifested a narrow HOMO-LUMO gap, implying the molecular reactivity of the inhibitor moieties interacting with the receptors active site amino acid residues. NSC23766 exhibited the lowest energy gap with the HOMO-LUMO gap energy of 0.74 eV, and the highest energy gap was observed in EHop-016 with the gap energy of 1.22 eV. HOMO-LUMO analysis revealed the highest proportion of charge transfer, supporting our hypothesis that it strongly retains the inhibitor molecules within the receptor (Yadav et al., 2020).
Further, the higher HOMO-LUMO energy gap of EHop-016 correlated with the higher stability of the protein-ligand complex, similar to the previous report (Anitha et al., 2020;Singh et al., 2020). Our PCA and FEL analysis also reflected a strong association of EHop-016 with WT protein which was confirmed by better binding energy in docking studies. Possibly, EHop-016 helped in containing the larger amplitude collective motions in WT protein which are often slower than local (fast) motions. This large-scale motion represents stochastic motion behavior rather than a smooth motion behavior of the protein (Amadei et al., 1993). As explained above, EHop-016 attained a stable conformation in the GEF site and pacified the internal atomic motion during the MD simulation. The key interactions like non-covalent bonding between EHop-016 and WT were lost after the induction of mutations. The mutations (G12V and Q61L) impacted the structural and functional nature of WT protein. Despite lower interactions, EHop-016 managed to efficiently bind to the mutants and fairly contained the conformational changes by restricting the atomic fluctuations (Singh et al., 2020).
In summary, using a refined and carefully modelled structure of RhoG and computational analysis, we show that EHop-016 and NSC23766, a proven inhibitor of Rac1, had the dual potency to inhibit the binding with GEF sites in both WT and mutated RhoG and could block the downstream signaling. Secondly, EHop-016 and NSC23766 have higher binding energy compared to ITX3. Thus, our in silico analysis predicts that these compounds inhibit not only WT but also mutant RhoG. The findings of this study will form a fundamental basis for designing anticancer drugs targeting RhoG mediated dysfunction in cell migration. Nevertheless, the quest for better RhoG inhibitors can continue from searching large databases (chemical, natural, marine, etc.) and performing high throughput virtual screening.