Ceylon Journal of Science 54 (4) 2025: 963-977 RESEARCH ARTICLE
Computational Assessment of Guanidine and Arginine Isomers as Inhibitors of Caseinolytic Peptidase B (ClpB): Targeting Bacterial Chaperones for Novel Antimicrobial Strategies R. S. J. Udari, H. M. S. Shamodhi, N. R. M. Nelumdeniya, R. J. K. U. Ranatunga, S. P. N. N. Senadeera and C. B. Ranaweera
Highlights • The inhibitory potential of GuHCl, L-arginine, and D-arginine against bacterial ClpB protein was evaluated using computational approaches. • Report the stable binding supported by H-bonds and affinities from docking and MD simulations. • ADMET analysis shows good oral bioavailability of test ligands with no violations of Lipinski’s Rule of Five. • Low toxicity and acceptable LD₅₀ values supported the test ligand’s potential as safe ClpB inhibitors. • Findings suggested that targeting ClpB can offer promising strategies for antimicrobial therapies, warranting further in vitro and in vivo validation.
Ceylon Journal of Science 54 (4) 2025: 963-977 DOI: https://doi.org/10.4038/cjs.v54i4.8675 RESEARCH ARTICLE
Computational Assessment of Guanidine and Arginine Isomers as Inhibitors of Caseinolytic Peptidase B (ClpB): Targeting Bacterial Chaperones for Novel Antimicrobial Strategies R. S. J. Udari1, H. M. S. Shamodhi1, N. R. M. Nelumdeniya2, R. J. K. U. Ranatunga3, S. P. N. N. Senadeera4 and C. B. Ranaweera1 Department of Medical Laboratory Sciences, Faculty of Allied Health Sciences, General Sir John Kotelawala Defense University, Werahera, 10718, Sri Lanka 2 Department of Pharmacy, Faculty of Allied Health Sciences, General Sir John Kotelawala Defense University, Werahera, 10718, Sri Lanka 3 Department of Chemistry, Faculty of Science, University of Peradeniya, Peradeniya, 20400, Sri Lanka 4 Department of Zoology and Environmental Sciences, Faculty of Science, University of Colombo, Colombo 03, 00300, Sri Lanka 1
Abstract: Bacterial infections pose a significant global health threat due to their resistance to broad-spectrum antimicrobials, necessitating novel therapeutic strategies. Caseinolytic Peptidase B (ClpB), a bacterial chaperone absent in human cells, plays a crucial role in bacterial survival under stress by disaggregating protein aggregates, making it a promising antimicrobial target. This study utilized computational methods to evaluate guanidine hydrochloride (GuHCl), D-arginine, and L-arginine as potential ClpB inhibitors. Molecular docking studies using AutoDock Vina against ClpB (PDB ID: 1QVR) identified possible binding poses of the ligands. Both D-Arginine and L-Arginine showed a binding affinity of -5.9 kcal/mol to Nucleotide Binding Domain-1 (NBD-1), while GuHCl exhibited a binding affinity of -3.5 kcal/ mol to Nucleotide Binding Domain-2 (NBD-2). The formation of favorable conventional hydrogen bonds between the protein and the ligands primarily contributed to the observed binding affinities in the docking results. SwissADME predicted drug-likeness and pharmacokinetics, while ProTox-II assessed toxicity. None of the ligands violated Lipinski’s rule, indicating their suitability for oral administration. In silico toxicity, predictions classified GuHCl as slightly toxic (Class IV; LD50: 350 mg/kg) and L-arginine and D-arginine as possibly harmful (Class V; LD50: 245,050 mg/ kg), yet overall reflecting a favourable safety profile. Molecular dynamics (MD) simulation trajectories for 100 ns revealed stable root-mean-square deviation (RMSD) and consistent hydrogen bond formation, indicating stable ligand binding. However, solvent-accessible surface area (SASA), radius of gyration (Rg), and root-mean-square fluctuation (RMSF) analyses revealed minor alterations in the NBD-1 domain. These structural changes align with experimental findings for GuHCl, suggesting impaired ClpB activity. Binding free energy calculations using MM/ PBSA confirmed favourable interactions, with all test molecules showing negative free energy values. These findings suggest that GuHCl, L-arginine, and D-arginine have potential as ClpB inhibitors, warranting further in vitro and in vivo validation for antimicrobial therapy. Keywords: ClpB; GuHCl; L-Arginine and D-Arginine; Molecular docking; Molecular dynamics
INTRODUCTION Molecular chaperones are a class of proteins that facilitate the stabilization and proper folding of other proteins into their functional native conformations without becoming part of the final structure (Doyle et al., 2012). Newly synthesized polypeptides emerging from ribosomes may occasionally misfold in the cellular environment, though most correctly fold into their active native states. Similarly, some proteins may also misfold and lose their native conformations under stress conditions. Molecular chaperones play a crucial role in reversing protein aggregation by disassembling protein aggregates and converting them into unstructured polypeptides. These released polypeptides can subsequently refold into their active native structures with the assistance of additional molecular chaperones or be targeted for degradation by the cellular protease machinery (Doyle et al., 2012; Ranaweera, 2021). Caseinolytic Peptidase B (ClpB) in bacteria, Hsp101 in plants, and Hsp104 in yeast are molecular chaperones classified under the Hsp100 family due to their monomeric molecular weight of approximately 100 kDa (kiloDaltons) (Doyle et al., 2012; Ranaweera, 2021). The physiologically active form of ClpB exists as a hexamer, with an approximate molecular weight of 575 kDa, comprising six ClpB monomers, each around 95 kDa. These monomers assemble in the presence of nucleotides such as Adenosine triphosphate (ATP) or Adenosine diphosphate (ADP). During the assembly, a narrow central channel is formed within the hexamer, facilitating the movement of extracted polypeptides from aggregates as ClpB hydrolyzes ATP (Zolkiewski et al., 2012). The Escherichia coli ClpB monomer has 857 amino acids in it. The first 146 amino acids (AA) comprise the N-terminal domain, which is joined to the Nucleotide Binding Domain-1 (NBD-1) by the conserved linker starting from 147 AA to 163 AA. NBD-1 starts at 164 AA *Corresponding Author’s Email: cbr2704@kdu.ac.lk https://orcid.org/0000-0003-4173-0547
This article is published under the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
964 and ends at 413 AA by joining to the middle (M) domain (414 AA to 525 AA long), which resides between NBD-1 and Nucleotide Binding Domain-2 (NBD-2). NBD-2 starts from 526 AA and ends at 769 AA by connecting to the last 87 amino acids, which make up the C-terminal (Figure 1) (Ranaweera, 2021). Antimicrobial resistance arises when bacteria, viruses, fungi, and parasites undergo genetic changes that render them unresponsive to antimicrobial agents. This phenomenon makes infections more challenging to treat, increases the risk of disease transmission, leads to more severe illnesses, and can result in fatalities (Antimicrobial Resistance, 2021; Capozzi et al., 2019). The rapid escalation of antimicrobial resistance, coupled with the surge in infectious bacteria, emphasizes the urgent need to develop new antimicrobials and identify novel antimicrobial targets. In a laboratory setting, a target is considered an antimicrobial target if it meets the following criteria: it must be essential for the microorganism’s survival, be widely distributed among the target organisms of concern, ideally have no homologs in humans or other eukaryotes, and be druggable, meaning it can interact with drugs in vivo and be inhibited by small molecules or biotherapeutics. (Alksne & Dunman, 2008). Interestingly, metazoan proteomes lack Hsp100 chaperones, which are exclusively found in bacteria, protozoa, fungi, and plants (Ranaweera et al., 2018). ClpB, has no orthologs in higher eukaryotes, including mammals and humans. Consequently, ClpB represents an attractive target for the development of novel antibiotics, as its inhibition could selectively disrupt microbial function without adversely affecting mammalian cells (Glaza et al., 2021; Kędzierska-Mieszkowska & Zolkiewski, 2021). Organisms that possess ClpB include clinically significant pathogens such as Staphylococcus aureus, Klebsiella pneumoniae, Pseudomonas aeruginosa, Enterococcus faecium, Acinetobacter baumannii, and Enterobacter spp. These are collectively known as the ESKAPE pathogens.
Ceylon Journal of Science 54 (4) 2025: 963-977 In these bacteria, ClpB is essential for survival, replication, and pathogenicity, highlighting its critical role in their stress response and virulence mechanisms (Udari et al., 2025). ClpB is also found in other obligate intracellular bacteria and protozoan pathogens, such as Anaplasma phagocytophilum and the malaria-causing Plasmodium falciparum (Ranaweera et al., 2024). In vivo and in vitro data demonstrate that ClpB is druggable by Guanidine hydrochloride (GuHCl) (Ranaweera, 2021). Studies on the ClpB 4HSE structure, including a crystallized Gdm+ ion with ADP, reveal that GuHCl can bind to and inhibit Hsp104/ClpB through two mechanisms. First, GuHCl disrupts the interaction between DnaK cochaperone and the M-domain, a crucial component for ClpB activity, thus stabilizing Hsp104/ClpB in a suppressed conformation by reinforcing the M-domain/NBD-1 link. Second, GuHCl inhibits the continuous ATP turnover by NBD-1 (Kummer et al., 2013). In yeast, GuHCl cures prions by inhibiting Hsp100 chaperones. This inhibition occurs because the Guanidinium ion (Gdm+) of GuHCl interacts with the bound nucleotide and conserved GLU, affecting nucleotide binding affinities and interfering with essential GLU required for ATP hydrolysis, by specifically binding to the N-terminal nucleotide-binding regions (Zeymer et al., 2013). For many years, drug discovery efforts have relied heavily on computer-aided drug design (CADD), significantly reducing the time and cost required to plan, execute, and test experiments in the laboratory (Garg et al., 2020). By utilizing in silico methods, such as molecular docking and molecular dynamics, researchers can predict the binding affinities and stabilities of potential drug candidates with target proteins. Additionally, in silico absorption, distribution, metabolism, excretion, and toxicity (ADMET) analysis helps identify the most promising compounds for further studies (Tabani, 2023).
Figure 1: The crystal structure of ClpB (PDB ID:1QVR). The N-terminal domain is shown in light blue, the linker in magenta, the NBD-1 in blue, the M-domain in red, the NBD-2 in yellow, and the C-terminal domain in light green. The figure was generated using structural data from the Protein Data Bank (PDB ID: 1QVR; Lee et al., 2003) and adapted with reference to Ranaweera (2021).
965
Udari et al. This study aimed to evaluate inhibitor molecules targeting bacterial ClpB using an in-silico approach. GuHCl was selected based on previous in vivo and in vitro findings demonstrating its ability to bind and inhibit ClpB, as discussed previously. Due to the structural similarity of the guanidine group in L-arginine and D-arginine to the guanidinium ion (Gdm⁺) in GuHCl, both enantiomers were also considered as candidate ligands. The selected molecules were assessed for their binding potential and pharmacokinetic properties. Computational analyses, including molecular docking and molecular dynamics simulations, were conducted to gain insights into the strength and stability of protein-ligand interactions. MATERIALS AND METHODS Preparation of the protein structure The X-ray diffractive crystallographic structure of ClpB (PDB ID: 1QVR; resolution of 3.00 Å), derived from Thermus thermophilus (Lee et al., 2003), available in the RCSB Protein Data Bank (PDB) (http://rscb.org) was used in this study. Prior to docking, the target protein was cleaned and refined by removing trapped water molecules and cocrystalized ligands. Then, they were screened for missing amino acids. The FASTA sequence (amino acid sequence) of the structure reveals that each monomer has a sequence length of 854 amino acid residues. Still, the PDB format structure consists of 803 amino acid residues, leaving 51 amino acids missing in each monomer. Modeller 10.4 tool was used to fill the missing amino acid residues of the protein structure (Šali & Blundell, 1993). After fixing the missing amino acid residues, the structure was refined using the ModRefiner algorithm (Xu & Zhang, 2011) available at https://zhanggroup.org/ModRefiner/. This algorithm minimizes the atomic-level energy of the protein through a two-step process. First, it utilizes c-alpha traces and minimizes the main chain energy in a lowresolution step. Subsequently, it minimizes the energy of the entire atomic structure by incorporating initial side chain atoms at a high resolution and higher speed. At
the end of the process, the algorithm calculates the Root Mean Square Deviation (RMSD) and Template Modelling score (TM-score) for the refined structure to compare and validate the resulting refined model. A Ramachandran plot (Ramachandran & Sasisekharan, 1968) was generated for the refined complete monomer of 1QVR to evaluate the protein backbone conformation using the BIOVIA Discovery Studio v21.1.0.20298 package (Systèmes, 2020). For structural validation, the native and modeled monomeric structures of 1QVR were superimposed using the Visualization Molecular Dynamics (VMD) software package, allowing a comparative analysis of structural deviations. Retrieval of ligand structures PubChem (https://pubchem.ncbi.nlm.nih.gov/) database was used to obtain the 3D structures of the ligands in a structure data file (.sdf) format. Gdm+ ion of GuHCl, L-arginine, and D-arginine were used as test molecules. ADP, known to co-crystallize with ClpB in the 4HSE, PDB structure, was used as a reference ligand (Zeymer et al., 2013). The canonical simplified molecular-input line-entry system (SMILES) of the testing ligands was used for the insilico ADMET and toxicity analysis. Molecular Docking AutoDockTools 1.5.7 (Goodsell & Olson, 1990) was used to generate Protein Data Bank, Partial Charge (Q), & Atom Type (T) / PDBQT format files of the ClpB PDB structure while introducing hydrogens and partial atomic charges (Kollman charges). Energy minimization of the ligands was carried out using the option available on PyRx 0.8 (Dallakyan & Olson, 2015) using the MMFF94 force field. After the energy minimization of each ligand, they were saved separately in pdb, pdbqt, and mol2 formats for docking and MD studies. Docking was continued using the PyRx open-source platform using AutoDock Vina (Trott & Olson, 2010). The grid box was defined to cover the whole protein. X× Y × Z sizes to 111.6939 Å × 124.4924 Å × 74.0728 Å, and center grid coordinates X, Y, Z were set to 45.8962, 23.4164, 33.1118, respectively.
Figure 2: Ligand structures investigated in this study. (a) D-arginine, (b) L-arginine, (c) Adenosine diphosphate (ADP), and (d) Guanidine Hydrochloride (GuHCl) 966 Ceylon Journal of Science 54 (4) 2025: 963-977
The binding poses produced by the docking process were further analyzed, and their docking scores and interactions with the key amino acids of protein were observed using BIOVIA Discovery Studio v21.1.0.20298 package (Systèmes, 2020).
Gcomplex is the energy of the protein-ligand complex, whereas Gprotein and Gligand represent the energy of the protein and ligand in an aqueous solvent, respectively. The free energy is calculated using the following equation for each of the terms in the equation.
Physicochemical, ADME, analysis, and drug-likeness and toxicity prediction Gx=Ebonded +(Evdw +Eelec)+Gpolar + Gnon-polar
In parallel to the docking study, the SwissADME online web server (Daina et al., 2017) was used for the screening and predicting the physicochemical, pharmacokinetic, and drug-likeness of ligands, while ProTox-3.0 webserver (Banerjee et al., 2018) was used to analyze the toxicity. Molecular dynamics simulation studies After the analysis of docking results, MD simulations were conducted for both free protein and all ligand-protein complexes using the GROMACS-2024 software package (Berendsen et al., 1995) with the CHARMM36 force field (Huang & MacKerell, 2013) to find out the ligand impact on the stability and structural changes of the ClpB protein. The topology file of the clean protein was generated using inbuilt GROMACS commands using CHARMM36 allatom forcefield, while the ligand topology files were created by SwissParam (http://www.swissparam.ch/) online server (Zoete et al., 2011). Subsequently, protein-ligand complex files were made by merging the protein and ligand files. The complex was solvated in a truncated octahedron periodic box (111.01 x 124.01 x 74.91 Å), using a distance of 12 Å between the protein and the boundary of each side using the TIP3P water model, keeping the complex in the middle. Then, the system was neutralized by adding Na+ ions, and an energy minimization step was carried out. The system was equilibrated using NVT and NPT conditions at 300 K and 1 bar pressure, respectively, for 100 ps. Finally, the production run was done for 100 ns, keeping the time step at 2 fs (Bhattarai & Emerson, 2021). In the end, the trajectory files of the free protein and protein-ligand complexes were analyzed by calculating RMSD, Root Mean Square Fluctuation (RMSF), H-bonds, Radius of Gyration (RG), and Solvent-accessible surface area (SASA) with GROMACS inbuilt tools gmx rms, gmx rmsf, gmx hbond, gmx gyrate, and gmx sasa, respectively (Kushwaha et al., 2021). VMD software package (Humphrey et al., 1996) was used to analyze the trajectories, and the plots were produced using GraphPad Prism 8.4.2 software. Total Binding free energy calculations For the calculation of total binding free energy (ΔGbind) of the protein-ligand complexes, the g_mmpbsa package was used with GROMACS (Valdés-Tresanco et al., 2021). The g_mmpbsa uses the Molecular Mechanics/ PoissonBoltzmann Surface Area (MM/PBSA) approach for the ΔGbind calculation of the complexes (Kushwaha et al., 2021). The ΔGbind was calculated for all the protein-ligand complexes using the entire MD simulation trajectories of the 100 ns period. The following equation provides the protein-ligand complex’s ΔGbind in an aqueous solvent. ΔGbind = Gcomplex -(Gprotein +Gligand)
In this equation, Gx can be Gcomplex , Gprotein or Gligand, and Ebonded represents the energy given by bonded contacts within the molecule, which is considered as zero always (Kollman et al., 2000). Evdw represents van der Waals energy, and Eelec represents electrostatic energy. Gpolar, is the energy associated with the electrostatic interactions between the molecule (protein or ligand) and the surrounding solvent. In Gnon-polar represents nonpolar or apolar solvation energy, which is the free energy contribution due to non-electrostatic interactions, such as hydrophobic effects. Gnon-polar is derived as follows, Gnon-polar = γ⋅SASA + b In this equation, SASA stands for solvent-accessible surface area, whereas b is the fitting parameter and γ is a surface tension coefficient. Therefore, the g_mmpbsa tool determines the total binding free energy by computing each energy component of this equation for the protein, ligand and the total complex in an aqueous solvent environment (Sundar et al., 2019). RESULTS AND DISCUSSION This study investigated the ClpB protein using an in silico approach, focusing on its interactions with the guanidinium ion (Gdm⁺) of GuHCl, as well as with L-arginine and D-arginine, using ADP as a control. The research was conducted in three major steps: molecular docking, ADMET analysis, and molecular dynamics (MD) simulation. Molecular docking served as a virtual screening step to identify ligand binding sites and their affinities. ADMET analysis evaluated the pharmacokinetics and toxicity profiles of the test ligands. Finally, the stability of the free protein and the protein-ligand complexes was assessed through MD simulations under predefined conditions. Refining the protein After filling the missing amino acid residues in the monomeric 1QVR structure using Modeller tools, the resulting model required refinement to eliminate artifacts such as steric clashes caused by unnatural overlaps between non-bonded atoms. This refinement can be achieved through short molecular dynamics (MD) simulations, which allow the structure to relax into a more energetically favorable conformation (Adiyaman & McGuffin, 2019). Alternatively, refinement can be performed using validated computational tools such as ModRefiner, which is a fast and efficient algorithm for protein structure optimization. ModRefiner operates in two stages: the first reconstructs a realistic backbone using predefined Cα trace conformations, and the second enhances structural accuracy by adding side-chain atoms and minimizing structural energy using a composite of physics-based and knowledge-based force fields. The quality of the final refined model is assessed
Udari et al. by metrics such as lower RMSD and higher TM-score, which indicate improved folding accuracy and structural realism (Xu & Zhang, 2011). Figure 3 shows the refined structure of 1QVR. RMSD value was within 0-2 Å, indicating reasonably higher structural similarity to the initial modelled structure. The TM-score being closer to 1 indicates higher similarity and accuracy of the folding patterns.
Figure 3: Refined 3D structure of the ClpB monomer (PDB ID: 1QVR) after filling missing amino acid residues using Modeller and optimizing the geometry with ModRefiner. Following optimization, the structure achieved a root mean square deviation (RMSD) of 1.446 Å and a template modeling score (TM-score) of 0.9881
Ramachandran plot The refined model of the 1QVR monomeric structure was validated using a Ramachandran plot and compared with the original 1QVR structure (Figure 4). Ramachandran plot helps us understand the conformational space available to amino acids and shows their theoretically favoured regions based on the backbone dihedral angles (φ and ψ). The rotational angles between the Cα-N and Cα-C bonds of the protein backbone are denoted by φ (phi) and ψ (psi), respectively (Ramachandran & Sasisekharan, 1968). ψ is usually plotted on the y-axis and φ on the x-axis. In both pots triangles represent glycine, squares represent proline, and circles represent all the other amino acids. Three main regions can be identified in these plots. The areas that are margined with light blue are the areas that represent alpha-helical and beta-sheet conformations, which are the permitted conformations where steric conflicts do not occur. The areas within the pink margin are the partially allowed zones representing left-handed alpha-helical conformations, which are uncommon (Ramachandran & Sasisekharan, 1968). Amino acids in these regions are shown in green, with most residues in the refined structure (Figure 4-a) located within sterically favorable zones. The regions not margined with a coloured line indicate sterically forbidden conformations where the atoms in the polypeptide are closer together than the total of their van der Waals radii, except for glycine (red triangles), which lacks a side chain (Ho & Brasseur, 2005). Non-glycine residues in these forbidden regions (red circles) are minimal in the refined structure. Therefore, overall, the Ramachandran plots indicate that the refined 1QVR monomeric structure folds with minimal steric clashes and adopts energetically favorable backbone conformations.
Figure 4: Ramachandran plots of (a) the refined 1QVR monomeric structure and (b) the original 1QVR structure. φ (phi) angles are on the x-axis, and ψ (psi) angles are on the y-axis. Triangles, squares, and circles represent glycine, proline, and other amino acids, respectively. Green symbols denote residues in allowed regions. Red symbols represent residues in sterically disallowed regions, with glycine (red triangles) permitted in these regions due to its unique conformational flexibility. The refined structure (a) demonstrates most residues in favorable conformations, with minimal steric clashes
968 Molecular Docking Gdm+ ion of GuHCl, L-arginine, and D-arginine were taken as test ligands, while ADP was taken as the control ligand for the molecular docking. All ligands were docked against the 1QVR structure using a blind docking approach. As depicted in Figure 5, the crystal structure 4HSE, which includes NBD-1 to M region (residues 141 to 534) with a crystalized Gdm+ ion and ADP. In this structure, Gdm+ ion make hydrogen bonds with GLU 209, ASP 170, and PRO 171 while ADP bonds with THR 205, GLY 201, ALA 206, GLY 203, and ILE 73, respectively (Zeymer et al., 2013). Blind docking was intentionally employed to assess the reliability and accuracy of the docking protocol by evaluating whether ADP would still bind to the same functionally relevant residues observed in the 4HSE structure. Moreover, this approach ensured an unbiased evaluation of the test ligands, allowing them to explore and bind to the most energetically favorable sites in the 1QVR structure without predefined constraints. In docking results with 1QVR (Figure 6), ADP binds to the NBD-1 forming conventional hydrogen bonds with GLN 8, ARG 11, GLU 12, GLU 144, ILE 215, VAL 216, LYS 225, GLY 226, LYS 227. In contrast, the Gdm+ ion binds to the NBD-2 by forming conventional hydrogen bonds with SER 539, LYS 540, GLU 543, and GLU 545 and attractive charge interactions with GLU 545, ASP 614. As mentioned earlier, this Gdm+ ion interacts with negatively charged amino acid GLU, the same as in the 4HSE crystal structure, even though it shows high affinity towards NBD-2 in docking results. Both positively charged guanidine groups in the L-arginine and D-arginine interact with negatively charged GLU in the 1QVR structure NBD-1, as shown
Ceylon Journal of Science 54 (4) 2025: 963-977 in Figure 6 as well. Table 1 shows the binding affinities between all the ligands and the protein and their binding interactions with adjacent residues of the most favourable pose. ADMET results The SwissADME server provides results of analyzed chemicals in three main categories: physicochemical, pharmacokinetic, and drug-likeness. Lipinski’s rule of five (RO5) was considered when analyzing the drug-likeness of a ligand for oral bioavailability. For the compound to be orally bioavailable, the ligand should not have more than one violation of the RO5. To predict that the absorption and penetration is ideal and to be accepted as an orally active ligand, the RO5 is as follows: no more than five H-bond donors, no more than ten H-bond acceptors, a molecular weight (MWT) of less than 500 Daltons and a computed Log P (ClogP) of less than five (Lipinski et al., 2001). GuHCl, L-arginine, and D-arginine showed zero violations thereby considered orally bioavailable. When assessing pharmacokinetic properties, Blood Brain Barrier (BBB), gastrointestinal (GI) absorption, skin permeation coefficient (Log Kp), and substrate for cytochromes (CYPs) isoforms were taken into account. All three ligands showed favourable properties while also having high GI absorption capacities. The toxicity was evaluated using the ProTox-II server. GuHCl was predicted to be slightly toxic (Class IV) with an LD50 of 350 mg/kg, while L arginine and D-arginine were categorized as may be harmful (Class V) with an LD50 of 245050 mg/kg. Even though GuHCl has a lower LD50, it showed inactive results as same as both L and D arginine for hepatotoxicity, cytotoxicity, heat shock factor response
Figure 5: (a)- Crystalized ADP in 4HSE structure, bonding with THR 205, GLY 201, ALA 206, GLY 203, and ILE 73. (b)- Crystalized Gdm+ ion in 4HSE structure, making hydrogen bonds with GLU 209, ASP 170, and PRO 171. Conventional hydrogen bonds are depicted with green dashed lines
969 Udari et al.
Figure 6: The central 3D diagram shows binding sites of Gdm⁺, L-arginine, and D-arginine docked to the ClpB monomer (1QVR). Individual panels display ligand-specific 3D interactions with hydrogen bond donors in magenta and acceptors in green. Corresponding 2D diagrams illustrate ligand–residue interactions, with hydrogen bonds shown as green dashed lines Table 1: Binding affinity and interacting amino acids of the control and test ligands with the 1QVR structure Ligand
Binding Interacting Amino Acids Affinity kcal/ mol Control ADP -8.2
Conventional Hydrogen Bonds – GLN 8, ARG 11, GLU 12, GLU 144, ILE 215, VAL 216, LYS 225, GLY 226, LYS 227 Attractive Charge – GLU 144 Test Ligands Gdm+ ion -3.5 L-arginine -5.9 D-arginine -5.9
Conventional Hydrogen Bonds – SER 539, LYS 540, GLU 543, GLU 545 Attractive Charge – GLU 545, ASP 614 Unfavorable Donor-Donor – ARG 370 Conventional Hydrogen Bonds – GLN 8, ARG 11, GLU 144, LYS 227 Attractive Charge – GLU 12, ASP 159 Conventional Hydrogen Bonds – GLU 12, LEU 38, LYS 40, ASN 151, GLU 154, ARG 228 Unfavorable Positive-Positive/ Unfavorable Donor-Donor – LYS 40, ARG 162
element (HSE), effect on mitochondrial membrane potential (MMP), phosphoprotein tumor Suppressor (p53) activity, and ATPase family AAA domain-containing protein 5 (ATAD5). Therefore, from overall ADMET results, the test ligands have the potential to be modified as drug candidates accordingly.
MD results After completing the docking study, MD simulations were carried out for the free protein and the proteinligand complexes, namely ADP, GuHCl, L-arginine and D-arginine. As mentioned previously, RMSD, RMSF, Hydrogen Bonds, Rg, and SASA calculations were used
970 Ceylon Journal of Science 54 (4) 2025: 963-977 GuHCl L-arginine D-arginine High High High to analyze the dynamic behaviour of the free protein and to check the stability of the protein-ligand complexes throughout the 100 ns period. Root Mean Square Deviation (RMSD) measures the average displacement of atoms in the system over a specified period compared to a reference structure. RMSD is an indicator of the stability of a complex; higher fluctuations suggest lower stability. Initially, RMSD may gradually increase during the simulation, but it typically stabilizes once the complex reaches conformational equilibrium. Root Mean Square Fluctuation (RMSF) evaluates the structural changes in a protein by assessing the variability of specific residues during the simulation. Lower RMSF values indicate more stable residues, providing insights into how the ligand impacts the protein’s structure. This aids in understanding the ligand’s effect on the protein’s overall stability (Opo et al., 2021). According to Figure 7, the free protein displayed stable RMSD starting from around 45 ns to 100 ns timeline averaging an RMSD value of 0.5185 nm where the fluctuations are within 1 Å. Variations in the RMSF were observed at a few positions of the free protein, and they were distributed among the NBD-1, M-domain and NBD2. Significant changes were observed between residues numbers 270-290 (NBD-1), 415-500 (M-domain) and 620-650 (NBD-2), giving a similar RMSF value ranging between 0.75 nm to 0.8 nm. According to the plots in Figure 8, the protein-ADP complex displayed stable RMSD averaging at 0.5826
Table 2: Pharmacokinetic properties (ADME) and Lipinski’s rule of five properties of the test ligands. 0 0 0
nm. Over time, the complex portrayed no significant fluctuations, indicating stability. When considering the RMSF of the protein-ADP complex, no major oscillations were displayed compared to the free protein. The RMSD of GuHCl (Figure 9) displayed stability until 60 ns. From 60-90 ns, severe fluctuations were seen. It was also noted that no significant fluctuation was seen for the last 10 ns of the production run, suggesting that the proteinGuHCl complex might stabilize if the run was continued over 100 ns. Overall, the Protein-GuHCl complex averaged at an RMSD of 0.7355 nm. The RMSD of the protein-L-arginine complex stabilized after 45 ns, with an average RMSD of 0.7771 nm, fluctuating with a difference of around 0.2 nm. Similarly, the D-arginine complex stabilized after 30 ns and remained stable until the end of the simulation, with an average RMSD of 0.6851 nm, fluctuating with a difference of around 0.3 nm. Both complexes demonstrated minimal fluctuations and maintained stability throughout the simulation (Figure 9). As already known, arginine structure consists of a guanidine group. Even though that fact might suggest that arginine and Gdm+ ion should have the same effect on protein stability due to their structural similarity, the RMSD results suggest that both L and D arginine bindings are more stable than the Gdm+ ion binding to the protein. This could be due to other functional groups, such as the carboxyl group in the arginine structure.
Figure 7: RMSD and RMSF profiles of the free ClpB monomer obtained from the 100 ns MD simulation. RMSD stabilized around 45 ns to 100 ns with an average value of 0.5185 nm. RMSF analysis revealed notable flexibility at residues 270–290 (NBD-1), 415–500 (M-domain), and 620–650 (NBD-2), each showing values between 0.75 and 0.8 nm
Udari et al. 971
Figure 8: RMSD and RMSF profiles of the ClpB–ADP complex obtained from the 100 ns MD simulation. The complex maintained an average RMSD of 0.5826 nm and no significant fluctuations over time. RMSF analysis showed no major oscillations compared to the free protein The RMSF of the protein GuHCl and L-arginine-complexes shows similar deviations in the residues across the overall protein structure, and more prominent fluctuations can be observed towards NBD-1 and M-domain. In contrast, the D-arginine complex exhibits no significant changes in the RMSF plot compared to the free protein (Figure 9). Notably, the Gdm⁺ ion binds and remains in the NBD-2 domain, while both L-arginine and D-arginine remain in the NBD-1 domain.
According to the study conducted by Cathleen Zeymer and colleagues using the 4HSE ClpB structure, they have suggested that Gdm+ binding can affect ATP hydrolysis due to binding to the conserved GLU residue present in the NBD-1. This interaction affects ATP binding to NBD1, inducing a conformational change in the domain that leads to inhibitory effects. They have also suggested that this inhibitory change only occurs in NBD-1 and not in NBD-2 in the presence of GuHCl (Zeymer et al., 2013).
972 Ceylon Journal of Science 54 (4) 2025: 963-977
Figure 9: RMSD and RMSF profiles of ClpB in complex with GuHCl, L-arginine, and D-arginine from 100 ns MD simulations. The GuHCl complex was stable until ~60 ns, fluctuated between 60–90 ns, and regained stability in the final 10 ns (average RMSD: 0.7355 nm). L-arginine and D-arginine complexes stabilized approximately after 45 ns and 30 ns, respectively, and remained stable (average RMSDs: 0.7771 nm and 0.6851 nm). RMSF deviations were most pronounced in NBD-1 and the M-domain for GuHCl and L-arginine, while D-arginine showed minimal changes A study conducted by Eva Kummer and colleagues also suggests that Gdm+ ion binds to the NBD-1 and interferes with ClpB function by reducing ATP turnover from NBD1 via making conformational changes there. They also suggest that Gdm+ ion can cause conformational changes in the middle domain, thereby interfering with the DnaK activation of ClpB (Kummer et al., 2013). Thus, this information can be used in our study as a justification for the RMSF obtained for Gdm+ ion, which displayed significant oscillations between residues 160-460 (NBD-1 and middle
domain), even though in contrast to their findings, Gdm+ ion remained in the NBD-2 during whole 100 ns of MD simulation, which was the best binding site suggested by our docking results. Therefore, we suggest that although Gdm+ ion binds to the NBD-2 with more affinity, it may cause conformational changes in NBD-1 and middle domain. Hence, this binding can inhibit ClpB by interfering with ATP hydrolysis in the NBD-1 and disrupting the connection between the middle domain and DnaK, which is essential for activating ClpB, as the experimental data
Udari et al. suggested. Moreover, L-arginine portrays guanidinelike effects because the RMSF results showed similar fluctuations to those of the Gdm+ ion. We can also suggest that it might be a good candidate molecule to target ClpB as the RMSD curve displayed more favourable stability in the binding than the Gdm+ ion. H-bond analysis, Rg, and SASA were also conducted and compared with the free protein to understand more about the interactions of the test ligands. Hydrogen bonds are formed in the vicinity of an electronegative atom, and hydrogen atoms are covalently bound to another electronegative atom. The stability of a protein-ligand complex depends on the number of intermolecular hydrogen bonds (da Fonseca et al., 2023). Hydrogen bond formation within the protein (intraprotein hydrogen bonds) in the presence and
973 absence of the primary ligands and between the ligand and the protein was assessed to evaluate the protein stability and protein-ligand complex stability throughout the 100 ns period of the simulation (Figure 10). After the ligand binding, intraprotein hydrogen bond formation has been favoured with all three test ligands and especially with GuHCl in the first half of the simulation. The region of a protein surface that comes into contact with its surrounding solvent molecules is known as the SASA (Mazola et al., 2015). According to Figure 11, the computed SASA of the entire protein illustrated increasing trends for GuHCl, L-arginine and D-arginine complexes. Furthermore, the calculation of the SASA of the entire protein for GuHCl displayed a relatively similar but higher SASA to that of the free protein. Steeper peeks in SASA of
Figure 10: Protein–ligand hydrogen bond formation of the ClpB monomer during 100 ns MD simulations in the presence of primary ligands (GuHCl, L-arginine, and D-arginine), and intraprotein hydrogen bond formation in the free protein. Ligand binding increased intraprotein hydrogen bond formation for all three ligands, with GuHCl showing the greatest effect
974 Ceylon Journal of Science 54 (4) 2025: 963-977 the entire protein within the first 20 ns were indicated for L-arginine and D-arginine, demonstrating the hydrophobic core of ClpB becoming more accessible to the surrounding aqueous environment. Overall, these observations point toward the fact that there might be a possibility of conformational changes, which leads to an increase in the SASA of the protein with all the primary ligands.
the gap between the terminal of each protein atom and its centre of mass during a certain period. The fewer fluctuations there are, the better the stability of the folded protein structure (Choudhary et al., 2023). Figure 11 represents variations in Rg value over the simulation period and indicates that all the protein-ligand complexes were more extended than the free protein except for GuHCl.
The Rg showed the protein structure’s overall compactness throughout the simulation. It is determined by measuring
Considering the results of both SASA and Rg, decreasing protein compactness (higher Rg) and increased SASA were
Figure 11: SASA profiles of the ClpB monomer during 100 ns molecular dynamics simulations in the presence of primary ligands (GuHCl, L-arginine, and D-arginine) and in the free protein. The protein–ligand complexes showed increasing SASA trends, with GuHCl exhibiting a consistently higher SASA compared to the free protein. Early pronounced peaks in SASA for L-arginine and D-arginine occurred within the first 20 ns
975
Udari et al. observed. In GuHCl, it is noticeable that Rg has decreased even though the SASA remains high compared to the free protein. This could suggest a conformational change causing lower Rg, which may lead to concealed residues of the protein being more accessible, resulting in higher SASA. Total Binding free energy calculations The 100 ns trajectory obtained during MD simulations was used for computing the total binding free energy of the ligand in the complex. The g_mmpbsa tool, which was used with GROMACS, implies that the MM/PBSA method was used to compute the binding energy. For each complex of the ClpB protein with the control ligand ADP and test ligands GuHCl, L-arginine and D-arginine, Evdw, Eelec, Gpolar, Gnonpolar, and △Gbind were calculated using the g_mmpbsa module and are shown in Table 3. Gpolar contributes positively to the total binding free energy, while Evdw, Eelec, Gpolar, Gpolar contribute negatively to the calculation. According to the results depicted in Table 3 and Figure 12, the total binding free energy of all four compounds is below zero, which indicates a good affinity towards ClpB, and it shows that the three test molecules show more affinity towards the protein structure than the control ligand ADP itself.
CONCLUSIONS ClpB, a bacterial molecular chaperone critical for protein homeostasis under stress, represents a compelling target for novel antimicrobial development. This study computationally evaluated GuHCl, L-arginine, and D-arginine as potential inhibitors of ClpB. Docking analyses identified favourable binding sites, and ADMET predictions suggested oral bioavailability, high GI absorption, and safe toxicity profiles. MD simulations studies revealed ligand-induced conformational changes in the NBD-1 and M-domain of ClpB. These changes may impair ATP turnover and disrupt co-chaperone DnaK interactions necessary for aggregate reactivation. Hydrogen bond analysis indicated that ligand binding, favored intraprotein hydrogen bond formation supporting enhanced protein stability and stable complex formation. Post MD binding free energy calculations further confirmed all three test ligands’ stability and favourable interaction with ClpB, exhibiting higher affinity than the control ligand ADP itself. These findings suggest that GuHCl, L-arginine, and D-arginine could serve as potential ClpB inhibitors, offering a novel approach to antimicrobial therapy. The observed conformational changes, particularly in NBD-1 and M-domain, align with inhibitory mechanisms suggested by prior experimental studies, though discrepancies in
Table 3: Binding energy calculations of ADP, GuHCl, L-arginine and D-arginine Ligand Name Evdw (kcal/mol) Eelec (kcal/mol) Gpolar (kcal/mol) Gnon-polar (γSASA + b) (kcal/mol) ADP GuHCl L-arginine D-arginine
-29.95 -3.52 -14.88 -13.12 -49.54 -293.97 -245.04 -279.22 65.73 278.14 237.90 256.38 -3.44 -1.30 -2.36 -2.56 Figure 12: The binding energy of ADP, GuHCl, L-arginine and D-arginine △Gbind (kcal/mol) -17.20 -20.65 -24.39 -38.52
976 binding site preferences warrant further investigation. Future in vitro and in vivo studies are essential to validate these findings and explore their therapeutic potential. DECLARATION OF CONFLICT OF INTEREST The authors declare no conflicts of interest