2.1. Construction of Library

The compounds were drawn by using the ChemDraw workspace [34]. A library of fifteen ILs was created by replacing the anion moiety from the 1,3-benzodioxole compound (Figure 1). The anions were selected from established literature based on pharmacokinetic properties. Energy minimisation of ligands was performed using MMFF94 force field in Chem3D [35, 36]. The maximum number of iterations was set to 10,000, and the minimum RMS gradient was set to 0.100.

2.2. Modelling and Energy Minimisation of Targeted Protein

The FASTA sequence of targeted protein tubulin was obtained from RCSB Protein Data Bank followed by submission to the SWISS MODEL server (https://swissmodel.expasy.org) [37, 38]. The protein model was prepared from the template having a hundred percent identity. MolProbity score, QMEANDisCo and local quality estimate plot did the structural assessment of the prepared model [39–41]. The unwanted ligands were removed from the receptor protein, and then, energy minimisation of protein was performed with the GROMOS96 implementation of Swiss-PdbViewer [42, 43].

2.3. Molecular Docking Analysis

The optimised ligands, encompassing both the anionic and cationic constituents as a complex, were subsequently subjected to blind docking analyses with tubulin protein utilising the HEX server [44, 45]. The parameters used for docking control were correlation type—shape only; FFT Mode—3D; sampling method—range analysis; grid dimension—0.6; solutions—2000; receptor range—180, step size—7.5; ligand range—180, step size—7.5; twist range—360, step size—5.5; distance range—40, box size—10; translation step—0.8, substeps—0; score threshold—0.0 and final search—25. The amino acid residues involved in nonbond interactions were observed using the PLIP server [46]. The docked complex of the top five potent candidates was analysed in UCSF Chimera and PyMOL Molecular Graphics System [47–50].

2.4. ADMET Analysis

Many drugs are withdrawn from the market owing to their detrimental effect witnessed during clinical studies. Hence, the performance of drugs after administration can be anticipated via in silico ADMET studies. The physicochemical and ADMET properties were predicted using ADMETlab 2.0 (https://admetmesh.scbdd.com/) [51–54]. The behaviour of pharmacological candidates inside the body is described by pharmacokinetic properties which in turn are influenced by physicochemical properties [55]. The online server adumbrates several properties mandatory for screening of biologically active compounds such as how effectively the compound is absorbed and distributed in the human body; how the compound would be metabolised and cleared out from the biological system and how much it will affect the biological system either positively or adversely [56].

2.5. Molecular Dynamic Simulation

To validate the docking results obtained from molecular docking, dynamic simulation was performed with GROMACS open-source software using CHARMM27 (all-atom force field) force field [57–59]. Before the simulation, topology files for ligands were created by using SwissParam [60–62]. The unit cell was created, and the recommended TIP3PBOX solvent model was chosen. Energy minimisation of structures was performed before conducting the NVT and NPT ensemble. The NVT and NPT equilibration were conducted for 100 ps, and MD simulation was run for 1 ns to obtain the output values for root mean standard fluctuations, root mean square deviations, the hydrogen bond between protein and ligand, total protein and ligand energy and values for kinetic energy [63, 64]. XMgrace software was used to plot the graph from these output values [65].

3.1. Ligand Optimisation

The minimum energy conformations of all compounds were examined for change in bond length and bond angle; details are provided in Supporting Information (Supporting Information-Optimisation Details). The three-dimensional structures of ligands optimised by Chem3D are given in Figure 2.

3.2. Modelling of Protein

The three-dimensional structure (1TUB) downloaded from RCSB Protein Data Bank has some missing amino acid residues; hence, modelling of protein was conducted through the online open-source software SWISSmodel. The sequence alignment for both chains is shown in Figure 3(a). The MolProbity score was calculated as 2.62, and the local quality estimate plot is presented in Figure 3(b). The MolProbity score suggests the quality of protein by combining all geometric scores clashscore, rotamers and Ramachandran evaluations, and the local quality plot shows the similarity of the model structure to its native structure. The Ramachandran plot predicts the validity of the three-dimensional structure of a protein by discriminating the phi and psi space into allowed and disallowed regions, and here, the plot shows that Ramachandran favoured, Ramachandran outliers and rotamer outliers are 78.79%, 8.34% and 5.97% (Figure 3(c)). The amino acids with poor chemical parameters are presented by C-β deviation (38), bad bonds (0/6944) and bad angles (161/9418). The reliability of quality estimation can be improved by the QMEANDisCo Global score which was found to be 0.61 ± 0.05 (Figure 3(c)).

3.3. Molecular Docking Analysis

The ligands and receptor tubulin protein were subjected to docking through the HEX CUDA server. The binding energy values are mentioned in Table 1, interpreting that Compound 7 has scored the lowest with −361.33 kcal/mol followed by Compound 10 (−345.4 kcal/mol), Compound 4 (−333.8 kcal/mol), Compound 6 (−328.5 kcal/mol) and Compound 15 (−321.85 kcal/mol). Compounds 3 and 13 have almost similar values for binding energy. The outcome of docking studies of Compounds 7 and 10 are better than our previous 1,3-benzodioxole-tagged dacarbazine derivatives [9].

3.4. Visualisation of Docked Complexes

The resulting complex obtained from docking was analysed in Chimera and PyMOL. Compounds 4, 6, 7, 10 and 15 have shown hydrophobic, salt bridge, pi-cationic and hydrogen bond interactions with receptor protein (Table 2). The amino acid residues of tubulin protein involved in nonbonding interactions are supported by literature [31]. Compound 4 has shown hydrophobic interactions with HIS406 and VAL409 amino acid residues of Chain A and hydrogen bond with HIS406 amino acid residue (Figures 4(a) and 4(b)). The protein—Compound 6 complex is stabilised through hydrophobic interactions with amino acid residues ILE163B, TRP407A and LEU263B; salt bridge formation with GLU411A and hydrogen bonding with ARG105A, GLY410A and GLU411A amino acid residues (Figures 4(c) and 4(d)). Compounds 7 and 15 both have hydrophobic interactions with VAL258B, TRP407A and GLU411A and pi–cation interaction with TRP407A. Compound 7 has formed salt bridge with HIS406A and three hydrogen bonds with ARG251B, LEU263B and HIS406A (Figures 4(e) and 4(f)), whereas Compound 15 has one hydrogen bond with LEU263B (Figures 4(g) and 4(h)). Compound 10 interacts with PRO261B, VAL258B, GLU411A and HIS406A amino acid residues through nonbond interaction, and the stability of the complex is enhanced by pi–cation interactions with TRP407A and hydrogen bonding with HIS406A and GLY410A amino acid residues (Figures 4(i) and 4(j)). The protein–ligand interactions are presented in two-dimensional format in Figure 5(a) (Compound 4), Figure 5(b) (Compound 6), Figure 5(c) (Compound 7), Figure 5(d) (Compound 10) and Figure 5(e) (Compound 15).

3.5. Physicochemical and Pharmacokinetic Analysis

In silico studies were carried out with ADMETlab 2.0 server to predict the physicochemical, absorption, distribution, metabolism, excretion and toxicity profile of designed compounds. Here, we presented these properties for potent compounds only in Tables 3, 4, 5, 6, 7, 8, 9, 10. The same data for Compounds 1, 2, 3, 5, 8, 9, 11, 12, 13 and 14 are given in supporting table (SI) (available here).

Table 4. The pharmacokinetic properties required for medicinal use of possible drug candidates.

• Note: The green-coloured cells predict excellent results, and red-coloured cells predict poor results. QED, quantitative estimate of drug-likeness; SA score, synthetic accessibility; Fsp³, number of sp³ hybridised carbon; NPscore, natural product-likeness score.

Table 5. The absorption properties are predicted by in silico method.

• Note: The green-coloured cells predict excellent results, and yellow- and red-coloured cells predict poor results. Caco-2, human colon adenocarcinoma cell lines; F20%, human oral bioavailability 20%; F30%, human oral bioavailability 30%.
• Abbreviations: HIA, human intestinal absorption; MDCK, Madin-Darby canine kidney cells; Pgp-inh, P-glycoprotein inhibitor; Pgp-sub, P-glycoprotein substrate.

Table 6. The distribution properties predicted by in silico method.

• Note: The green-coloured cells predict excellent results, and red-coloured cells predict poor results. Fu, fraction unbound in plasma.
• Abbreviations: BBB, blood–brain barrier; PPB, plasma protein binding; VD, volume distribution.

Table 8. Excretion properties predicted by in silico method.

• Note: The green-coloured cells predict excellent results, and red-coloured cells predict poor results. CL, clearance of drugs; T_1/2, half-life of a drug.

Table 9. Toxicity profile of selected compounds.

• Note: The green-coloured cells predict excellent results, yellow cells indicate medium results, and red-coloured cells predict poor results. hERG, human ether-a-go-go; Ames, Ames test for mutagenicity; ROA, rat oral acute toxicity (mg/kg); FDAMDD, maximum recommended daily dose (mmol/kg-bw/day); carcinogenicity (TD50 values); EC/EI, eye corrosion/irritation.
• Abbreviations: DILI, drug-induced liver injury; H-HT, human hepatotoxicity; SkinSen, skin sensitisation.

For a compound to be a biologically suitable candidate, it should possess number of hydrogen bond acceptors (nHA)—0 to 12; number of rotatable bonds (nRot) 0–11; topological polar surface area (TPSA)—0 to 140 based on Veber rule; logarithm of aqueous solubility value (logS)—4 to 0.5 log mol/L; logarithm of n-octanol/water distribution coefficient (logP)—0 to 3 log mol/L and logarithm of n-octanol/water distribution coefficient at pH 7.4 (logD)—1 to 3 log mol/L. The compounds with molecular weight less than 500 g/mol are more potent to be drug candidate as they are transported and diffused. The logP value measures the lipophilic potency; the compound with higher logP does not pass through the cellular membrane [66]. Bioavailability radar gives a quick glance at physicochemical properties (Figure 6).

The quantitative estimate of drug-likeness (QED) value is output from the integration of eight drug-likeness properties, with a higher score for attractive compounds. QED values lie between 0 (all properties unfavourable) to 1 (all properties favourable) [67]. Synthetic accessibility (SA score) less than 6 indicates easily synthesizable compounds. The higher the natural product-likeness (NP) score, the higher the probability of a molecule being a NP. Halides and dichloroacetate have a higher score of 0.274 and 0.203. The compounds following the Lipinski rule MW ≤ 500; logP ≤ 5; Hacc ≤ 10; Hdon ≤ 5, Pfizer rule, and golden triangle have more favourable ADMET profiles [39, 68, 69]. None of the compounds has violated the PAINS alert. Among these seven compounds, Compounds 7, 10 and 15 should possess favourable properties. Synthetic accessibility score was good for all ligands.

Human colon adenocarcinoma cell lines (Caco-2 cell) permeability is considered an important index for an oral drug due to morphological and functional similarity with the human intestinal epithelium [70, 71]. All selected candidates have a little bit of poor permeability. The apparent permeability coefficient (P_app) for Madin-Darby canine kidney (MDCK) cell permeability is used to effectively absorb chemicals into the body. The compounds have high passive MDCK permeability as the P_app value > 20 × 10⁻⁶ cm/s [72]. P-Glycoprotein belongs to the ATP-binding cassette transporter family. The P-glycoprotein inhibitor (P-gp-inh) and P-glycoprotein substrate (P-gp-sub) values denote the probability of being the corresponding molecule [73]. The p-gp substrate may either enhance or inhibit the pharmacokinetic implications [74]. The red, green and yellow colours in the human intestinal absorption (HIA) column indicate the probability of being a molecule with low, medium and high absorption, parading that Compounds 4, 6 and 10 have better absorption. F20% (human oral bioavailability 20%) and F30% (human oral bioavailability 30%) indicate how efficiently the drug is delivered to the human biological system [75]. Compounds 4 and 10 have better oral bioavailability than 6, 7 and 15 (Table 5).

The compounds that bind to plasma protein (plasma protein binding) in higher amounts have a low therapeutic index and lower fraction unbound (Fu) because they diffuse through the cell membranes less efficiently [76]. The volume of distribution (VD) was considered proper if it lies in the range 0.04–20 L/kg as the value denotes the in vivo distribution of drugs [77]. The peripheral targeting drugs should have little blood–brain barrier (BBB) permeability [78]. Compounds 7, 10 and 15 are found to have better distribution in systematic circulation.

The isozymes of cytochrome P450 family 1A2, 2C19, 2C9, 2D6 and 3A4 metabolise two-thirds of known drugs in Phase I process of drug metabolism and catalyse cholesterol and steroid synthesis [5]. The study of being inhibitor and substrate of cytochrome isozymes is necessary to know the effectiveness and to improve the pharmacokinetics of anticancer medication. The values in Table 7 denote the probability of being substrate/inhibitor [79].

The clearance of drugs (CL) informs about the frequency of dosing required for the drug [80]. CL values lying in the range of 5–15 mL/min/kg denote moderate clearance. The values lying between 0 and 1 indicate the probability of being a molecule with T_1/2 less than or equal to 3. The half-life of a drug (T_1/2) is predicted by involving both factors—clearance and VD (Table 8).

The human ether-a-go-go (hERG)–related gene encodes a voltage-gated potassium channel which is responsible for proper cardiac activity, and blockade of hERG may cause other life-threatening diseases [81]. The output value denotes the probability of being a molecule with IC₅₀ more than 10 μm. All other factors of toxicity are predicted in terms of the probability of a molecule being toxic, within the range of 0 to 1. Human hepatotoxicity (H-HT) and drug-induced liver injury (DILI) factors can predict the adverse effects on the liver which may cause the late withdrawal of drugs from the market. Mutagenicity and carcinogenicity (TD₅₀ values) are interrelated; the mutagenic effect is predicted by AMES toxicity. Rat oral acute toxicity (ROA in mg/kg), skin sensitisation (skinSen), eye corrosion (EC) and eye irritation (EI) factors are evaluated to determine how much drug candidate is safe for our skin and eyes. The maximum recommended daily dose (FDAMDD in mmol/kg-bw/day) value predicts the threshold amount of toxic dose. The diagnosis of respiratory toxicity is delayed because it does not show distinct symptoms but may lead to mortality (Table 9).

Bioconcentration factor is the ratio of the chemical concentration in biota to that in water at a steady state. Bioaccumulation is the absorption of chemical concentrations present in water bodies via a nondietary route, and it may lead to health issues. IGC₅₀ is the concentration of chemical causing 50% growth inhibition of Tetrahymena pyriformis after 48 h, LC₅₀FM is the concentration of compound causing 50% fathead minnow to die after 96 h, and LC₅₀DM is the concentration of drug causing 50% of Daphnia magna to die after 48 h. The environmentally toxic compounds may possess secondary poisoning potential and affect human health via the food chain.

Based on pharmacological evaluation, Compounds 4, 7, 10 and 15 were found more suitable for further studies.

3.6. MDS Analysis

The RMSD, RMSF and radius of gyration were examined as a function of time, and the kinetic energy of the complex was also evaluated in kj/mol. The RMSF value determines the flexibility of individual residues during simulation. The average RMSF value for Compounds 4, 7, 10 and 15 was lying in the range of 0.05–0.15 nm presenting little fluctuation and better stability of binding amino acid residues (Figure 7(a)). The lower RMSF value indicates the less flexibility of amino acid residues lying on the binding site.

RMSD was calculated for four potent ligands—Compounds 4, 7, 10 and 15. The RMSD value steadily increases to 200 ps for all ligands and then reaches a maximum of 1 nm throughout the simulation for Compounds 4, 7 and 10, but Compound 15 has shown a bit more deviation (Figure 7(b)). The MD of Compound 4 has shown consistent stability as its RMSD value is lower than 0.5 nm throughout the simulation. The protein and Compound 15 complex presented the most instability because its RMSD value swayed considerably.

The radius of gyration values (a measure of compactness) for receptor–ligand complexes was determined between 3 and 3.05 nm till 700 ps and then remained stabilised for Compounds 4 and 7 but increased up to 3.09 for Compound 10 and Compound 15 (Figure 7(d)). The lower value of Rg indicates the higher compactness that results from increased interaction between protein and ligand.

The kinetic energy of the complexes was lying between 2.13 × 10⁵ and 2.16 × 10⁵ kJ/mol. The results interpreted from molecular dynamic simulation confirmed the stability of receptor–ligand complexes. Compound 7, Compound 10 and Compound 15 can be possible drug candidates with good binding energy values (Figure 7(c)).

4.1. Potential Synthesis Route

Piperonyl bromide or 5-(bromomethyl)-1,3-benzodioxole can be prepared from piperonyl alcohol using a bromination agent. Then, piperonyl bromide can be further reacted with 4-(dimethylamino) methyl benzaldehyde to obtain 1-(benzo[d][1,3]dioxol-5-yl)-N-(4-formylbenzyl)-N,N-dimethylmethanaminium bromide, followed by metathesis reaction.