Thermodynamic stability, in-vitro permeability, and in-silico molecular modeling of the optimal Elaeis guineensis leaves extract water-in-oil nanoemulsion

Monitoring of MDS, PDI, and zeta potential for thermodynamic stability

The stability of nanoemulsions is an important character that explains the shelf life of any given formulation. Monitoring of MDS, PDI, and zeta potential during 90 days storage at 25 °C is crucial for the optimal nanoemulsion to determine the thermodynamic stability of the system against coalescence and Ostwald ripening phenomena²⁰. Figure 1 illustrates the changes in MDS, PDI, and zeta potential of the optimal nanoemulsion within 90 days storage period at 25 °C. MDS of the optimal nanoemulsion on the first day of preparation was slightly higher than the acceptable range (< 200 nm) (Fig. 1a). This was likely due to the colloidal dispersion system that has yet to stabilize and reach equilibrium. The same trend was observed by another study that prepared a glutathione-loaded water-in-oil nanoemulsion where the MDS increased steadily after 30, 60, and 90 days, showing values of 96.05 nm, 155 nm, 181.3 nm, and 190 nm, respectively.

Furthermore, the stability of the said nanoemulsion was closely associated with the nanodroplets. Ostwald ripening caused the increase in MDS, thus increasing the tendency to undergo phase separation. A drastic rise in MDS indicates low stability of the nanoemulsion system²¹. The MDS of the optimal nanoemulsion was maintained around 180 nm until the 60th day. However, there was an increase in MDS on 90th day to 244.8 nm, likely due to two phenomena: coalescence or Ostwald ripening. This is because the surfactant type and its composition in the formulation play an important for optimal nanoemulsion stabilization. The surfactant is essential to ensure that the water phase is distributed into very small droplets. The outcome can be achieved in the optimal nanoemulsion by the surfactant mixture (Brij L23, Span 80) encompassing and protecting the water droplets from coalescence²².

As can be seen in Fig. 1b, PDI values until 90 days of storage stabilized at ~ 0.25, hence well within the acceptable benchmark (< 0.25) for a stable nanoemulsion system²³. The results thus communicated that the optimal nanoemulsion still exhibited a monodisperse characteristic. The good stability seen in the optimal nanoemulsion was likely contributed by surfactant in the colloidal dispersion system. Contrariwise, the PDI was 0.34 on the 30th day, slightly higher than the acceptable specification. The outcome seen here was presumably due to a certain percentage of the droplets in the nanoemulsion system combining to form slightly bigger droplets. A study was done by Sulaiman et al. observed that the PDI on the 30th day decreased slightly compared to the 1st day. In short, the said nanoemulsion maintained a PDI below 0.25 throughout the 90 days storage period^22,24.

As can be seen in Fig. 1c, the optimal nanoemulsion’s zeta potential changed between − 30.0 to − 40.0 mV over 90 days. This demonstrated the optimal nanoemulsion’s adequate stability to overcome destabilization phenomena through the repulsive forces formed between negatively charged nano-sized water droplets. The optimized nanoemulsion results indicated the good physical stability of its dispersed systems, as similarly observed by a similar work²⁵. Previously, studies have shown that the zeta potential of any given nanoemulsions ought to be highly negatively charged for good stability. A water-in-oil nanoemulsion described by a recent study showed that droplets with a surface charge would get attracted to the droplets of opposite charge, which then formed an electric double layer. The zeta potential of the nanoemulsion was –32.0 mV to –38.2 mV. This proved that the negatively charged droplets were highly stable to overcome coalescence, flocculation, creaming, Ostwald ripening, or sedimentation²⁶. Strong repulsive forces between the nano-sized droplets in the colloidal dispersion system cause a highly negative/positive zeta potential. This agreed with an observation by Roselan et al., which obtained a nanoemulsion loaded with kojic monooleate with a zeta potential of − 30 mV after 90 days of storage²⁷.

Rate of coalescence

Notably, nanoemulsions are proven to be capable of withstanding all types of storage instabilities viz. creaming, sedimentation, aggregation, flocculation, coalescence, and Ostwald ripening compared to microemulsions due to nano-scaled droplets. Nevertheless, their stability could depreciate with time via two well-known distinctive irreversible demulsification processes. They are the Brownian-induced coalescence and Ostwald ripening. Coalescence describes a storage instability phenomenon involving two or more droplets fusing into bigger droplets in nanoemulsion systems²⁸. For instance, the optimal nanoemulsion could undergo a coalescence effect over time as the small water droplets tend to collide with each other. Given enough time, this increases the MDS of the system and affecting the storage stabilities, and eventually result in phase separation.

In this study, the coalescence rate of the optimal nanoemulsion was estimated by plotting the reciprocal of the r² of the MDS obtained from Fig. 1a. A linear graph is obtained for a coalescence affected nanoemulsion stored over time, causing the MDS to increase linearly with increasing time. Figure 2a depicted the plotted graph of 1/r² versus storage time was not linear, revealing that the optimal nanoemulsion stored at 25 °C was stable. The overall trend seen here indicated that the change in MDSs as a function of time was not coalescence-related²⁹. A plausible explanation might be due to the contribution of the surfactant mixture, Span 80 and Brij L23, used in this optimal nanoemulsion. The results thus indicated that both emulsifiers were adequate in preventing droplets from coalescing due to the high viscosity. The wax-like Brij L23 and the highly lipophilic Span 80 were good emulsifiers for coating the water nanodroplets. These surfactants restrict droplets’ movements and avert coalescence³⁰.

Rate of Ostwald ripening

Alternatively, another stability degradation mechanism in nanoemulsion is Ostwald ripening. This process involves the irreversible growth of larger droplets in the nanoemulsions at the expense of smaller ones. In fact, nanoemulsions with a considerably soluble dispersed phase in the continuous phase are prone to destabilization by Ostwald ripening. Ostwald ripening differs from coalescence because the former does not need any contact between the droplets. The changes mentioned above are brought about by the difference in Laplace pressure inside the droplets³¹. Figure 2b presents the plotted graph of r³ versus storage time, which revealed that the optimal nanoemulsion stored at 25 °C was under significant influence of the Ostwald ripening, based on the increase in MDS from 206.3 nm on day-1 to 244.8 nm on day-90. However, the r³ value of the optimal nanoemulsion somewhat plateaued between the 30th and 60th day. In the first 2 months, the nanoemulsion stability likely resulted from the lipophilic surfactant Span 80 that halted the Ostwald ripening effect. The surfactant molecules restricted the water droplets’ diffusion into the continuous oil phase.

On the other hand, the increasing MDS in the last 30 days of this study was due to the diffusion of the smaller water droplets through the bigger water droplets (Fig. 2b). It was apparent that the ambient temperature promoted the Ostwald ripening phenomenon in the optimal nanoemulsion, related to the water droplets’ high kinetic energy in the dispersion phase that increased their movement³². The outcome seen above was in good agreement with the earlier report by Gupta et al. They found the oil phase’s chemical potential (sunflower seed and olive oils) was higher in smaller droplets than in bigger ones. This factor allowed the mass transfer of the smaller oil droplets to the larger droplets³³. A curcumin-loaded MCT nanoemulsion prepared by Kim et al. also demonstrated similar behavior. In their work, the sorbitan monooleate/Span 80 (HLB: 4.3) was the selected emulsifier, which successfully retarded the Ostwald ripening phenomenon toward the MCT oil phase by forming an interfacial layer, possibly from within the droplets³⁴.

Heavy metal analysis

Nanocosmeceutical products should be free from the presence of heavy metals such as arsenic, cadmium, lead, and mercury due to their high toxicity, skin irritation, and cancerous effect on consumers. The FDA has set acceptable limits for these heavy metal presence in cosmetics which are arsenic (< 3 ppm), cadmium (< 0.3 ppm), lead (< 20 ppm), and mercury (< 1 ppm) (FDA, 2018)^35,36. In this study, the heavy metal analysis on the optimal nanoemulsion confirmed that arsenic, cadmium, lead, and mercury was detected at low levels, namely < 0.5 ppm, < 0.1 ppm, < 0.5 ppm < 0.1 ppm, respectively. The results indicated that the ultrasonically extracted polyphenolic compounds (gallic acid and catechin) from the E. guineensis leaves in 50% ethanol:water yielded safe active ingredients for the topically applied optimal nanoemulsion. The heavy metal analysis results corroborated that the leaves extract was safely purified, thus indicating the extraction procedure’s suitability to isolate the polyphenolic compounds from the raw oil palm leaves.

It is pertinent to indicate here that there are frequent reports about the heavy metals in creams, mostly from the contaminated raw material. This was due to the lack of compliance by small-scale manufacturers and the lack of strict FDA regulations. Specifically, color additives and perfumes may contain these residual impurities (arsenic, cadmium, lead, and mercury), which eventually contaminate the nanoemulsions during preparation³⁷. Henceforth, the aforementioned heavy metals are carcinogenic and prohibited in cosmetics. In conclusion, the optimal nanoemulsion is safe for topical use because the heavy metals are within the allowable range³⁵.

Antimicrobial assay

Topical nanoemulsion should be investigated for antimicrobial assay to determine their inhibitory effects against different microorganisms or gauge whether the emulsions formulation might promote microbial growth. The cosmetic products are usually stored at varying storage conditions, largely between 4 °C to 25 °C. Cold and ambient environments are suitable for microbial growth in any medium; thus, it is important to test the cosmetics³⁸. The antimicrobial assays confirmed the absence of Candida albicans, Pseudomonas aeruginosa, Staphylococcus aureus, Total Yeast & Mould, and Total Plate Count in the optimal nanoemulsion. These microorganisms were used for the study’s antimicrobial assay as they are the microbial agents of concern in cosmetic products, which might degrade and jeopardize the product’s quality and safety.

The Total Plate Count and Total Yeast & Mould colonies were detected below 10 CFU/g in the optimal nanoemulsion, while growths of S. aureus, P. aeruginosa, and C. albicans were not detected on the test plates. Henceforth, the outcome corroborated that the optimal nanoemulsion is an unsuitable medium for pathogen growth if kept in cold and/or ambient environments during long-term storage. As a precaution, the added antimicrobial agent, 2-phenoxyethanol, probably inhibited pathogen growth in the optimal nanoemulsion.

Supposedly, the nanoemulsions were not assayed for antimicrobial properties. However, pathogenic Gram-positive bacteria and Gram-negative bacteria might thrive in them during prolonged storage conditions. This is because warm temperatures are suitable environments for the growth of pathogenic bacteria^39,40. Also, the larger surface area in smaller droplets promotes their greater interaction with microbial cells⁴¹.

In-vitro permeability

The in-vitro permeability of the optimal nanoemulsion was carried out to study the bioavailable amount of polyphenolic compounds for skin penetration when applied topically⁴². The commercially available synthetic Strat-M membrane was used in this in-vitro study. This is due to its multiple polyethersulfone layers that are morphologically similar to the human skin and have a very tight surface layer. This synthetic membrane is composed of two layers of polyethersulfone lying on top of another layer of polyolefin. This membrane type is the most suitable synthetic membrane to simulate human skin for active compounds or drug diffusion experiments⁴³. The calibration curve of the TPC of EgLE (R² = 0.9925) (Fig. 3) was constructed to determine the amount of the TPC of EgLE diffused through the Strat-M membrane from the donor compartment to the receptor compartment of the Franz diffusion cell. The amount of TPC of EgLE permeated hourly was determined by multiplying the volume of the receptor compartment with the concentration of the corresponding hourly sample obtained from the plotted calibration curve, followed by division of the yielded amount with the diffusional area. As can be seen, the TPC of EgLE permeation increased with increasing time, reaching up to 15% of the loaded nanoemulsion after 8 h (Fig. 4). Overall, the TPC of EgLE permeated across the membrane almost in a linear relationship, indicating a controlled and sustained release of the TPC of EgLE from the optimal nanoemulsion.

Literature has shown that the active ingredients’ properties and the delivery system’s type are factors that influence their controlled release from the nanoemulsion¹³. As seen in this study, the permeation of the targeted compounds of EgLE, namely, gallic acid and catechin, in the form of encapsulated water droplets from the optimal nanoemulsion through the Strat-M membrane increased as a function of time. The outcome seen here was likely due to their minute droplet size and low PDI alongside the stability of the water–oil interface. Furthermore, the surfactant mixture (BrijL23 and Span 80) used in the optimal nanoemulsion might also influence the water droplets’ permeability or release. This led to the controlled mass transfer of the encapsulated gallic acid and catechin from the optimal nanoemulsion through the Strat-M membrane, as similarly reported by recent studies^42,44. In this study, the surfactant mixture (29%) is higher than the oil phase (10%) in the optimal nanoemulsion. This feature was important to delay the rapid mass permeation of the encapsulated polyphenolic/hydrophilic compounds in the water phase (58%).

Recently, Strat-M membrane is widely used as a synthetic membrane for in–vitro permeability assay to simulate human skin. The polyethersulfone layer is more resistant to diffusion, whereas the polyolefin layer is more diffusive⁴⁵. On the other hand, the receptor medium contained 50% ethanol because it acts as a permeation enhancer⁴⁶. The graph for the cumulative amount of the TPC of EgLE permeated per unit area versus time was plotted as depicted in Fig. 5. The total amount of the TPC of EgLE permeated across the membrane per unit area after 8 h of study is 1935 ± 45.7 µgcm⁻², with the steady-state flux (Jss) of 241.9 ± 5.7 µgcm⁻² h⁻¹ and permeation coefficient (Kp) value of 1.15 ± 0.03 cm.h⁻¹.

Kinetic release study

In this study, the kinetic release study was performed to determine a possible release mechanism and best fitted kinetic release models viz. zeroth-order, first-order, Higuchi, Hixson-Crowell, and Korsmeyer-Peppas for the optimal nanoemulsion. The best fitted kinetic release model was chosen based on the coefficient of determination (R²) approaching 1. The R² for all the five kinetic release models for the TPC of EgLE from the optimal nanoemulsion is tabulated in Table 1, and Fig. 6 shows the plotted graph for (a) zeroth order, (b) first-order, (c) Higuchi, (d) Hixson-Crowell, and (e) Korsmeyer-Peppas models.

In this study, the kinetic release mechanism of the optimal nanoemulsion was best fitted to the Korsmeyer–Peppas model, as seen from the high linearity (R² = 0.9961) (Fig. 6). According to the equation, Korsmeyer–Peppas equation described the type of diffusion by the release exponent, n = 13.867, higher than 1. The findings implied that the release of the TPC (gallic acid and catechin) of EgLE from the optimal nanoemulsion exhibited a supercase II transport behavior¹⁴. The outcome meant that the release of the TPC of EgLE from the optimal nanoemulsion was controlled by the swelling and relaxation of the crosslinked network of the thickener (HPMC), surfactant mixture, and oil phase. The diffusion rate of the TPC of EgLE was also governed by the concentration gradient and polymer relaxation resulting from water uptake and the osmotic pressure occurring during the swelling process⁴⁵.

Characterization of the AQP-3 protein

The AQP-3 Protein (EC 2.7.11.11) has 292 amino acid residues, and the BRENDA analysis revealed it to be a true protein. The Protparam analysis of the AQP-3 protein primary structure seen in Table 2 showed the amino acid sequence of the AQP-3 protein to comprise 4461 atoms with a corresponding molecular formula and molecular weight of C₁₄₅₉H₂₂₃₃N₃₇₃O₃₈₃S₁₃ and 31543.83 Daltons, respectively. Notably, a computed theoretical pI value below 7 indicates acidic characters, whereas the AQP-3 protein has a low theoretical pI (6.74). This indicated the acidic nature of the AQP-3 protein, which correlated with the high number of acidic residues (Asp + Glu = 16).

Additionally, the instability index of AQP-3 protein, 22.68, conveyed that the protein is stable in–vitro and has a high aliphatic index of 105.86. This high index verifies the protein’s thermal stability, as stability- and aliphatic indices of < 40 and > 40, respectively, are the collective attributes of a thermally stable protein⁴⁷. The negative values of the grand average of hydropathicity (GRAVY) of AQP-3 protein (0.535). This value signified that the AQP-3 protein is hydrophobic. Notably, GRAVY is used to measure the hydrophobicity and solubility of protein¹⁸. A negative GRAVY value denotes hydrophilic protein, while a positive value indicates a hydrophobic protein.

Analyses of the quality of the modeled structure of AQP-3 protein (before and after energy minimization) were performed using PROCHECK, ERRAT, and Verify3D. Meanwhile, the homology model of the AQP-3 protein PROCHECK-generated Ramachandran plots analyzed the polypeptide backbone torsion angles phi (U) and psi (w) of amino acids. The results of the Ramachandran plot (Fig. 7) before and after energy minimization are summarized in Table 3. Residues in the most favorable region were high at 91.2%, followed by 7.3% in the additional allowed region, 1.5% in the generously allowed region, and no residue was found in the disallowed region after energy minimization. Additionally, 100% of the non-glycine and non-proline residues were distributed within the allowable regions. Thus, a good quality protein homology model should have over 90% of residues residing in the most favorable region⁴⁷. Hence, the 3D model of AQP-3 protein was proven stereochemically satisfactory.

The general quality of the AQP-3 protein 3D model was measured using ERRAT, wherein the ERRAT was 99.16% (Fig. 8). Literature has shown that a good protein model should have an ERRAT score of > 50% (15,48). Meanwhile, the ERRAT histogram illustrates the correct regions as grey, and the incorrect regions are colored black; white bars signify the region with a lower error rate of protein folding. The data revealed that the two lines were drawn at 95% and 99% as the confidence level to reject the regions which exceeded the latter error value. This meant that the energy minimization step improved the structural quality of the 3D model of the AQP-3 protein. The results thus corroborated the excellent quality of the predicted AQP-3 protein structures (> 50%) and therefore are reliable for subsequent structural analysis.

Consequently, the local environment of the AQP-3 protein was checked using Verify-3D. The program determines the compatibility of an atomic model (3D) with its own amino acids sequence. It is germane to indicate here that a satisfactory protein model has a Verify-3D score of > 80%⁴⁸. Since the Verify-3D scores (Fig. 9) for the AQP-3 protein after energy minimization scored 84.46%, this indicated that the side-chain environments of each residue in the AQP-3 protein model were satisfactory (Table 3).

Molecular docking

A molecular docking study was attempted to examine the in-silico ability of the ligands (gallic acid and catechin) in the optimal nanoemulsion to bind with the AQP-3 protein to improve skin moisturization and it was done using the AutoDock Vina tool. Here, binding energies, binding interactions, and the AQP-3 protein binding sites to gallic acid and catechin were monitored and compared to gauge the protein-substrates complexes’ binding quality. Therefore, the ligands have to bind to AQP-3 protein to increase the expression of the proteins to hydrate the human skin better.

Similarly, a patent (FR2874502A1 dated 26th August 2004) explained the effect of different ligands from pomegranate extract on the expression of AQP-3 protein. It revealed that the pomegranate extract induced the expression of the AQP-3 protein for the moisturization effect, which is an important criterion for cosmetic products. On the other hand, the binding energies of AQP-3 protein with compounds through computational modeling for skin cancer were also investigated⁴⁹. The study found that the two ligands’ viz. gallic acid and catechin binding energies were − 5.6 kcal/mol and − 7.4 kcal/mol, respectively (Table 4), thus showing a higher preference of the latter to bind to the AQP-3 protein.

It is important to note that the lowest binding energy (kcal/mol) represents a simple estimate of a better or stronger substrate affinity towards the protein for protein-substrate conformation¹⁶. Also, the protein–ligand complex’s specific docking helps predict the ligand’s preferred orientation when bound to the protein¹⁵. In this case, a stronger moisturization effect is expected by the preferable binding of catechin to the AQP-3 protein over gallic acid. The effect was inferred from the higher number of intermolecular interactions between the AQP-3-catechin complex. Figure 10 showed the corresponding LigPlot analysis of the AQP-3-ligand complexes. Gallic acid was bonded to the AQP-3 protein by a single amino acid, Asn 215, via hydrogen and hydrophobic interaction. While more amino acid residues interacted with catechin through five hydrogen bonds via Arg 95, Ser 78, Glu 28, Arg 23, and Gln 24 located nearby the catalytic site of AQP-3 protein.

Based on Table 4, the lowest binding energy for the AQP-3-catechin complex (− 7.4 kcal/mol) indicated a stronger AQP-3-ligand complex interaction compared to gallic acid (− 5.6 kcal/mol). The lower binding energy of the AQP-3-catechin complex corresponded well with the five hydrogen bonds that held the complex together at the distances of 3.01 Å, 2.76 Å, 3.06 Å, 2.92 Å, and 2.86 Å. Conversely, the AQP-3-gallic acid complex’s higher binding energy contributed to a single hydrogen bond with a distance of 3.04 Å to Asn 215. It is pertinent to indicate here that < 3.5 Å is the cut-off distance for an intermolecular hydrogen bond. Hence, a longer bond distance (> 3.5 Å) conveys a lower binding affinity of a protein towards a ligand and is less likely to bind the ligand.

Molecular dynamics simulation

MD simulation is crucial in silico investigation to demonstrate and analyze the behavior, structural flexibility, and stability of the protein when bonded to different ligands⁵⁰. In this study, the structural changes, stability, and flexibility of AQP-3 protein with the ligands (gallic acid and catechin) were monitored by comparing the values of RMSD, RMSF, Rg, and hydrogen bonds distance. The following results were the mean value of triplicated analyses of the AQP-3-ligands complexes simulated for 100 ns.

Root mean square deviation

It is important to indicate here that a low RMSD value (< 0.3 nm) conveys a strong binding affinity and the formation of a stable complex⁵¹. The analysis of RMSD of the AQP-3 protein backbone was calculated to describe the conformational changes of the two different ligands, gallic acid and catechin. As can be seen, the low RMSD values (RMSD ~ 0.1–0.35 nm) for the AQP-3-catechin and AQP-3-gallic acid complexes indicated the strong binding between the two complexes, with the former being a stronger one. The outcome was suggestive of better skin hydration through catechin in the optimal nanoemulsion. The RMSD value of the AQP-3-gallic acid complex reached equilibration considerably earlier after 50 ns at ~ 0.25 nm than the AQP-3-catechin complex, which fluctuated for 75 ns before achieving equilibrium around 0.30 nm during the production simulation (Fig. 11a).

Based on the MD simulation results, although AQP-3 protein binds preferably to catechin over gallic acid, as shown in the molecular docking studies (Table 4), it took a while to form the stable AQP-3-catechin complex. It was plausible that a longer duration was needed for the five hydrogen bonds to bind the two moieties than only a single hydrogen bond in the AQP-3-gallic acid complex. The results conveyed that the moisturizing ability of the optimal nanoemulsion was contributed by the rapid action of catechin to bind to the AQP-3 protein. Although gallic acid eventually bonded to the AQP-3 protein, the rate was slower than that of catechin. Similarly, α-glucosidase bonded more favorably to catechin than gallic acid. This was reported by Choudhary et al., where the RMSD observed was among the more prevalent MD trajectories for the catechin-α-glucosidase and the gallic acid-α-glucosidase, corresponding to 0.2 and 0.3 nm, respectively. They confirmed that catechin inhibited α-glucosidase activity more efficiently than gallic acid⁵².

Root mean square fluctuation

RMSF analysis imparts information on residue-specific flexibility as the parameter calculates the individual residue flexibility or the extent of any particular residue moves (fluctuates) during an MD simulation. The prediction was made by measuring the amount of movement along a principal axis. It is important to indicate here that the highest value of RMSF represents a higher degree of movement, whereas the lowest RMSF value implies a more stable structure from limited structural fluctuation during the MD simulation^53,54. In this study, the average RMSF plot (Fig. 11b) showed moderately high structural stability levels for the complexes in the range of 0.05–0.45 nm.

Notably, the RMSF value of the AQP-3-gallic acid complex was slightly lower than that of the AQP-3-catechin complex, thus indicating that the gallic acid bonded more tightly to AQP-3 protein at residue positions, 128, 135, 153, 180, 215, and 230. The data demonstrated that AQP-3 protein is bonded slightly less flexible with catechin, which is in good correlation with the RMSD values in the previous subsection 3.10.1, exhibiting the fast equilibrating AQP-3-gallic acid complex (50 ns) over the AQP-3-catechin complex (75 ns). This could be explainable that the gallic acid is much smaller in structural size compared to catechin. This would limit the structural movement and flexibility of gallic acid upon binding with the AQP-3 protein. Albeit the overall RMSF values indicated that both the complexes exhibited low structural movements and flexibilities during simulation trajectories, it is found that the AQP-3-catechin complex formed more stable binding and increased the expression of AQP-3 protein for better skin hydration. Similarly, Anuar et al. concluded earlier that the mutated LipKV1-tributyrin exhibited greater stability at pH 8.0 because it recorded a lower RMSF value than the LipKV1-tributyrin complex^15,16.

Radius of gyration

In this investigation, the average Rg value of the AQP-3 complexes fluctuated in the range between 1.70 nm–1.80 nm (Fig. 12). Remarkably, AQP-3-catechin decreased from 30,000 ps onwards before achieving equilibration with small fluctuations at ~ 1.755 nm. Contrariwise, the AQP-3-gallic acid complex fluctuated before reaching equilibrium at 1.75 nm from 70,000 ps onwards. The catechin and gallic acid were loosely packed with the AQP-3 amino acid residues based on the Rg values. These AQP-3-ligands are less compact, hence can move more freely. According to the literature, the highest Rg values indicate a looser packing of amino acids with ligands and vice versa. Likewise, Rg values for all the DehH2-ligand studied fluctuated between 1.75–1.85 nm. These values revealed less compact conformation of the DehH2-ligand complex, which conformed to the poor binding to the enzyme’s catalytic residues^17,55.

Hydrogen bonds analysis

Hydrogen bonds and their relative strength in a water environment are vital to enable protein–ligand binding. This is particularly true when the mechanism of action involves hydrolysis, where water plays an important role in the compound’s breakdown. Hydrogen bonds are formed when an electronegative atom of a hydrogen-bond acceptor binds to a hydrogen atom directly bonded to a hydrogen-bond donor⁵⁶. Figure 13 depicted the number of hydrogen bonds in the AQP-3-gallic acid and –catechin complexes. Catechin appeared to bind strongly to AQP-3 protein compared to gallic acid, despite the formed four hydrogen bonds at 70 ns in the latter, before stabilizing to a two-hydrogen bonded AQP-3-catechin complex from 88 ns onwards. The study proved that the AQP-3-catechin complex was stabilized much later, equilibrating at a two-intermolecular hydrogen-bonded complex from 88 ns onwards.

The AQP-3-gallic acid output seen in this study exhibited a higher number of intermolecular hydrogen bonds than the corresponding docking study in subsection 3.9. On the other hand, there were only two hydrogen bonds in the AQP-3-catechin complex instead of five, as predicted by molecular docking. The findings are somewhat justifiable that water molecules were included in the system during MD simulation. Hence, the produced output is a close resemblance to the actual protein hydrolysis system. Therefore, AQP-3-catechin exhibited stronger hydrogen bonding in comparison with AQP-3-gallic acid. The results proved that catechin was the better polyphenolic compound to hydrate the skin. The earlier study also observed the same outcome wherein the number of hydrogens changed throughout the simulation, but the changes were more profound¹⁷.

MM-PBSA binding free energy

Based on the previous studies, data derived from the average binding energy (kcal/mol) calculations can offer a better insight into the protein–ligand complexes’ interactions. Consequently, the binding strengths between the AQP-3 protein and the tested ligands (gallic acid and catechin) were assessed through binding free energy calculation via MM-PBSA based on MD trajectories of van der Waals, electrostatic, polar solvation, and nonpolar solvation energies. MM-PBSA offered more reliable and accurate predictions on AQP-3-gallic acid and –catechin complexes than molecular docking. This is because of the molecular docking system’s rigidity in the absence of water molecules that limit the protein–ligand binding’s flexibility to certain types of motions. Moreover, the energy derived from MD trajectories is more responsive and flexible when the protein–ligand complex is interchange effectively throughout 100 ns simulation^15,17.

As tabulated in Table 5, AQP-3-catechin complex exhibited the most favorable binding energy, − 57.514 kcal/mol compared to AQP-3-gallic acid (− 8.553 kcal/mol). The data corresponded with the AQP-3-gallic acid complex’s highest van der Waals energy (− 0.063 kcal/mol) compared to AQP-3-catechin (− 63.335 kcal/mol). The highest electrostatic energy of AQP-3-gallic acid (− 0.108 kcal/mol) indicated that this complex is less positively charged compared to AQP-3-catechin, which showed the lowest electrostatic energy (− 1.985 kcal/mol)¹⁶. As can be seen, the MM-PBSA calculation validated MD simulation data where catechin interacted spontaneously with the AQP-3 protein, as seen by the negative binding value of Gibbs free energy. The more negative binding energy is an indication that the reactions are spontaneous. Hence, the more negative the reactants’ electrostatic energy, they tend to undergo reactions without an external energy source¹⁵. In this milieu, the protein–ligand complex is converted quicker into the product at a higher reaction rate. In conclusion, AQP-3 protein prefers to bind with catechin to increase the expression of AQP-3 protein for skin hydration.