Unbiased picture of the ligand docking process for the hevein protein–oligosaccharide complex

Abstract

The ligand-docking behavior of hevein, the major latex protein from the rubber tree Hevea brasiliensis (Euphorbiaceae), has been investigated by the unguided molecular dynamics (MD) simulation method. An oligosaccharide molecule, initially placed in an arbitrary position, was allowed to move around hevein for a prolonged simulation time, on the order of microseconds, with the expectation of spontaneous ligand docking of the oligosaccharide molecule to the binding site of hevein. In the binary solution system consisting of a hevein molecule and a chito-trisaccharide (GlcNAc₃) molecule, three out of the six separate simulation runs successfully reproduced the complex structure of the observed binding from. It appeared that the surface topology formed by two aromatic side chains of the hevein molecule played a role in orienting the GlcNAc₃ molecule in the correct direction. We also performed MD simulations of the ternary solution system containing a cello-hexasaccharide (Glc₆) molecule in addition to hevein and a chito-hexasaccharide (GlcNAc₆) molecule. Formation of hevein–GlcNAc₆ complex structures was exclusively observed, while the Glc₆ molecule remained in the solvent phase throughout the simulations. Obviously, the acetamide groups of GlcNAc play a role in detecting the binding site and its vicinity on the protein surface.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-87407-8.

Introduction

Many plants protect themselves from pathogenic attack by producing defense proteins, some of which have the ability to reversibly bind to chitin, β-(1→4)-linked N-acetyl D-glucosamine (GlcNAc), which functions as a structural polysaccharide in the cell walls of fungi and the exoskeletons of invertebrates. Most plant proteins with antifungal activity contain a noncatalytic, plant-specific chitin-binding domain, which is called the hevein domain after hevein, the major latex protein of the rubber tree, Hevea Brasiliensis (Euphorbiaceae), involved in the coagglutination of rubber particles¹. This domain shares a common structural motif of 30–43 residues rich in glycine and cysteine residues at highly conserved positions and organized around a three to five conserved disulfide bond core². In addition to hevein itself, this domain has been detected in pseudohevein, Urtica dioica agglutin, wheat germ agglutin, Amaranthus caudatus antimicrobial peptides, and the chitin-binding modules of class I chitinases^3,4.

Three-dimensional structure analyses of hevein and its chito-oligosaccharide complexes have been performed by NMR and X-ray measurements, which have provided the atomistic details of the ligand-binding structures^5–9. Aboitiz et al.¹⁰ proposed the three-dimensional NMR structure of HEV32, truncated hevein lacking eleven C-terminal amino acids, and the triacetylchitotriose (GlcNAc₃) complex form. In the structure of the hevein–GlcNAc₃ complex (Fig. 1), two of the tryptophan side chains of the hevein molecule (W21 and W23) are bound to the pyranose rings of the GlcNAc₃ molecule, which appears to be stabilized by CH–π stacking interactions and van der Waals interactions at the interface of the complex. The GlcNAc residue at the nonreducing end is attached to the side chain of W23, which can be defined as the subsite + 1. The hydroxyl groups of residues serine 19 (S19) and tyrosine 30 (Y30) are involved in hydrogen bonding with the carbonyl group of the acetamide moiety and the O3–H group of the nonreducing GlcNAc residue.

Fig. 1.

Reference model of the complex of hevein with a chito-trisaccharide (GlcNAc₃) molecule that is accommodated at subsites + 1 to + 3 from a nonreducing residue¹⁰. The three aromatic amino acids composing the binding site are shown in space-filling representation.

Many noncatalytic carbohydrate-binding proteins, including hevein, provide a shallow ligand-binding surface¹¹ for carbohydrate ligands whose polar functional groups are almost symmetrically aligned with respect to the molecular chain, suggesting that the proteins can form different ligand binding modes with similar binding energies. Jiménez-Barbero and co-workers^12,13 discussed the docking characteristics of the hevein complex, which they referred to as a typical example of protein–carbohydrate interactions. In particular, they pointed out that this feature can be attributed to the amphiphilic character of carbohydrate ligands^12,13. The hydroxyl groups with hydrophilic nature and the aliphatic C–H region forming a hydrophobic patch on the pyranose ring faces account for the formation of hydrogen bonds with the side-chain of the polar amino acid and the occurrence of hydrophobic interactions with the aromatic side chain, respectively. The latter interactions result from the mutual shielding from bulk water at the interface. The interfaces may also involve attractive electrostatic interactions derived from the positive net charge of the C–H groups and the delocalized π electrons of the aromatic rings.

Theoretical calculations, such as molecular dynamics (MD) simulations and quantum mechanics calculations, have provided further insights into the motional behavior at the binding site, including the bound oligosaccharide ligand conformation and thermodynamics associated with ligand binding^{10,12,14–18}. These theoretical studies were performed based on the experimental structure as a starting structure or as a target structure to simulate the process of ligand docking. Shan et al.¹⁹ reported an unguided MD simulation study of the prediction of the complex structure of Src kinase and the kinase inhibitor PPI, in which the MD simulations were continued until the PPI ligand initially located at an arbitrary position in the simulation box reached the target binding site on the protein surface. Several of the simulations successfully provided the ligand-binding structure, which was virtually identical to the crystallographically determined complex structure. The simulation trajectories provided a continuous atomistic view of the entire ligand-docking process, including the intermediate ligand conformations and the role of solvent water molecules in the ligand binding. The same simulation tactic was applied to the MD system consisting of HEV32 and chito-oligosaccharide molecules, either GlcNAc, GlcNAc₂, or GclNAc₃, by Solanke et al.²⁰ While most of the simulation trajectories showed irreversible binding of the chito-oligosaccharide molecules to the ligand binding sites, the MD simulation with a GlcNAc₃ molecule failed to reproduce the identical ligand-binding mode to that observed in the NMR complex structure¹⁰. Instead, the reducing-end residue of the GlcNAc₃ molecule was bound to the subsite + 1 while the disaccharide moiety remained in the bulk water.

In the present study, we performed unguided MD simulations combined with modified van der Waals parameters and extended the simulation time to 10 µs. We first confirmed that the MD simulations successfully reproduced the identical hevein–GlcNAc₃ complex structure to that observed in the NMR complex structure¹⁰. The ligand-docking processes were analyzed by visualizing the MD trajectories. In particular, we investigated the possibility that a GlcNAc₃ molecule binds to the substrate-binding site via a specific pathway during complex formation, including the possibility of the presence of a hot spot; this information cannot be obtained by means of experimental approaches such as NMR and X-ray measurements. The MD simulations were then applied to predict the structures of an unknown complex form with a longer substrate, a hevein–GlcNAc₆ complex, and to evaluate the substrate specificity of hevein to a chito-oligosaccharide compared to a cello-oligosaccharide.

Methods

The X-ray crystal structure of a hevein domain (4WP4) extracted from latex surgical gloves⁹ was used as the starting structure. The protein consisted of 43 amino acids and involved four disulfide bonds. The reference model of the carbohydrate–hevein complex was constructed by combining the two structures of a Hev b 6.02 molecule and a truncated hevein (HEV32) complex with a chito-trisaccharide (GlcNAc₃) (1T0W)¹⁰. The binding site of hevein is composed of the aromatic side chains W23 and W21, which serve as the subsite + 1 and + 2 for a chitin substrate, respectively (Fig. 1). The + 1 subsite accommodates the nonreducing GlcNAc residue, and its 2-acetamide group is the contact with the side chain of Y30. Two hydrogen bonds are formed between the 3-OH group of the nonreducing residue and the hydroxyl groups of the Y30 and S19 side chains¹⁰. Six binary solution systems consisting of a hevein molecule and an oligosaccharide ligand were constructed by arbitrarily placing an oligosaccharide ligand on each of the six sides of the hevein molecule (Fig. 2). The oligosaccharides used in the present MD simulations were a chito-trisaccharide (GlcNAc₃), a chito-hexasaccharide (GlcNAc₆), and a cello-hexasaccharide (Glc₆). The latter two saccharides were used in competitive unguided MD simulations for ternary solution systems in which GlcNAc₆ and Glc₆ molecules were first placed on opposite sides of a hevein molecule. These solute systems were solvated with 5000 TIP3P water models^21,22.

Fig. 2.

Six starting positions of the oligosaccharide (GlcNAc₃) molecule around the hevein molecule.

The solvated complex systems were subjected to energy minimization steps in which the water configurations were first optimized, followed by full structure optimizations. The systems were gradually heated from 100 to 300 K in 500 ps using the NVT ensemble. The density of the systems was equilibrated by 500-ps NPT ensemble simulations with a constant temperature of 300 K and pressure of 1 bar. The equilibrated systems were subjected to unbiased production MD simulations using the NPT ensemble for 10 and 5 µs for the hevein–GlcNAc₃ solution, and hevein–GlcNAc₆ and hevein–GlcNAc₆/Glc₆ solutions, respectively. For both the binary and ternary solution systems, six separate simulations were performed, each differing in the initial position(s) of the oligosaccharide ligand(s). The temperature and pressure of the system were regulated by a Langevin thermostat^23,24 with a collision frequency of 5 ps^− 1 and a Berendsen barostat²⁵ with a pressure relaxation time of 1.0 ps, respectively. A time step of 2.0 fs was used with the SHAKE option²⁶ to ensure rigid hydrogen atoms. The particle mesh Ewald method²⁷ was used for the long-range interactions, and the van der Waals interactions were cut off at 8.0 Å. All of the energy minimizations and MD simulations were performed with the PMEMD and PMEMD.CUDA modules^28,29 of the Amber 16 program package³⁰ using an NVIDIA Kepler GPU system. The hevein molecule was described by the ff14SB force field^31–33, and the oligosaccharide molecules were described by the GLYCAM06 force field³⁴. In the present MD simulations, the parameters describing the intermolecular interactions of the carbohydrate and amino group carbon atoms were corrected to weaken the excessively attractive interactions by using modified Lennard-Jones parameter values developed to reproduce the experimental osmotic pressures of mixed aqueous solutions of diglycine and sucrose^35,36.

The spatial motion of the oligosaccharide ligand(s) during the MD simulations was monitored by evaluating the root-mean-square deviation (RMSd) values of the positions of the pyranose ring and glycosidic oxygen atoms, with respect to those of the observed complex NMR structure¹⁰ as a reference structure. The value indicated how far an oligosaccharide ligand was located from the binding site, so that when the bound position of a GlcNAc₃ part was identical to that in the NMR structure RMSd = 0 Å. As an alternative assessment of the bound position, the binding energy (E_bind), representing the nonbonded interactions composed of van der Waals and electrostatic interactions between an oligosaccharide molecule and the hevein surface in the gas phase, and the binding free energy including Poisson-Boltzmann solvation energy of an implicit solvent model (G_bind^PBSA) were calculated from the MD trajectories using the MMPBSA module³⁷. The MD trajectories were analyzed using the CPPTRAJ module of AmberTools 19³⁸, and VMD 1.9.3³⁹ was used for visual analysis. The molecular graphics software PyMOL (version 2.5.0)⁴⁰ was used to visualize the system.

Results and discussion

Ligand-docking behavior of the binary solution containing hevein and GlcNAc₃ molecules

The ligand-docking behavior in the MD simulations of the binary solution containing hevein and GlcNAc₃ molecules is shown in Fig. 3, which is represented by the two-dimensional ligand-RMSd and binding energy (ΔE_bind) distributions (Fig. 3A) and ligand-RMSd profiles with respect to the simulation time (Fig. 3B). Among the six separate docking simulations, simulation runs 2, 3, and 6 showed the formation of the hevein–GlcNAc₃ complex, while hevein remained in the unbound state in the remaining simulation runs 1, 4, and 5 throughout 10 µs simulation times. In Fig. 3A, the three high-density dots with RMSd values close to the origin, namely, those near the reference complex structure, indicate the formation of the complex structures. All of the complex structures, labeled bound poses a, b, and c, survived for approximately 100–400 ns before they reversibly changed to an unbound state (Fig. 3B). The two complex structures of bound poses a and b obtained from the MD trajectories of simulation runs 2 and 3 are shown in Fig. 3C. Bound pose a, in which the GlcNAc residues were accommodated at subsites + 1 to + 3, was consistent with that of the observed NMR complex structure¹⁰. Although the complex structures of bound pose a appeared in higher ΔE_bind ranges than those of bound poses b and c, bound pose a can be considered to be the most stable in terms of the total complex lifetime. It should be noted that the contribution of the solvent effect was not taken into account in the calculation of ΔE_bind, which is the sum of the intermolecular van der Waals and electrostatic interactions in the gas phase. Compared with bound pose a, the nonreducing end residue of the GlcNAc₃ molecule shifted to subsite − 1 from subsite + 1 in bound pose b. A small distribution of ligand-RMSd and ΔE_bind dots, labeled c, appeared at approximately 8500 ns for simulation run 2 with a relatively short lifetime. Bound pose c was identical to bound pose b with respect to the subsites occupied by the GlcNAc residues, but the direction of the GlcNAc₃ molecule was reversed.

Fig. 3.

Ligand docking behavior of the binary solution containing hevein and GlcNAc₃ molecules obtained from the trajectories of simulation runs 2, 3, and 6. (A) Two-dimensional ligand-RMSd (Å) and binding energy (ΔE_bind, kcal/mol) distributions. Each dot represents a spatial arrangement of the two solute molecules with the average ligand-RMSd and ΔE_bind values obtained from a 1 ns MD trajectory. The labels a, b, and c denote the bound poses. (B) Ligand-RMSd profiles with respect to the simulation time (ns). (C) Representative complex structures of bound poses a and b.

As mentioned previously, the structure of hevein is distinguished by rich in conserved cysteine and glycine residues, which are a common structural motif observed in hevein domains². Hevein involves five disulfide bonds to secure stability of the three-dimensional structure, whereas glycine residues may contribute flexibility to the peptide backbone. Fig. S1 shows a hevein backbone-RMSd profile with respect to the simulation time obtained from the final 5000 ns MD trajectory of the simulation run 3. The observed complex NMR structure¹⁰ was used as a reference structure. To emphasize RMSd variations, the residues subject to the RMSd calculations were those belonging to the amino acid sequence between S19 and Y30. When comparing the RMSd profiles between the state of the bound pose a and an unbound state, a slight reduction in RMSd values and the absence of distinct pulse peaks are observed in the bound state. About 200 ns after the time of the complex formation of bound pose b, the RMSd values increased discontinuously by about 0.25 Å. This indicates that the excessive binding energy occurring in the complex structure of bound pose b led to local steric distortions of the peptide backbone, which could increase the whole steric energy of the hevein structure.

The three ligand docking processes leading to the complex structures of bound pose a were visually examined by inspecting the MD trajectories taken from simulation runs 2, 3, and 6. In the trajectory from 657 to 664 ns of simulation run 2 (Movie S1), the GlcNAc₃ molecule showed a multistage process as it bound to the binding site. When it first approached the binding site, the GlcNAc₃ molecule appeared to climb up the side chain of Y30 as if it were a handhold. The GlcNAc₃ molecule first attached to the side chain of W21 and then made a combined movement of flipping and turning on the aromatic side chains to fit into the binding site with bound pose a. In the trajectory from 3316 to 3325 ns of simulation run 6, the GlcNAc₃ molecule directly reached the binding site and remained on the upper edge of the aromatic side chains, where the aromatic rings slightly overlapped for a few nanoseconds. The GlcNAc₃ molecule then docked into the binding site in concert with the elimination of the overlapping part to better interact with the binding surface (Movie S2). The ligand docking process in the trajectory from 9677 to 9695 ns of simulation run 3 was similar to that observed in simulation run 6. The GlcNAc₃ molecule bound to the binding site from the top edge of the aromatic side chains in a rather straightforward manner (Movie S3). The ligand docking motion shown in the three movies suggests that the surface topology formed by the two aromatic side chains W21 and W23 primarily regulates a GlcNAc₃ molecule to bind in a proper fashion, especially in terms of its direction with respect to the order of the subsites. The ligand docking motion shown in Movies M2 and M3 suggested that the intermolecular hydrogen bonds involving the hydroxyl groups of the S19 and Y30 side chains and the acetamide moiety of the nonreducing end residue may play a role in reinforcing the complex structure after the GlcNAc₃ molecule correctly binds to the binding site.

Our initial unguided MD simulation study of the hevein–GlcNAc₃ binary solution system using the original nonbonded interaction parameters was affected by the excessively high affinity between the sugar and amino acid residues. In all of the six simulation runs, the GlcNAc₃ molecule irreversibly bound to the binding site in the course of 1 µs simulations only to provide the undesired bound pose b with RMSd values of approximately 5 Å (see Fig. S2). From Fig. 3C, the ligand binding structure of bound pose b appears to involve a larger binding interface than that of bound pose a, which is consistent with lower ΔE_bind values of the complex structures of bound pose b, regardless of the nonbonded interaction parameters used. Clearly, the solvent effect is an additional effect that makes the complex structure of bound pose a the most stable. Table S1 lists the thermodynamic values calculated from the MD trajectory frames involving the bound poses a and b complex structures. The average values of ΔE_bind for bound pose a and b calculated from the corresponding trajectory frames of simulation runs 2, 3, and 6 were − 11.8 ± 2.4 kcal/mol and − 26.5 ± 4.0 kcal/mol, respectively. Aboitiz et al. reported the van’t Hoff enthalpy of -8.7 kcal/mol for the hevein–GlcNAc₃ complex formation determined by the NMR titration measurement, in addition to the observed binding free energy of -5.5 kcal/mol¹⁰. However, the ΔG_bind^PBSA calculations provided undesirable positive values, suggesting that the modified van der Waals parameters may not have worked properly on the MM_PBSA.py module. On the other hand, the MD simulations using the original van der Waals parameters for the NMR complex structure suggested an excessively low ΔE_bind value of -51.4 kcal/mol, accompanied by the apparently reasonable ΔG_bind^PBSA value of -22.8 kcal/mol. It should be noted that in the present MD simulations the evaluation of the thermodynamic values, in combination with the RMSd evaluations of the ligand positions, was rather adopted as a means of monitoring complex formation.

Ligand-docking behavior of the binary solution containing hevein and GlcNAc₆ molecules

Originally, hevein binds to the macromolecular chitin that composes the fungal cell wall. Therefore, the question arises as to whether the ligand docking behavior observed for the GlcNAc₃ binary solutions is inherently different from that for polymeric substrates. Unguided MD simulations of a binary solution containing a hevein molecule and a GlcNAc₆ molecule were performed in a similar fashion to obtain a more realistic picture of the ligand docking behavior. Compared with the docking behavior observed for the hevein–GlcNAc₃ binary solutions, the GlcNAc₆ molecule more effectively bound to the binding site in four out of the six simulation runs during a simulation time of 5 µs, as indicated by the two-dimensional ligand-RMSd and ΔE_bind distributions (Fig. 4A) and the time course of the ligand-RMSd value (Fig. 4B). Three bound poses were detected, all showing an identical direction of the GlcNAc₆ molecule to that of the observed hevein–GlcNAc₃ complex structure (Fig. 4C). All of the three subsites, + 1 to + 3, were occupied by GlcNAc residues in bound pose a. Note that the RMSd values, using the observed hevein–GlcNAc₃ complex as the reference structure, reflect the deviations in the atomic coordinates of the GlcNAc residues accommodated at the three subsites from those in the observed structure. The bound poses with one vacant subsite, bound poses b and c, tended to have relatively large RMSd values. Similar to the ligand-docking behavior observed for the hevein–GlcNAc₃ binary solutions, neither transient ligand binding to the protein surface other than to the binding site nor a specific intermediate complex structure was observed. It appeared that the complex structures of bound pose b were the most stable, as suggested by their long lifetimes of more than 99% and 60% of the simulation time in simulation runs 2 and 5, respectively, accompanied by ΔE_bind values of mostly approximately − 40 kcal/mol. The GlcNAc₆ molecule unevenly bound to the binding site in all three bound poses with either the reducing or nonreducing moiety of the GlcNAc₆ molecule exposed to the bulk water (Fig. 4C). It is likely that the solvent effect acting on the unbound three to four residue moiety of the GlcNAc₆ molecule promotes complex formation.

Fig. 4.

Ligand docking behavior of the binary solution containing hevein and GlcNAc₆ molecules obtained from the MD trajectories of simulation runs 2, 3, 4, and 5. (A) Two-dimensional ligand-RMSd (Å) and binding energy (ΔE_bind, kcal/mol) distributions. Each dot represents a spatial arrangement of the two solute molecules with the average ligand-RMSd and ΔE_bind values obtained from a 1 ns MD trajectory. The labels a, b, and c denote the bound poses. (B) Ligand-RMSd profiles with respect to the simulation time (ns). (C) Representative structures of bound poses a, b, and c.

Competitive ligand-docking behavior of a ternary solution containing hevein, GlcNAc₆, and Glc₆ molecules

The present unguided MD simulation method for a binary solution system containing a ligand molecule and its target protein was originally proposed for identifying an unknown binding site and predicting a complex structure¹⁹. As an extended application of the unguided MD simulation method, we propose using a ternary solution system containing hevein, an oligosaccharide substrate, and a substrate analog to compare their relative affinities for the binding site, where the two ligand molecules are expected to competitively bind to the binding site. The oligosaccharide ligands used were GlcNAc₆ and Glc₆ molecules, the latter playing the role of a substrate analog molecule. A hevein–GlcNAc₆ complex successfully formed in two out of the six simulation runs, while the Glc₆ molecule remained circling around the hevein molecule throughout the 5 µs simulation time in all of the six simulation runs (Fig. 5A and B). In the two hevein–GlcNAc₆ complex structures, all three subsites were occupied by GlcNAc residues (Fig. 5C). The bound pose b corresponds to the bound pose a obtained in the previous MD simulations of the GlcNAc₆ binary solution system. The position of the GlcNAc residue at the subsite + 3 slightly deviated from that of the corresponding residue in the reference complex structure, resulting in larger RMSd values. The nonreducing residue of the GlcNAc₆ molecule shifted from the subsite + 1 to the subsite − 1 by changing the ligand-binding mode from bound pose b to bound pose a. We did not anticipate that the competitive MD simulations would be able to distinguish the two oligosaccharide molecules so clearly, owing to the limited simulation system. Interestingly, the complex structures with the GlcNAc₆ molecules were retained for longer simulation times (up to a few microseconds) than those in the GlcNAc₆ binary solutions (hundreds of nanoseconds), which was obviously caused by the double effective concentration of the GlcNAc₆ molecule in the ternary solution. It should be noted that both of the oligosaccharide molecules with the common β-1→4 glycosidic linkage prefer to have a similar extended overall structure, even in the solution state. Therefore, the competitive MD simulations again confirmed that the 2-acetamide groups significantly contribute to the substrate specificity of the hevein binding site. This group may act as a guide to lead a chito-oligosaccharide to the hevein binding site, in addition to forming intermolecular hydrogen bonds with the hydroxyl groups of the S19 and Y30 side chains¹⁰, probably mediated by water molecules.

Fig. 5.

Competitive docking behavior of the ternary solution containing hevein, GlcNAc₆, and Glc₆ molecules obtained from the MD trajectories of simulation runs 2 and 3. (A) Two-dimensional ligand-RMSd (Å) and binding energy (ΔE_bind, kcal/mol) distributions. Each dot represents a spatial arrangement of the two solute molecules, hevein and GlcNAc₆, with the average ligand-RMSd and ΔE_bind values obtained from a 1 ns MD trajectory. The labels a and b denote the bound poses of the hevein and GlcNAc₆ complex structures. (B) Ligand-RMSd profiles with respect to the simulation time (ns). (C) Representative structures of bound poses a and b, accompanied by a Glc₆ molecule.

Conclusions

The present unguided MD simulations of a binary solution system containing GlcNAc₃ and hevein molecules successfully predicted the complex structure of the observed bound pose¹⁰ as the most stable form. The MD simulations also gave alternative bound pose b in which the GlcNAc₃ molecule, shifted by one residue, was accommodated in the subsites − 1 to + 2 with the same orientation. Using weakened nonbonded interactions with modified van der Waals parameters, combined with the solvent effect, allowed the GlcNAc₃ molecule to reversibly bind to the binding site for several hundred nanoseconds. In the hevein–GlcNAc₆ binary solution systems, the GlcNAc₆ molecule more effectively but unevenly bound to the binding site. The terminal moiety of the GlcNAc₆ molecule was exposed to the bulk water, suggesting that the nonbonded interactions at the binding interface and the solvent effect on the unbound GlcNAc residues were balanced to form a stable complex structure. In our previous docking MD simulations using the original van der Waals parameters, in some simulation runs, a chito-oligosaccharide molecule bound to the hevein surface other than to the binding site. Such an undesirable complex structure was not observed in the present MD simulations, indicating that using the modified van der Waals parameters allowed the chito-oligosaccharide molecule to successfully identify the binding site from the rest of the protein surface.

The docking processes of a chito-oligosaccharide molecule were evaluated by visualizing the MD trajectories. After moving around the hevein molecule, the chito-oligosaccharide molecule rapidly approached the binding site from a different direction in each simulation run, without a specific hot spot where the ligand molecule was temporarily bound before moving to the binding site. It appeared that the chito-oligosaccharide molecule approaching the binding site was oriented in the appropriate bound pose on its own accord, although it was in fact mainly guided by the surface topology of the aromatic side chains W21 and W23. While intermolecular hydrogen bonds with the side chains of S19 and Y30 are then formed to secure the complex structure, as mentioned above, surrounding water may assist in guiding the GlcNAc₃ molecule to the binding site and stabilize the final complex structure.

We also performed competitive docking simulations for a ternary solution containing hevein and two different oligosaccharide substrates, GlcNAc₆ and Glc₆ molecules, as an extended application of the unguided MD simulation method. To our surprise, hevein clearly distinguished the GlcNAc₆ molecule from the Glc₆ molecule, yielding exclusively a hevein–GlcNAc₆ complex. In addition to participating in the formation of the intermolecular hydrogen bonds, the acetamide groups, which are bulkier than the hydroxyl groups of a glucose residue, may act as a detector to explore the binding site and its vicinity on the protein surface. However, we are reluctant to conclude that the binding site of hevein has no affinity for the Glc₆ molecule at all, considering that the present MD simulations were performed for limited simulation time on the order of nanoseconds or microseconds, a short time compared with the real time scale. One of the drawbacks of the unguided MD simulation method is that it is essentially impossible to rule out complex formation for a given ligand. When performing a similar competitive docking simulation for hybrid substrates with closer affinities for a target protein, the simulation conditions, such as the solute concentration and simulation time, must be carefully examined; otherwise, a false conclusion could be drawn from an accidental result.

The present docking MD simulations were not intended to be an unbiased search for a ligand binding site of a target protein, contrary to the original objective of the unguided MD simulation method¹⁹. Because hevein is a small protein and its binding site has been defined, the purpose of the present MD simulation study was to confirm the reproducibility of the observed hevein–GlcNAc₃ complex structure, which was successfully achieved by using modified van der Waals parameters to weaken the nonbonding interactions between the GlcNAc₃ and hevein molecules. With this information, it is expected that a future MD study of the hevein complex using an advanced MD method, such as thermodynamic integration, will be performed with high reliability.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

T.Y. conceptualized the study, visualized the data, developed the methodology used, applied the software, performed the MD calculations, wrote the original draft, and reviewed and edited the manuscript. T.U. developed the force field parameters and developed the methodology used.

Funding

This research was partly supported by JSPS KAKENHI (Grant No. JP21K05187 to T.Y. and Grant No. 23K18510 to T.U.) of the Japan Society for the Promotion of Science (JSPS).

Data availability

The video files, Movie S1, S2, and S3, and the word file containing Fig. S1 and S2, and Table S1 are available in the Supplementary Material. The AMBER topology, input, and initial coordinates, and the PDB files of the starting structures used for the present study are also provided in the Supplementary Material.

Declarations

Competing interests

The authors declare no competing interests.