Exploration of structural space of Cus17 lectin with glycans using molecular docking and simulation studies

Abstract

The Cus17 phloem protein, in the case of Cucumis sativus species, plays an important role in the phloem-based defense of the plant. Cus17 can bind to various carbohydrates present on insect exoskeletons or fungal cells. The recent experimental structure of chitotriose bound Cus17 elucidates the carbohydrate interacting residues of Cus17. Higher chito-oligosaccharides are also known to interact with Cus17, but the lack of experimental structure impedes our understanding of their interaction. In this study, we have employed in-silico methods to explore the binding interactions of higher chito-oligosaccharides with Cus17. Chitoheptaose forms stable interactions with canonical binding site residues Trp48 and Asp50. Smaller chito-oligosaccharides were observed to be relatively unstable at the canonical binding site of Cus17. Further, the chito-oligosaccharides were inspected for interactions with predicted ligand binding sites. We also generated different tetramers of Cus17 and docked the chito-oligosaccharides to the tetrameric Cus17. All chito-oligosaccharides were found to make persistent interactions with Cus17 tetramer. Interestingly, chitotriose shows the best binding affinity and maintains stable interactions with Cus17 tetramer upon extended simulations with the canonical site.

Graphical Abstract

1. Introduction

In higher plants, phloem is an indispensable part of the vascular network and whole plant communication system [1]. Phloem tissue is engaged in the transport of plant sap, which contains water, minerals, proteins, carbohydrates, mRNA, and small RNA. As plants are sessile organisms, they need to cope with a wide variety of stresses including injury caused by abiotic (wind, fire etc.) or biotic (stinging, pruning, browsing, chewing etc.) agents. As an immediate response to injury, plants have developed a defense mechanism called phloem-based defense (PBD) in which their sieve pores are plugged immediately by the disulphide cross-linked polymers of the structural protein-phloem protein 1 (PP1) with the dimers of phloem protein-2 (PP2) that act as a physical barrier against bacteria, viruses and fungi as well as help in impeding the loss of nutrients rich phloem sap by mass flow [2], [3], [4], [5], [6], [7]. The growth of fungal pathogens of cucurbits, eg. Phytophthora infestans and Botrytis cinerea, is shown to be curtailed significantly by binding of PP2 protein of Cucumis sativus (PP2-A1; Cus 17) to fungal cell walls [2]. PBD also protects plants against sap sucking insects such as Myzus persicae and Apium cinera [4], [8]. Further, PP2-A1 of Arabidopsis thaliana retarded the growth of the nymphs of sap sucking insects A. pisum and M. persicae, presumably by binding with high affinity to N-linked glycoproteins abundantly expressed in their midgut [9], [10], [11]. Remarkably, the colonization of leaves of Arabidopsis thaliana by aphid M. persicae was dramatically diminished in plants over-expressing AtPP2-A1, as compared to its knocked out counterparts, because of the reduction in stylet puncturing of sieve elements by aphids and in their phloem feeding activity [8]. Treatments of plants by Harpin N_Ea, which is known to induce PBD increased expression of PP2, significantly suppressed the colonization by the aphid as compared to the wild type plants [8]. Further, PP2- A1 protein (Cus17) has been shown recently to increase aphid resistance by enhancing ROS homeostasis in cucumber plants [4].

Many PP2 proteins contain a heterogenous N-terminal domain exhibiting TIR or AIG or F box or SKPF motifs that are involved in the signalling of plant immune defences, development etc. [12], [13], [14] and a highly conserved lectin domain at their C terminus, which can recognize a wide variety of N-linked glycans, chitin and its degradation products, chito-oligosaccharides [15], [16], [17]. The lectin binding domain can also contain a tetralysine motif to facilitate its localization into the nucleus, as observed in Nictaba [13]. N-linked glycans act as cell surface markers for most cells, including those of pathogens and sap-sucking insects. Chitin is an integral component of insect exoskeletons and fungal cell walls. Moreover, chito-oligosaccharides have been known to elicit general immune responses in plants [18]. It is thus clear that PP2 could play a pivotal role in triggering the immune response of plants by sensing glycans associated with infection and damage.

An interesting property of PP2 is the protein’s ability to interact with a huge repertoire of glycans ranging from chito-oligosaccharides to a vast number of N-linked glycans, including hybrid, complex with and without fucosylation and or sialylation, and poly-N-acetyllactosamine chains of o-linked glycans [10]. Indeed, the activity of PP2 is intricately related to how the glycan ligands bind to PP2. The structure of dimeric PP2 from Cucumis sativus, Cus17, representing the archetype member of this family of proteins, was recently elucidated [10]. The study also utilized a glycan array to determine the binding specificity of 586 glycans. Subsequently, structures of the Cus17 dimer bound to the sugars LacNac and Chitotriose were also obtained, offering a deeper insight into the glycan binding site of Cus17. Further, detailed isothermal titration calorimetry (ITC) experiments from the study demonstrated that Cus17 exhibits the greatest affinity for Man3GlcNAc2 - the core component of N-linked glycans - followed by chitoheptaose, chitohexaose, chitotriose, chitopentaose, and chitotetraose respectively. This suggests an extended binding site on Cus17 that can accommodate three or more consecutive GlcNAc residues and Man3GlcNAc2, the motif common to N-linked glycans, to accommodate even larger structures [10]. A recent crystal structure of Nictaba, another member belonging to the PP2 family of proteins, reinforces the structural similarities and mode of glycan binding, although detailed ITC experiments are not reported in this study [19].

While we do have information regarding the binding mode of two different glycans to the Cus17 dimer, one cannot assume that binding modes for all the different carbohydrates recognized by the protein would be equivalent. However, the determination of the PP2 crystal structure was quite a challenge due to their heterogeneity, post-translational modifications, and their tendency to aggregate when isolated from natural sources [20]. As such, we overcome this challenge by employing an in-silico approach to investigate the potential binding modes of different chito-oligosaccharides and a number of glycans to the Cus17 PP2 dimer. Unbiased molecular docking and molecular dynamics simulations were performed to explore the structural space of glycan binding to PP2 complexes.

Moreover, as chito-oligosaccharides have repeating symmetric motifs, they are known to cross-link binding sites intra and intermolecularly. Studies in CIA24, an ivy gourd PP2-like lectin with homology to Cus17, have established its ability to form tetramers and have also hinted at the role of longer chito-oligosaccharides, particularly chitohexaose, in PP2 oligomerization [21], [22], [23]. So, we also discuss the possibility of the potential role of such multivalent glycan binding in dimer-dimer cross-linking at the interface region of an in-silico model of a Cus17 tetramer [24], [25]. Our findings suggest that the Cus17 PP2 dimer may harbor additional binding sites beyond the canonical one, with chito-oligosaccharides of varying lengths potentially favoring these alternative sites.

2. Methods

2.1. Protein and ligand preparation and binding site prediction

The protein–ligand interaction profiler (PLIP) tool was used to examine the interactions within the crystal structure of Cus17-chitotriose complex (PDB 7W4B) and Nictaba-chitotriose complex (PDB 8AD2) [10], [19], [26]. The protein structure (PDB 7VS6)[10] was prepared using the protein preparation wizard of Schrodinger [27]. The pH 7.4 was used for the protonation of protein structure. Missing residues were filled, and termini were capped (N-acetyl group was added at the N-terminus, and N-methyl amide was added at the C-terminus). H-bond optimization was performed followed by restrained structure minimization. The structure of chitotriose was extracted from the PDB structure of chitotriose bound to Cus17 dimer (PDB 7W4B)[10]. SDF files for the higher chito-oligosaccharides - chitotetraose (CID 10985690), chitopentaose (CID 14055040), chitohexaose (CID 6918014), and chitoheptaose (CID 14605620) - were downloaded from PubChem. Chitotriose, chitotetraose, and chitohexaose structures contained 3, 4, and 6 repeating units of N-acetylglucosamine (GlcNAc), whereas chitopentaose and chitoheptaose comprised 5 and 7 repeating units of glucosamine, respectively. The ligands were prepared at the same pH using the ligprep module of Schrodinger. The possible binding sites were predicted using the sitemap tool [24], [25]. The residues that constitute each of the predicted binding sites are listed in Supplementary Table 4.

2.2. Molecular docking

The prepared protein was used for receptor grid generation. Aromatic hydrogens were considered as donors and all the rotatable groups were checked for the receptor grid generation. The hydroxyl groups of residues Tyr, Ser, and Thr, along with the thiol group of Cys, are allowed to rotate, allowing them to adopt different orientations upon docking. The Ser, Thr and Cys can orient in three local minima, while the Tyr hydroxyl can orient in two local minima. To define the canonical site, the ligand-bound dimer crystal structure (PDB ID: 7W4B) was superimposed onto the prepared Cus17 dimer using the protein backbone atoms. The ligand from the superimposed structure was then used to generate the receptor grid for the canonical binding site. The grid was centered on the ligand, with the docking ligand length set to 36 Å. Additionally, the ligand diameter midpoint box size was set to 15 Å, such that the box is big enough to accommodate the larger ligands in the study. Separate receptor grids were also generated for predicted binding sites 1 and 2, as identified using the Sitemap tool. These sites, along with their defining residues, are listed in Supplementary Table 4.

Protein-ligand docking was performed using Glide’s XP protocol [28]. Options for reward intramolecular H-bonds and enhance planarity of conjugated pi groups were considered for the docking. Aromatic hydrogens were also included as H-bond donors with partial charge cutoff of 0.1. The option for strain correction was included for the docking score calculation. The same docking protocol was followed for docking at each of the generated receptor grids. The prepared ligands were used for this docking procedure. The binding energy (ΔG) was calculated for the docked complex using the prime MM-GBSA tool of Schrodinger. Residues within 5 Å were made flexible along with a constraint on flexible residues.

Blind docking experiments were carried out using AutoDock. In this case the protein and ligands were prepared using AutoDock Tools (ADT) [29]. The grid box was constructed to include the entire protein, and the same grid box was used for docking all the ligands. Grid parameter file and docking parameter file were made using ADT with default parameters, except for the number of GA runs (set to 100) in the docking parameters, and Lamarckian genetic algorithm was used. Hundred blind docking poses were generated and the docked poses were categorized into canonical site and other predicted sites as follows: If residues within 4 Å of a given ligand pose are part of any site, then that pose is categorized into that site. Threshold for number of common residues within 4 Å of ligand pose and residues of that site are set as:

$$\mathrm{Threshold}=(\mathrm{number}\mathrm{of}\mathrm{nearest}\mathrm{residues}/2)+1$$

Where, number of nearest residues represent number of residues within 4 Å.

2.3. Molecular dynamics simulation

The protein-ligand complexes were subjected to all-atom Molecular Dynamics (MD) simulation in an explicit aqueous environment, using Desmond, to assess their interaction stability [30]. The complexes were prepared in a simulation box using the system builder tool of Schrodinger. The TIP3P water model was used for solvation. The system was neutralized by adding counter ions and default ionic concentration (0.15 M NaCl) was used. The prepared system was subjected to 100 ns of production MD simulation at 298.15 K temperature in the NPT ensemble. OPLS4 force field was used and default relaxation protocol was run before the production run. The default relaxation protocol begins with energy minimization performed using the steepest descent method for a maximum of 2000 steps. This process is carried out both with solute restraints, applying a force constant of 50 kcal/mol/Å on all solute atoms, and without restraints. Following energy minimization, a series of short molecular dynamics (MD) simulations are conducted: (1) 12-picosecond simulation at 10 K under the NVT ensemble, employing the Berendsen thermostat with restraints on non-hydrogen solute atoms, (2) 12-picosecond simulation at 10 K and 1 atm pressure under the NPT ensemble, using both the Berendsen thermostat and barostat with non-hydrogen solute atom restraints, (3) 24-picosecond simulation at 300 K and 1 atm pressure under the NPT ensemble, using the Berendsen thermostat and barostat while maintaining non-hydrogen solute atom restraints, and (4) 24-picosecond simulation at 300 K and 1 atm pressure under the NPT ensemble, using the Berendsen thermostat and barostat without any restraints.

MD simulation results were analyzed using simulation interaction diagrams (SID). The structure at 0th ns was used as the reference frame for calculating root mean square deviation (RMSD). The RMSD of ligand heavy atoms is calculated after the protein-ligand complex is aligned to the reference frame. The higher values of ligand-RMSD than protein-RMSD indicates the diffusion of ligands away from its initial binding site. Along with RMSD, the RMSF (Root Mean Square Fluctuation) and RoG (Radius of Gyration) of the protein and ligands were calculated for each MD run. Protein RMSF shows the fluctuation of the protein backbone for each residue, while ligand RMSF shows the fluctuation of each heavy atom in the ligand. RoG indicates the compactness of the protein and it was also monitored across the simulation trajectory. Increasing values of RoG can indicate the unfolding of protein or large scale conformational change. Non-covalent interactions (H-bond, Hydrophobic interaction, ionic interaction and water bridges) between protein and ligand were monitored throughout the simulation time. Hydrophobic interaction includes Cation -Π and Π- Π interactions. The calculated raw data was plotted using custom python scripts. Post MD simulation, structural clustering of all protein-ligand conformations throughout the trajectory was performed to find the binding energy of the conformation most adopted during the MD simulation. For this, the largest cluster was selected and ΔG of binding for the cluster representative of the selected cluster was then calculated using MMGBSA. The ΔG was also calculated for conformations at every 2 ns interval, starting at 1 ns, using the python script “thermal_mmgbsa.py”.

2.4. Unbinding event analysis

Unbinding kinetics were quantified by two complementary metrics applied to the simulation trajectories (frame interval = 0.1 ns). First, contact-based retention time was determined by monitoring the number of interatomic contacts between the ligand and binding-site residues using a 4.5 Å cutoff (via Schrodinger’s “trajectory_asl_monitor.py” script). For the canonical binding site, the analysis was restricted to residues 17, 18, 19, 20, 21, 25, 46, 47, 48, 56, 63, 79, 80, 139, 140, 141, and 143 of the relevant chain. For the predicted sites (site 1 and site 2), binding-site residues corresponded to those listed in Supplementary Table 4.

The retention time was defined as the first time point after which no atomic contacts persisted for at least 50 consecutive frames (equivalent to 5 ns), to account for transient fluctuations. Second, we computed the percentage retention, defined as the fraction of total frames in which the ligand maintained at least one contact with the specified binding-site residues.

2.5. Protein-protein docking

Cus17 tetramers were generated through protein-protein docking of Cus17 dimers using HDOCK [31]. Top 100 models were evaluated for protein-protein interaction strength using our in-house tool PPCheck [32], [33]. Top two tetramer models were used for blind docking experiments.

3. Results

3.1. The asymmetry of chitotriose binding to protomers of Cus17 in crystal structure

Initially, the crystal structure of the Cus17-chitotriose complex was examined for the interactions of chitotriose. There are subtle differences in the interactions between chitotriose with each protomer of Cus17 in the crystal structure (Supplementary Table 1). One of the chitotriose molecules forms more H-bond interactions with its corresponding protomer, in comparison to the other molecule. The H-bond interaction of chitotriose with Trp19 is present in only one protomer. The rotamers of binding site residues in both protomers were found to be very similar, however, the conformations of the bound chitotriose ligands are not identical (Figs. 1a, 1b).

Fig. 1.

MD simulation of Cus17-chitotriose complex from the crystal structure. (a) Comparison of side chain conformations at the chitotriose binding site in the first [cyan] and second [green] protomers. (b) Comparison of the conformation of chitotriose ligand bound to the first [magenta] and second [lilac] protomer (c, f) RMSD of Cus17 and chitotriose ligand bound to first protomer (c) and to the second protomer (f) (d, g) Interactions of chitotriose with Cus17. Persistence of each interaction in terms of simulation time has been indicated on the Y-axis. (e, h) Total number of interactions at each time point.

To further probe these differences, MD simulation was employed to investigate the dynamics of chitotriose interactions with Cus17. Interactions of chitotriose with Cus17 were monitored across the simulation trajectory, and it was observed that chitotriose does not form stable interactions with Cus17 (Fig. 1). H-bond interactions are the major type of interactions between chitotriose and Cus17, with residues Trp19, Ser47 and Trp48 forming H-bonds which persist for less than 60 % of simulation time. There are hydrophobic interactions as well with residues Trp19, Trp48 and Trp141. The other protomer (protomer B) involves the H-bond interaction of chitotriose with Ser127, Asp93 and Glu98, which persists for less than 20 % simulation time, though this interaction is absent in the first protomer (Fig. 1g). Both molecules of chitotriose were observed to lose interaction with Cus17 at the end of the simulation (Figs. 1e, 1h). The simulation was performed in three replicates with different initial seeds. In one of the runs, both the molecules of chitotriose were observed to leave the binding site completely and diffuse away (Supplementary Figure 1) while in another, one of the two chitotriose molecules left the native binding site during the simulation and the other molecule lost its interaction with Cus17 at the end of the simulation (Supplementary Figure 2). The mean retention time for chitotriose bound to the first and second protomers was 62 ± 44.8 ns and 207 ± 118.1 ns, respectively (Supplementary Table 9). Since a difference was observed between the two molecules of chitotriose bound to Cus17 in the crystal structure, an MD simulation was also carried out after swapping the two chitotriose poses. It was observed that there are persistent differences in the interaction of chitotriose with each protomer. Interacting residues of each protomer are similar, but the persistence of their interaction varies (Supplementary Figure 3, Supplementary Table 9). The simulation with swapped chitotriose was also repeated with different random seed numbers and it was observed that the swapped complex is relatively more stable than the native complex (Supplementary Figure 4). In case of the second run, both ligands remain in contact with the binding site residues throughout the 500 ns simulation (Supplementary Table 9). To make sure that this difference in interaction is not an artifact of protein dimer instability, we also monitored the residue RMSF and radius of gyration (RoG) of the Cus17 dimer. The Cus17 residues do not show large RMSF across replicates and it is less than 3.5 Å for all residues (Supplementary Figure 5). The Cus17 dimer shows small conformational variability across the simulation, but it does not undergo dissociation as assessed through RoG calculation (Supplementary Figure 5). Together, these results indicate that the starting conformation of chitotriose can affect its interaction stability with Cus17.

3.2. Interaction of higher chito-compounds with Cus17 canonical binding site

As mentioned in the previous section and due to inherent challenges in structure determination of Cus17 complexes, there are no crystal structures available for the complex of higher chito-oligosaccharides (chitotetraose, chitopentaose, chitohexaose and chitoheptaose) with Cus17. Therefore, to gain structural insights into the interaction of these higher chito-oligosaccharides with Cus17, computational docking experiments were conducted. Initially, the chitotriose binding site (canonical binding site) from the crystal structure was utilized for docking using the extra precision (XP) docking protocol of Glide [28]. Additionally, chitotriose was docked to assess how closely the predicted binding pose aligns with the crystal structure. While the major interacting residues remained largely the same, with the exception of Gly139 in the crystal structure being replaced by His137 in the docked pose, a conformational difference was observed between the docked and crystal structure ligands (Supplementary Figures 6, 7a). Chitoheptaose was observed to have the best docking score which agrees with the experimental binding studies (Supplementary Table 2). The docked poses were used for ΔG calculation using MMGBSA. However, here the binding energy of chitotriose is −62.48 kcal/mol, while that of chitoheptaose is −52.68 kcal/mol (Supplementary Table 2). The higher chito oligosaccharides do not bind in a similar fashion (Fig. 2). Chitotriose forms H-bonds with Ser47 and Trp48, while chitoheptaose interacts with many residues through H-bonds in the docked pose (Supplementary Figure 7).

Fig. 2.

The best docking pose of chito-oligosaccharides at the canonical site of Cus17. Chito-oligosaccharides are represented as spheres (Pink: Chitotriose, Blue: Chitotetraose, Orange: Chitopentaose, Yellow: Chitohexaose, Red: Chitoheptaose).

Initially, the MD simulations were performed for 100 ns for all the complexes, and a summary of the results is presented in Table 1. It was observed that chitoheptaose remains stable for the initial 100 ns and makes persistent H-bond interactions with Trp48 and Asp50 (Fig. 3). Many water bridge interactions were also observed. Other compounds, except Man9, were found to lose initial contact with Cus17 upon simulation (Supplementary Figures 9–17). The chitotriose dissociated from the canonical binding site after 58.20 ns and remained in contact with binding-site residues for 67.43% of the trajectory (Supplementary Table 9), indicating a short residence time. Given that this early unbinding of chitotriose was observed in simulations starting from both the holo and apo crystal structures, whose binding site conformations are very similar (Supplementary Figure 18a), we wanted to ascertain that the observed instability reflects an intrinsic feature of the ligand, rather than a bias introduced by the protein conformation. To rigorously rule out the possibility that hidden conformational dynamics of the protein influenced ligand behaviour, we simulated the ligand-free apo dimer for 500 ns and extracted five representative conformations at 100-ns intervals. The RMSD, RoG and RMSF plots for the simulation were also monitored and no significant global structural changes were observed (Supplementary Figures 18b, 18c, 18d). Chitotriose was docked onto each of these conformations at the canonical site (Supplementary Figure 18e). The best-scoring pose in each case was subjected to a new 100-ns MD simulation (Supplementary Table 3; Supplementary Figures 19, 20, 21). As such, chitotriose leaves the binding site in three out of five of these simulations, each time within 25 ns (Supplementary Table 9). These results reinforce that the observed instability is a consistent feature of chitotriose–Cus17 interactions at the canonical site.

Table 1.

Summary Table - Binding affinity of chito-oligosaccharides at the canonical binding site. The results correspond to 100 ns of MD simulations for each ligand-Cus17 dimer complex.

Compound	ΔG of docked pose (kcal/mol)	ΔG of cluster representative from MD (kcal/mol)	Average ΔG from MD simulation (kcal/mol)
Chitoheptaose	−52.68	−51.24	−52.17
Chitohexaose	−49.28	−42.18	−43.98
Chitopentaose	−65.59	NS	NS
Chitotetraose	−57.04	NS	NS
Chitotriose	−62.48	NS	NS

NS: Not Stable. These ligands leave the binding site upon simulation and therefore the ΔG was not calculated for these compounds.

Fig. 3.

MD simulation of Cus17-chitoheptaose complex at canonical site (a) RMSD of Cus17 dimer (green) and chitoheptaose (red). (b) 2D structure of Chitoheptaose (c) Interactions of chitoheptaose with Cus17. Persistence of each interaction in terms of simulation time has been indicated on Y-axis (d) Total number of interactions at each time point.

Chitotetraose and chitopentaose were also found to be very unstable in their interaction with Cus17 at the canonical binding site, while chitohexaose maintains interactions with Cus17 at all time points along the simulation. Chitohexaose interacts with Lys96, through H-bonds, which does not interact with chitoheptaose (Supplementary Figure 12, Fig. 3c). The MD trajectories were then clustered to get the most persistent structure and the binding energy of all chito-oligosaccharides was again calculated for the cluster representative of the largest cluster. Chitoheptaose and chitohexaose were found to have the best binding energy for the cluster representative compared to chitotriose (Table 1). To further investigate the stability of chitohexaose and chitoheptaose at the canonical binding site, we extended their MD simulations to 500 ns. Both ligands eventually dissociated from the site, with chitohexaose unbinding at 204.5 ns and chitoheptaose much later, at 478.2 ns (Supplementary Table 9). These results are consistent with experimental trends, where chitoheptaose exhibits higher binding affinity than chitohexaose, and both bind more strongly than the shorter chito-oligosaccharides [10].

Some of the N-linked mannooligosaccharides are experimentally known to be the strong binders of Cus17. These compounds were found to have strong bonding affinity in the docking experiment as well. Indeed, we observed through docking at the canonical binding site that Man9 has a binding energy of −97.38 kcal/mol. MD simulation of Cus17-Man9 complex shows the interaction stability of Man9 through multiple H-bond interactions (Supplementary Figure 13). Though detailed studies on N-linked mannooligosaccharides are not addressed here, a summary of the docking results for N-linked mannooligosaccharides is given in Supplementary Table 2. Similarly, some compounds are experimentally known to be non-binders of Cus17, like lactose and cellobiose. Our molecular docking protocol, followed by simulation, recapitulated this behavior. The docked complexes of lactose and cellobiose exhibited pronounced instability during simulation, as indicated by high ligand RMSD values and short contact retention times, each under 12 ns (Supplementary Table 9; Supplementary Figures 14, 15).

3.3. Interaction of chito-compounds with Cus17 at non-canonical binding sites

It is possible that the different chito-oligosaccharides bind at different sites on Cus17. Hence, to address this possibility, other ligand binding sites were explored for the Cus17 dimer followed by docking of all compounds to those sites. The exploration of structural space, like the presence of cavities by Schrodinger suite’s SiteMap tool, suggested alternate binding sites (please see Fig. 4a). Five binding sites were predicted and each of the binding sites has some overlapping residues with that of the canonical chitotriose binding site but provide different protein surfaces (Fig. 4a). The residues that make up each predicted site are summarized in Supplementary Table 4. Chito-oligosaccharides and Mannoligosaccharides were then docked onto these predicted sites using XP docking. The docked compounds were re-ranked based on the ΔG calculation using MMGBSA, and their binding energy results have been summarized in Supplementary Table 5. The detailed interactions of each of these compounds at both sites have been shown as ligand interaction diagrams (Supplementary Figures 22–25).

Fig. 4.

Binding of chito-compounds to non-canonical sites on the Cus17 dimer. (a) Non-canonical binding sites predicted by SiteMap. (b) Different docked poses of chitotriose generated using blind docking with AutoDock. (c) Binding pose of chitotetraose at the predicted sites 1 and 2 on the Cus17 dimer.

Surprisingly, our docking studies suggested that the non-canonical binding sites were more favorable for all the compounds (Supplementary Table 5). To determine whether this trend extends to other PP2 lectins, SiteMap analysis was applied to the Nictaba dimer, the only other PP2 homolog with a resolved crystal structure [19]. This analysis revealed three additional non-canonical binding sites with SiteScores greater than 0.8 (Supplementary Figure 26a). Despite differences in the dimerization patterns of Cus17 and Nictaba, chitotriose exhibited similar or even stronger predicted binding affinities at sites 1 and 2 compared to the canonical site (Supplementary Figures 26b–e). These findings suggest that non-canonical binding sites may play some role in ligand recognition across PP2 lectins, warranting further investigation.

Among the predicted binding sites on the Cus17 dimer, sites 1 and 2 were selected for further analysis based on their high Dscore (>1) and large volume, which suggest a favorable binding environment (Supplementary Table 4). Additionally, these sites extend across both chains of the dimer, enabling chito-oligosaccharides to adopt multivalent binding poses. Notably, chitotetraose exhibited the highest binding energy at both sites (Fig. 4c). To assess the stability of these interactions, initially, 100 ns MD simulations were performed for compounds docked at sites 1 and 2 (Supplementary Table 6). The complexes which did not experience unbinding within these 100 ns were extended to 500 ns. At site 1, chitotriose, chitopentaose and chitoheptaose displayed instability during simulation, as indicated by ligand RMSDs and short contact retention times (Supplementary Table 9, Supplementary Figures 27–32). Chitotetraose and chitohexaose maintained continuous contact with Cus17 and chitotetraose had a better average ΔG of binding for the simulation (Supplementary Table 10, Supplementary Figures 28, 30). At site 2, multiple ligands - specifically chitotriose, chitotetraose, chitohexaose and chitoheptaose - showed stable interactions for the first 100 ns. Their corresponding MDs were thus extended to 500 ns (Supplementary Figures 33–38). Chitohexaose demonstrated the highest interaction stability among the chito-oligosaccharides, maintaining contact throughout the simulation and with the best average ΔG of binding (Supplementary Table 9, 10). It formed persistent hydrogen bonds with Gln62 and Phe60 for over 90 % of the simulation time (Supplementary Figure 36). On the contrary, chitotriose, chitotetraose and chitoheptaose lose contact within less than 150 ns upon extended simulations. Chitopentaose failed to establish stable interactions, even for the first 100 ns, at either site (Supplementary Figures 29 and 35).

To further explore unbiased binding of chitotriose to Cus17, blind docking was carried out where the entire protein surface is sampled for molecular docking (Fig. 4b). The best docked pose of chitotriose binds at the dimer interface of Cus17, with a docking score of −6.35 kcal/mol. Further, the docked poses were categorized into canonical binding site poses and other predicted site poses. Most of the 100 poses belong to predicted site 1 or site 2 for all the compounds (Supplementary Table 7). Similarly, blind docking was also performed using the Nictaba dimer. The canonical site and predicted site 2 in Nictaba had an equal number of poses clustering near their site, comprising twenty percent of the generated poses when combined (Supplementary Table 7).

3.4. Binding of chito-oligosaccharides to Cus17 tetramer

As discussed in the previous section, studies on other PP2 lectins have demonstrated their ability to assemble into tetramers, suggesting a potential for similar oligomerization in Cus17 [25], [26], [27]. Additionally, repeating symmetric motifs in chito-oligosaccharides could mediate cross-linking between binding sites within and across molecules. With this in mind, the possibility of higher-order oligomerization in Cus17 was explored. For this, a Cus17 tetramer was modeled based on the chains C, D, E and F in the asymmetric unit of its crystal structure [10]. Chito-oligosaccharides were docked to this tetramer “CDEF” to investigate if the larger chito-oligosaccharides may bridge the two dimers by binding at the dimer-dimer interface of the Cus17 tetramer. Docking was performed by utilizing the canonical chitotriose binding site, at the dimer-dimer interface, where chitotriose was found to have the best binding energy.

Of all the chito-oligosaccharides, chitotriose displayed the most favorable ΔG of binding at the canonical site, near the dimer interface, of the CDEF tetramer (Supplementary Table 8, Fig. 5a). These results are interesting in the light of the hypothesis that larger chito-oligosaccharides could be favoured at the dimer-dimer interface. MD simulation of bound chitotriose to CDEF tetramer indicates the stability of chitotriose as it forms H-bonds with Trp19, Ser47 and Trp48 that persists for more than 90 % of simulation time (Fig. 5). Most of the interactions are with chain C residues. All the chito-oligosaccharides except chitoheptaose maintain stable interactions with the Cus17 during the first 100 ns of the simulations (Supplementary Figures 39–43). Chitotriose remains stable upon extension of simulation from 100 ns to 500 ns (Figs. 5b, 5c, 5d). Chitotetraose mainly interacts with Arg24 and Tyr25 of chain C, Gln62 and Asp63 of chain E and Trp19 of chain F, through H-bonds (Supplementary Figure 39). However, it leaves the binding site at 169.40 ns. In the case of the chitopentaose-CDEF tetramer complex, the tetramer itself was not very stable as per the protein RMSD and RoG (Supplementary Figure 40, 43c). However, chitopentaose maintains persistent interactions with Asp21 and Asp22 of chain C (Supplementary Figure 40). Chitohexaose interacts with residues of chain C and chain F (Supplementary Figure 41). It maintains contact with the binding site throughout the duration of the simulation and has the highest average ΔG of binding across the MD trajectory amongst all 5 chito-oligosaccharides (Supplementary Table 9, 10) In contrast, chitoheptaose leaves the binding site residues on chain C early on in the simulations and interacts mainly with residues of chain D, chain E and chain F (Supplementary Figure 42).

Fig. 5.

Docking and MD simulation of Cus17 CDEF tetramer-chitotriose complex (a) The docking pose of chitotriose (pink spheres) to the Cus17 tetramer (different colours represent each subunit) (b) RMSD of Cus17 and chitotriose (ligand) bound to Cus17-tetramer (c) Interactions of chitotriose with Cus17. Persistence of each interaction in terms of simulation time has been indicated on Y-axis (d) Total number of interactions at each time point.

4. Discussion

The recognition of carbohydrates by the members of PP2 gene family proteins is important for plant defence. Cus17 is a prototype of the PP2 family of proteins with a novel 3D structure fold and its interactions with chitotriose and N-acetyllactosamine have been elucidated through crystal structure determination. The binding of Cus17 with higher chito-oligosaccharides has been experimentally established but the lack of experimental structure of Cus17 in complex with these higher chito-oligosaccharides impedes our understanding in terms of binding of these compounds to Cus17. As such, computational modelling and molecular dynamics simulations were employed to overcome these challenges and gain insights into the binding of higher chito-oligosaccharides to Cus17. We considered the possibility of different chito-oligosaccharides binding at different sites on Cus17 using computational docking. In this study, we also explored binding of a variety of chito-oligosaccharides to Cus17 in its dimer and tetramer state.

ITC studies in the past have consistently reported one saccharide binding site per protomer in PP2 proteins, which have been validated by crystallographic studies on Cus17 and Nictaba recently. Hence, we have described here the interactions between the canonical site on Cus17 and different chito-oligosaccharides and mannans (Man3GlcNac2, Man5, Man7D1 and Man9) in detail.

Initially, the crystal structure of Cus17-chitotriose was inspected and was observed to have an asymmetric interaction between the chitotriose and each of the two protomers. A similar asymmetry is seen in the crystal structure of chitotriose bound to Nictaba, and the interacting residues at the binding site seem to be structurally equivalent [19] (Supplementary Table 11, Supplementary Figure 44). Interestingly, while the monomer structures of Nictaba and Cus17 are very similar (RMSD ∼1 Å), both proteins have starkly different dimerization interfaces.

The interaction stability of chitotriose was monitored through all-atom MD simulations and chitotriose was found to be unstable. The other chito-oligosaccharides were also docked at the canonical binding site and chitoheptaose was found to be the best compound for the canonical site, which agrees with the experimental binding studies. Chitoheptaose remains stable throughout most of the simulation time and majorly interacts with Trp48, Asp50, and His137. Docking and simulation experiments were carried out for experimentally proven strong binders like Man9 and negative control compounds like lactose and cellobiose. Man9 forms a stable complex upon simulation while the negative control compounds diffuse away from the binding site which strengthens our overall procedure.

Next, it was hypothesized that certain compounds might exhibit stronger binding affinity at alternative sites on the Cus17 dimer. To explore this possibility, potential ligand-binding sites were predicted, and all chito-oligosaccharides were docked at these sites. Molecular dynamics simulations were then performed to assess the stability of interactions at the two most promising sites. The results suggest that chito-oligosaccharides of varying lengths may exhibit site-specific binding preferences on the Cus17 surface.

Further, because of the reported proclivity of chito-oligosaccharide binding lectins from a diverse source to bind at the dimer-dimer interface, we explored the possibility of binding of larger chito-oligosaccharides to Cus17 tetramer [10], [20], [34], [35]. To this effect, chitohexaose showed stable binding at the canonical site on the tetramer throughout the simulation and had the best average ΔG of binding across the MD trajectory. Surprisingly, we observed that a smaller chito-oligosaccharide like chitotriose is also favored at the tetramer interface, both in terms of binding energy and its stability upon simulation. The residues involved in interaction with chitotriose are mainly from chain C, like Trp19, Ser47, Trp48 and Gly139, while it also interacts with Asp63 of E chain.

The additional binding sites uncovered on the surface of Cus17 dimer by means of these computational studies have been validated by an unbiased independent docking procedure (AutoDock) as well as by extension of such studies on Nictaba, the only homologue of Cus17 of which the structure has been reported [19] (Supplementary Figure 26, 44, Supplementary Table 7). This multi-site binding behavior is consistent with observations in other lectins. For example, snake gourd seed lectin (SGSL) crystallography and MD simulations revealed two sugar-binding sites per protomer, although only one remains stably occupied in dynamic conditions [36]. A mannose-binding jacalin-related lectin from pineapple (AcmJRL) also contains two carbohydrate-binding sites per monomer, corroborated by structural and biochemical data [37]. Similarly, BanLec, a tetrameric lectin, binds HIV glycans through bidentate interaction using multiple binding sites, with MD simulations identifying several distinct bidentate glycan-binding modes [38]. Since these sites have not been validated by an experimental method such as ITC it is difficult to speculate on their significance to the function of PP2 proteins. But, these non-canonical additional binding sites, should they be involved in any function, they might complement it by strengthening the glycan cross linking abilities of PP2 proteins to enhance their function [39].

4.1. Limitations of the study

The prediction of sugar binding pose has remained a difficult task as evident by recent CASP15 assessment of ligand binding to Nictaba protein [40]. In this study, we have performed docking and simulation of chito-oligosaccharides against the canonical binding site and predicted binding sites of Cus17. The higher chito-oligosaccharides have many rotatable bonds which impedes the exhaustive conformational search. Docking tools like Autodock limit the number of rotatable bonds to 32 and therefore higher chito-oligosaccharides cannot be made fully flexible for the blind docking.

5. Conclusions

The Cus17 protein recognizes a wide range of chito-oligosaccharides but the structures of Cus17 with higher chito-oligosaccharides are not available. In this study, we have utilized computational docking and simulation techniques to decipher the binding interaction of chito-oligosaccharides to Cus17. We observed that the chitotriose bound crystal structure of Cus17 is not very stable and chitoheptaose can bind to the same site with stable interactions with Cus17. Further, prediction of possible binding sites on Cus17 dimer surface and blind docking approaches revealed unexpectedly several additional binding sites which different chito-oligosaccharides prefer in terms of their interaction energy. These additional binding sites on the surface of Cus17 might be helpful in potentiating PP2 proteins’ defense function as multi-valency has been demonstrated to enhance the cross linking abilities of lectins [36]. Finally, we tested if the higher chito-oligosaccharides can be more stable upon docking to the Cus17 tetramer. Interestingly, chitotriose was also found to be very stable upon docking to the Cus17 tetramer, as it maintains its canonical interactions with the Cus17 upon extended simulation. This, in turn, could explain as to why in Cus17 and its homologous PP2 protein, Nictaba, chitotriose complexed with oligomeric states such as a dodecamer or a tetramer, respectively, only so far have been observed in the crystallographic studies [10], [19].

Author statements

Vikas Tiwari, Aditi Pathak and Tejaswini Poojary carried out all the modelling, docking, molecular dynamics simulations. Vikas Tiwari and Aditi Pathak performed the analysis and wrote the first draft of the manuscript. Ramanathan Sowdhamini and Avadhesha Surolia conceptualised the problem and looked into the final version of the manuscript.

CRediT authorship contribution statement

Avadhesha Surolia: Writing – review & editing, Supervision, Investigation, Conceptualization. Tejaswini Poojary: Investigation, Data curation. Vikas Tiwari: Writing – original draft, Visualization, Investigation, Data curation. Ramanathan Sowdhamini: Writing – review & editing, Supervision, Investigation, Conceptualization. Aditi Pathak: Writing – original draft, Visualization, Investigation, Data curation.

Declaration of Competing Interest

The authors declare no conflict of interest.

Appendix A. Supplementary material

Supplementary Table 1

Supplementary material