Molecular Docking and Dynamics Simulation Reveal Stable Binding of Tiliroside in the Colchicine Site of Tubulin

Abstract

Background:

The colchicine-binding site on tubulin is of particular interest for new drug development due to its role in microtubule destabilization and potential to overcome resistance to other agents. Tiliroside, a naturally occurring flavonoid glycoside, has demonstrated anticancer potential in vitro, but its interaction with tubulin has not been previously elucidated.

Objective:

This study aimed to investigate the binding of tiliroside to the tubulin site of colchicine, in comparison with colchicine.

Methods:

Molecular docking and molecular dynamics (MD) simulations were employed. Induced-fit docking predicted tiliroside binding, and redocking of colchicine was used to validate the docking protocol. MD simulations (100 ns) were conducted for both tubulin–tiliroside and tubulin–colchicine complexes.

Results:

Induced-fit docking predicted that tiliroside binds strongly in the colchicine site, with more favorable scoring metrics than colchicine (Glide GScore –16.77 vs –10.25, and MMGBSA ΔG_bind –50.46 vs –36.62 kcal/mol). Redocking of colchicine reproduced the binding pose (root-mean-square deviation (RMSD) ~0.6 Å). MD simulations further revealed that tiliroside forms a stable complex, remaining securely bound in the pocket. Tiliroside maintained multiple hydrogen bonds and hydrophobic contacts with tubulin, similar to or more persistent than those of colchicine.

Conclusion:

These results suggest that tiliroside can stably and snugly occupy the colchicine site of tubulin. In summary, our computational study provides structural and dynamic evidence that tiliroside is a high-affinity ligand for the colchicine site, supporting its potential as a lead compound for developing new tubulin-targeted anticancer agents.

INTRODUCTION

Microtubule‐targeting agents have been among the most successful chemotherapeutics due to their ability to disrupt mitosis in rapidly dividing cancer cells.[1] Tubulin, the dimeric protein building block of microtubules, offers multiple ligand binding sites for drug intervention, including the taxane site, the vinca alkaloid site, and the colchicine site.[1] Agents binding on the colchicine site on β-tubulin (colchicine-binding site inhibitors, CBSIs) prevent tubulin polymerization and induce microtubule depolymerization, leading to cell cycle arrest and apoptosis (McLoughlin and O’boyle, 2020).[1] Colchicine itself binds with high affinity at the interface of α- and β-tubulin, but its clinical use is limited by toxicity. Nonetheless, the colchicine site remains a compelling target for new anticancer drugs because inhibitors at this site can overcome multidrug resistance and have vascular-disrupting properties not shared by taxane-site agents.[1] For example, combretastatin A-4, a natural stilbene binding the colchicine site, has been developed into the prodrug fosbretabulin and entered clinical trials.[1] This demonstrates that natural products can inspire potent CBSIs. Additionally, numerous synthetic small-molecule scaffolds have been explored as tubulin inhibitors targeting the colchicine site. For instance, quinazoline and quinazolinone derivatives have shown excellent activity in inhibiting tubulin polymerization. Some of these compounds were even designed as dual inhibitors, concurrently targeting tubulin polymerization and other oncogenic pathways (e.g., EGFR signaling). These efforts underscore the diversity of chemotypes being investigated as colchicine-site agents in anticancer drug discovery, further supporting the relevance of this binding site.

Flavonoids are a broad class of natural polyphenols that exhibit diverse bioactivities, including antioxidant, anti-inflammatory, and anticancer effects. Tiliroside (kaempferol-3-O-(6’’-p-coumaroyl)-glucoside) is a flavonol glycoside found in many plants (such as Tilia species and berries) and has attracted attention for its cytotoxic and antitumor properties.[2] In particular, tiliroside has shown antiproliferative effects against several cancer cell lines and can modulate pathways related to apoptosis and drug resistance.[2] Given these promising biological activities, we hypothesized that tiliroside might exert anticancer effects by interacting with critical protein targets like tubulin. To date, however, the capacity of tiliroside to bind tubulin at the colchicine site has not been reported.

Here, we present a computational investigation of tiliroside binding to the colchicine site of tubulin, employing molecular docking and molecular dynamics simulations. We used a high-resolution crystal structure of the tubulin heterodimer with colchicine (Protein Data Bank ID: 6XER) as the receptor model.[3] Initially, we validated our docking approach by redocking colchicine into its binding pocket and comparing the pose to the crystal structure. We then performed induced-fit docking of tiliroside to allow protein flexibility and more accurate pose prediction, followed by rescoring with Prime MMGBSA for binding free energy estimation. Furthermore, to assess the stability of the predicted complexes, we carried out all-atom explicit-solvent MD simulations (100 ns) for both tubulin–tiliroside and tubulin–colchicine complexes. The simulation trajectories were analyzed for RMSD (to monitor complex stability) and protein–ligand interactions over time. Through this approach, we aimed to (1) confirm that tiliroside can indeed bind in the colchicine site, (2) compare its binding mode and affinity to that of colchicine, and (3) evaluate the dynamic stability of the tiliroside–tubulin interaction. Our findings provide molecular-level insight into tiliroside’s binding, suggesting it can serve as a stable tubulin inhibitor and a potential lead for anticancer drug development.

MATERIALS AND METHODS

Protein preparation

The X-ray crystal structure of the α, β-tubulin heterodimer in complex with colchicine and the stathmin-like domain (PDB 6XER, 2.5 Å resolution) was used as the receptor model.[3] The structure was prepared using Schrödinger’s Protein Preparation Wizard (Maestro 2021.4, Schrödinger, LLC, New York, NY, USA). All heteroatoms and crystallographic water molecules were removed, except for key structural waters. Missing side chains were fixed, and hydrogen atoms were added consistent with physiological pH. The protonation states of titratable residues were adjusted using PROPKA at pH 7.4. The tubulin structure includes the colchicine-binding pocket located at the interface of α-tubulin (chain A) and β-tubulin (chain B). For docking studies, colchicine was extracted from the binding site to generate the apo receptor. The binding site was defined as all residues within 10 Å of the colchicine-binding region (β-tubulin residues around helix H7–H8 and α-tubulin residues in the T5 loop, based on the crystal pose).

Ligand preparation

The 2D structure of tiliroside was obtained from the literature (confirmed as kaempferol-3-O-(6’’-p-coumaroyl)-β-D-glucoside) and sketched in Maestro. LigPrep (Schrödinger, LLC) was used to generate the 3D conformer and to perform energy minimization. Possible ionization states were evaluated, but tiliroside (a polyphenolic glycoside) is neutral at physiological pH, so no protonation state change was necessary except to ensure deprotonation of phenolic hydroxyls was turned off (treated as OH). The OPLS4 force field was applied for ligand parametrization. Colchicine was prepared similarly for redocking, starting from the crystal conformation.

Molecular docking procedure

Docking simulations were carried out with Glide (Schrödinger) using the standard precision (SP) and extra precision (XP) modes. Initially, a redocking validation was performed: colchicine was redocked into the colchicine site of tubulin. The Glide scoring grid was centered on the colchicine site, with a box size large enough (~20 ų) to accommodate colchicine or tiliroside. Default van der Waals scaling (0.8 for nonpolar atoms) was used for initial docking. The top-ranked colchicine pose from SP docking was compared to the crystal pose by RMSD to validate the accuracy of the docking protocol.

Induced-fit docking (IFD) was then employed for tiliroside to allow receptor flexibility. The IFD workflow (Schrödinger 2021-4) consists of an initial Glide docking (precision level: XP) to generate preliminary poses, protein side-chain refinement around the ligand using Prime, and a final redocking of the ligand into the refined site.[4] Key parameters included a 5 Å residue flexibility radius around the ligand and keeping residues βCys241, βAsn249, βLys254, αThr179, and αVal181 (among others at the pocket) free to move during refinement (as these were anticipated interaction hotspots). The top induced-fit poses were scored by the IFDScore (a combination of docking score and receptor strain energy) and the Glide XP GScore. The highest-ranked tiliroside pose was selected for further analysis. Glide scoring functions such as GlideScore and Emodel were recorded. Additionally, the Prime MMGBSA tool was used to estimate the binding free energy (ΔG_bind) of tiliroside and colchicine in their docked complexes, using the VSGB 2.0 solvation model and OPLS4 force field.

Molecular dynamics simulation

All-atom MD simulations were performed with Desmond v6.8 (Schrödinger) to evaluate the stability of the tubulin–ligand complexes. The selected induced-fit docked complex of tiliroside with tubulin and the crystal pose of colchicine with tubulin (from PDB 6XER) were used as starting structures for two separate simulations. Each complex was embedded in an orthorhombic periodic box of TIP3P water extending at least 10 Å from the protein in all directions. Appropriate numbers of Na⁺ and Cl⁻ ions were added to neutralize the system and mimic a 0.15 M physiological salt concentration. The systems were parametrized with the OPLS4 force field for protein and ligand. After building, each system was minimized with a relaxed restraint on the protein heavy atoms, then equilibrated using the default Desmond relaxation protocol (which includes gradual heating to 300 K and short NVT/NPT simulations with decreasing restraints).

Production MD was conducted in the NPT ensemble at 300 K and 1 atm for 100 ns for each complex. The temperature and pressure were controlled by a Nosé–Hoover thermostat and Martyna–Tobias–Klein barostat, respectively. A 2 fs time step was used with the RESPA integrator (bonded interactions at every step, short-range nonbonded every step, long-range every three steps). Long-range electrostatics were handled with the Particle Mesh Ewald method (grid spacing ~1 Å). Coordinates were saved every 50 ps for analysis.

Trajectory analysis

Simulation trajectories were analyzed using the Maestro Simulation Interaction Diagram tool. The RMSD of the protein Cα atoms and ligand heavy atoms (with respect to the starting minimized structure) was calculated over time to assess system stability. Protein–ligand contacts were characterized throughout the trajectory; hydrogen bonds, hydrophobic contacts, π–π stacking, π–cation interactions, and water bridges were monitored. An interaction was considered present if geometric criteria were met (e.g., hydrogen bond defined by donor–acceptor distance <3.5 Å and angle >120°). The percentage of simulation time each specific interaction was maintained was computed. These data were presented as an interaction timeline (a map of contacts vs. time) and a histogram of interaction frequencies (summarizing the fraction of the trajectory each residue was involved in a particular interaction with the ligand). All figures of molecular structures and surfaces were prepared with Maestro and PyMOL (Schrödinger and Open Source PyMOL).

RESULTS AND DISCUSSION

Docking validation and predicted binding modes

Our docking protocol accurately re-created the binding pose of colchicine. Glide XP docking positioned colchicine in the colchicine site with an RMSD of 0.2362 Å relative to the crystallographic pose, indicating excellent agreement. All key interactions of colchicine were preserved in the redocked pose. Notably, the trimethoxyphenyl ring of colchicine nestled into the deep hydrophobic pocket of β-tubulin, and the tropolone ring extended toward the α/β interface. The docked colchicine formed a hydrogen bond with βCys241 via its trimethoxyphenyl O-ester (mimicking the crystallographic interaction).[1] In addition, the tropolone carbonyl of colchicine was situated to form hydrogen bonds with αThr179 and αVal181 on the α-tubulin side of the interface.[1] These interactions are consistent with known colchicine–tubulin-binding features,[1] further supporting the validity of the Glide docking results. The successful redocking gave us confidence to proceed with docking tiliroside.

Docking results indicate that tiliroside binds in the colchicine site of tubulin with a pose overlapping that of colchicine, but with additional interaction capability due to its larger size and polyphenolic groups. Tiliroside’s planar kaempferol nucleus (flavonol aglycone) is inserted deeply into β-tubulin’s pocket, in the region where colchicine’s trimethoxyphenyl A-ring binds.[1] In this position, the flavonol ring system of tiliroside engages in π–π stacking and hydrophobic interactions with the surrounding nonpolar residues (e.g., βLeu242 and βAla250 on β-tubulin) that line the pocket. The 2D diagram highlights hydrophobic contacts with these residues. The p-coumaroyl substituent (a phenylpropanoid moiety) attached to the glucose extends into a sub-pocket and contributes additional hydrophobic interactions, further anchoring the ligand.

Importantly, tiliroside can form several hydrogen bonds within the pocket. The docking pose showed that the 4′-hydroxyl group of the flavonoid ring and the sugar hydroxyls act as hydrogen bond donors/acceptors to β-tubulin residues. Notably, a consistent hydrogen bond is predicted between tiliroside and the side chain of βCys241 – a residue known to interact with colchicine’s A-ring region (McLoughlin and O’boyle, 2020). Tiliroside also hydrogen bonds with βAsn249 and the backbone carbonyl of βAla250 (through its glucosyl OH groups), and with the side chain of βLys254 via a combination of hydrogen bonding and π-cation interaction (the aromatic rings of tiliroside’s coumaroyl group can interact with the Lys254 ammonium). Additionally, on the α-tubulin side, the hydroxyl group at C-5 of the flavonol and/or a sugar oxygen forms a hydrogen bond with αThr179]. αThr179 and αVal181 are part of the α-tubulin loop that closes over the colchicine site; in colchicine’s binding, these residues hydrogen bond to the tropolone ring.[1] In our tiliroside model, αThr179 serves as a hydrogen bond partner, whereas αVal181 (nonpolar) makes van der Waals contact with tiliroside’s sugar moiety. Overall, the docking suggests that tiliroside spans the colchicine pocket, contacting both β- and α-tubulin, in line with its larger polycyclic structure. This multivalent binding mode could confer higher affinity compared to colchicine.

In contrast, colchicine’s docking pose underscores the smaller ligand’s more limited interaction pattern. The 2D interaction diagram for colchicine reveals only one or two hydrogen bonds—typically one with βCys241 (through the methoxy or carbonyl group on ring A) and possibly one with αThr179 via the tropolone ring’s carbonyl. The remaining interactions are hydrophobic or van der Waals contacts with the pocket. Colchicine’s trimethoxyphenyl ring occupies the deep β-tubulin pocket and makes extensive hydrophobic contacts (with residues analogous to those tiliroside engages, like βLeu242), but it lacks polar groups to hydrogen bond in that region. Its tropolone ring (ring C) lies near αThr179 and αVal181, contributing one hydrogen bond and some aromatic contacts, but overall colchicine does not reach as far across the pocket interface as tiliroside does.

All the docking scores consistently favored tiliroside over colchicine. The Glide XP GScore for tiliroside was –11.462, significantly more favorable (lower) than colchicine’s –7.975. Glide standard scoring (GlideScore) similarly favored tiliroside (–16.773 vs –10.247 for colchicine). The Glide Emodel, which accounts for protein–ligand interaction energy and ligand internal strain, was quite favorable for tiliroside (–185.159) and equal to colchicine’s (the Emodel values were identical, which likely indicates a similarly well-resolved pose). More tellingly, the MMGBSA binding free energy ΔG_bind predicted by Prime was about –50.46 kcal/mol for tiliroside, compared to –36.62 kcal/mol for colchicine, suggesting tiliroside forms a significantly more stable complex with tubulin in silico. The induced-fit docking score (IFDScore)—a composite scoring of the IFD pose—was also slightly better for tiliroside (–1903.47 vs –1895.49, note that this score includes large negative terms and is useful mainly for rank ordering). These data support the idea that tiliroside not only fits well in the colchicine site but also interacts strongly, in accordance with the qualitative analysis of its binding mode.

The docking results thus strongly indicate that tiliroside can bind in the colchicine site of tubulin with equal or greater affinity than colchicine itself. However, docking alone provides only a static picture. To further validate tiliroside’s binding and to observe the complex behavior in a dynamic environment (e.g., accounting for protein flexibility and solvent effects), we carried out MD simulations of the tubulin–tiliroside complex, as well as the native tubulin–colchicine complex for reference. The following sections describe the stability and interactions of both complexes over 100 ns simulations.

Molecular dynamics simulation of tiliroside–tubulin complex

The MD simulation confirms that the tiliroside–tubulin complex is stable in an aqueous environment, with tiliroside remaining securely bound in the colchicine site. As shown by the RMSD analysis [Figure 1a], the protein Cα RMSD plateaued at around 2–3 Å after a brief period of adjustment (<10 ns), indicating that tubulin did not undergo any large conformational deviations; it largely retained its starting (curved) conformation associated with the colchicine-bound state.[5] The ligand tiliroside exhibited an initial RMSD rise to ~1.5–2.0 Å as it settled into the binding pocket during the early phase of the simulation, but thereafter it fluctuated only modestly (generally within 1 Å of its equilibrated position).

Figure 1.

Molecular dynamics simulation results for the tubulin–tiliroside complex (100 ns). (a) Root-mean-square deviation (RMSD) plots over time for the protein (tubulin Cα atoms, black line) and the tiliroside ligand (magenta line, heavy-atom RMSD) relative to the initial docked structure. Tubulin’s backbone RMSD stabilized around ~2.5 Å, indicating that the protein maintained its overall fold with minor fluctuations. Tiliroside’s RMSD remained low (under 2 Å for the majority of the trajectory after an initial equilibration period), demonstrating that tiliroside stayed tightly bound in the colchicine site without significant positional drift. In MD analyses, a consistently low ligand RMSD indicates the ligand stays close to its initial binding pose – a hallmark of stable, high-affinity binding – whereas a ligand that is weakly bound would show large or erratic RMSD deviations (b) Protein–ligand interaction timeline diagram. Each horizontal line represents a specific contact between tiliroside and a tubulin residue over the course of the simulation; colored blocks indicate frames where the interaction is present. Key interactions are annotated: persistent hydrogen bonds were observed with βCys241 and βAsn249 (blue lines nearly continuous across the simulation), as well as with βLys254 and αThr179 for a substantial portion of time. Hydrophobic interactions (green lines) with residues such as βLeu242 and βAla250 were consistently maintained. Occasional water-mediated bridges (orange lines) formed, e.g., connecting tiliroside to αVal181 via bridging water. (c) Interaction fraction histogram summarizing the percentage of simulation time each residue was engaged in a specific interaction with tiliroside. βCys241 and βAsn249 show ~90% hydrogen bond occupancy, indicating they formed H-bonds with tiliroside almost throughout the simulation. βLys254 and αThr179 also show high percentages (50–70%) for hydrogen bonds or salt-bridge/π-cation interactions. Hydrophobic contacts with βLeu242 and βAla250 were present >80% of the time. This profile highlights a network of interactions anchoring tiliroside in the binding site. From a drug-design perspective, tiliroside’s low RMSD and steadfast interactions suggest a longer residence time on tubulin and potentially higher binding affinity compared to a ligand that is more weakly bound

The interaction timeline [Figure 1b] provides a detailed view of how specific contacts between tiliroside and tubulin evolved over time. Several interactions identified in docking were maintained persistently. In particular, a hydrogen bond between tiliroside and βCys241 was present for almost the entire simulation (>95% of frames), confirming the importance of this anchoring interaction. βCys241 lies at the bottom of the pocket on β-tubulin, and the stability of this H-bond suggests tiliroside’s flavonoid core remained in that deep pocket without losing contact. Another hydrogen bond with βAsn249 was also highly stable (~90% occupancy); βAsn249 is located on a loop (βT7 loop) adjacent to the pocket, and it appears to form a steady H-bond with one of tiliroside’s sugar hydroxyls. βLys254, which was predicted to engage in hydrogen bonding and π-cation interactions, indeed showed an interaction with tiliroside for a significant duration (~50–60% of the time). The timeline indicates that sometimes the Lys254-tiliroside contact was direct hydrogen bonding, and other times it manifested as a π-cation stacking with the coumaroyl ring; both contribute to keeping the ligand in place at the pocket entrance. On the α-tubulin side, αThr179 formed a persistent hydrogen bond (~55% occupancy) with tiliroside (often via the 5-OH of the flavonoid), corroborating the docking prediction that tiliroside can bridge to α-tubulin’s loop.

Hydrophobic interactions also played a consistent role. Figure 1b shows continuous hydrophobic contact (green bars) between tiliroside and residues such as βLeu242 and βAla250 across the simulation. These nonpolar contacts, each present >80% of the time [Figure 1c], indicate that tiliroside’s aromatic rings remained snugly packed against these hydrophobic pocket residues. The phenyl ring of the p-coumaroyl group, for example, stayed in contact with βLeu242 deep in the pocket. Additionally, the simulation revealed dynamic water-mediated interactions: water bridges (orange bars in Figure 1b) occasionally formed, linking tiliroside to certain residues when direct hydrogen bonds broke transiently. One such water bridge was observed connecting tiliroside’s glycosidic oxygen to αVal181’s backbone, effectively mediating an indirect hydrogen bond for ~20% of the simulation frames. These water bridges illustrate the solvent’s role in reinforcing ligand binding.

The interaction frequency histogram [Figure 1c] quantifies these observations. βCys241 and βAsn249 each show ~90% hydrogen bond frequency with tiliroside, making them the top contributors. βLys254 shows around 60% (combining H-bond and π-cation interactions), and αThr179 about 55%, as noted. Several other residues (βLeu242, βAla250, αVal181) register significant contact percentages (mostly hydrophobic or van der Waals interactions). Importantly, no single crucial interaction drops to 0% early or is sporadic; instead, tiliroside is held by a redundant network of contacts. This network likely explains the low RMSD and overall stability of the complex – even if one interaction temporarily breaks, others still secure the ligand, and broken contacts often re-form (as evidenced by continuous or frequently recurring lines in the timeline).

In summary, the MD simulation demonstrates that tiliroside remains stably bound in the colchicine site, validating the docking pose under dynamic conditions. The ligand does not exhibit any tendency to dissociate or wander away from the pocket. Instead, it maintains a tight grip on tubulin via multiple interactions that persist over time. The ability of tiliroside to form simultaneous hydrogen bonds with both β- and α-tubulin (βCys241/βAsn249 and αThr179) and strong hydrophobic interactions in the β-subunit pocket underlies its stable binding. These MD results reinforce the notion that tiliroside is a promising tubulin binder, possibly with stronger residence or affinity than colchicine due to its polyfunctional binding mode.

Molecular dynamics simulation of colchicine–tubulin complex

The colchicine–tubulin simulation provides a baseline for comparison and further illustrates the importance of the interactions identified above. Tubulin with colchicine-bound remained structurally stable over 100 ns (protein RMSD ~2.5 Å), similar to the tiliroside complex. Colchicine also stayed in the binding pocket throughout the simulation; its RMSD fluctuated in the 1–3 Å range, which indicates it explored the pocket a bit more freely than tiliroside did (which is expected given colchicine’s smaller size and fewer anchoring points). There were a few transient increases in colchicine’s RMSD toward ~2.5 Å, suggesting slight reorientations or wiggles of colchicine within the site, though these settled back, and no unbinding was observed.

The contact analysis reveals that colchicine’s binding, while stable, relies on a simpler set of interactions. βCys241 emerged again as the most critical residue: colchicine formed a hydrogen bond with βCys241 for about 80% of the trajectory (often via the carbonyl on colchicine’s tropolone ring interacting with the cysteine’s backbone NH or side-chain SH as an H-bond donor). This interaction mirrors one of the main contacts seen in the crystal structure[1] and underscores βCys241’s role as a primary anchor for colchicine analogues. In the simulation, whenever the βCys241 H-bond broke, colchicine tended to shift slightly, but the bond usually reformed quickly, indicating a strong favorable interaction.

Colchicine’s other interactions were more intermittent. Hydrogen bonding with αThr179 occurred around 40% of the time—colchicine’s tropolone ring O7 (carbonyl) and O6 (methoxy) can, in some orientations, interact with αThr179’s side-chain hydroxyl or backbone, but these contacts were on-and-off during the simulation. Similarly, αVal181 (though nonpolar itself) contributed via backbone interactions or through water-bridged contacts ~30% of the time. The lesser persistence of these α-subunit interactions (compared to tiliroside’s case) suggests that colchicine did not constantly engage α-tubulin; at times, colchicine tilted or rotated such that its tropolone ring moved slightly away from αThr179/Val181, breaking the hydrogen bond, before later drifting back into a H-bonding position. This flexibility likely caused the slightly higher RMSD fluctuations noted above.

Hydrophobic interactions for colchicine were still significant, but a bit less uniformly maintained. βLeu242 and βAla250 each showed contact with colchicine roughly 60–70% of the time, which is high but a notch lower than in the tiliroside complex (which had >80% for the same residues). The timeline shows minor gaps in the green bars for these contacts, meaning colchicine occasionally shifted enough to momentarily reduce contact with one or the other hydrophobic residue. Colchicine’s trimethoxyphenyl ring, being rigidly attached, might pivot slightly within the pocket, changing its contact pattern transiently. Nonetheless, colchicine remained predominantly surrounded by hydrophobic pocket residues for most of the simulation, reflecting shape complementarity as a key factor in its binding.

Tiliroside had multiple interactions each above ~50% persistence, whereas colchicine has essentially one very strong interaction (βCys241) and a few moderate ones. This difference in interaction profile likely explains the more constrained behavior of tiliroside (lower RMSD) versus the somewhat greater mobility of colchicine in the pocket. In other words, tiliroside’s multi-point attachment secures it more tightly. From a drug-design perspective, this implies tiliroside (or analogues) could have a longer residence time on tubulin and potentially higher affinity, as our MMGBSA calculations also suggested.

Overall, the MD simulations corroborate the docking findings and provide dynamic validation: tiliroside is confirmed to bind stably at the colchicine site, and it remains engaged via a robust set of interactions. Colchicine, while also stably bound (as expected for a high-affinity ligand), shows that a smaller ligand with fewer functional groups relies on a couple of key interactions and hydrophobic fit. Tiliroside’s ability to maintain additional contacts translates into it being comparably, if not more, stably bound over time. This stable binding of tiliroside to tubulin could underlie potential antimitotic activity, as a ligand that strongly occupies the colchicine site would prevent tubulin polymerization similar to colchicine and combretastatin A-4. Our computational results thus strongly suggest that tiliroside, or derivatives thereof, merit experimental evaluation for antitubulin activity.

Comparing the two simulations, a clear pattern emerges: tiliroside’s larger, polyfunctional structure allows it to engage tubulin at multiple points simultaneously, leading to a very stable complex (low ligand RMSD and sustained interactions), whereas colchicine, with fewer contact points, shows a bit more mobility in the pocket (higher RMSD fluctuations). In essence, tiliroside’s flavonoid scaffold—with its glycosyl and p-coumaroyl extensions—behaves as a “molecular anchor” in the colchicine site, which may translate to higher binding affinity and longer target residence time than a smaller ligand like colchicine. This hypothesis is supported by the docking and MD data, and it aligns with prior observations that multivalent ligand interactions can enhance binding stability.

Our findings also highlight how tiliroside’s behavior compares to other flavonoids. Simple flavonols or flavones (aglycones without sugar moieties) have been reported to interact with tubulin, but generally with lower affinity. For example, the dietary flavonol quercetin (which is structurally related to tiliroside’s aglycone, kaempferol) is known to bind tubulin at the colchicine site with a dissociation constant in the low micromolar range. Quercetin can inhibit colchicine binding and perturb tubulin polymerization, but because it lacks the glycosidic and phenylpropanoid extensions, it engages fewer contact points on tubulin. In our study, tiliroside’s bulky glucose–coumaroyl substituent allowed additional hydrogen bonds and hydrophobic anchoring, effectively “locking” it into the pocket. Thus, compared to smaller flavonoids like quercetin, tiliroside forms a more extensive interaction network, which correlates with more stable binding. We postulate that tiliroside’s enhanced binding capability stems from this multivalent attachment – it occupies a larger portion of the colchicine site and interacts with both β- and α-tubulin subsites simultaneously, something a smaller flavonoid cannot easily do. Tiliroside may therefore represent an optimized flavonoid scaffold for tubulin inhibition. This insight could be valuable in the design of novel flavonoid-based CBSIs; adding functional groups (e.g. glycone or aromatic extensions) to flavonoid cores might improve their tubulin-binding affinity and specificity.

CONCLUSIONS

In this study, we combined molecular docking and molecular dynamics approaches to investigate the binding of tiliroside, a natural flavonoid glycoside, to the colchicine site of tubulin. The IFD results predicted a favorable binding pose for tiliroside, engaging both α- and β-tubulin residues in the colchicine pocket and achieving better docking scores than the native ligand colchicine. Key interactions included hydrogen bonds with βCys241 and αThr179, and extensive hydrophobic contacts, suggesting that tiliroside can anchor strongly within the pocket. Subsequent 100 ns MD simulations confirmed that the tiliroside–tubulin complex is indeed stable: tiliroside remained bound in the colchicine site throughout the simulation, with minimal drift and a persistent network of protein–ligand interactions. By comparison, colchicine, while also stably bound, showed a more limited interaction pattern and slightly higher positional fluctuation in the pocket. These findings indicate that tiliroside’s larger, polyfunctional structure allows it to bind tubulin in a multivalent manner, potentially conferring higher affinity and stability.

Our results support the notion that tiliroside is a viable lead compound for tubulin targeting. Given its natural origin and reported anticancer effects[2], tiliroside could inspire a new class of colchicine-site inhibitors. The strong binding of tiliroside observed in silico warrants further experimental validation, such as tubulin polymerization assays or cell-based tests of antimitotic activity. Additionally, the detailed interaction map provided by our simulations can guide the design of tiliroside analogues with optimized substituents to enhance specific interactions (for instance, strengthening the interaction with αThr179 or βLys254 could further improve affinity). In conclusion, this computational investigation provides a molecular rationale for the anticancer potential of tiliroside and lays the groundwork for future development of flavonoid-based tubulin inhibitors targeting the colchicine site.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

The Deanship of Scientific Research (DSR) at King Abdulaziz University (KAU), Jeddah, Saudi Arabia has funded this project, under grant no. (RG-12-166-43).