Xanthatin and 8-epi-xanthatin as new potential colchicine binding site inhibitors: a computational study

Phytocompounds xanthatin and 8-epi-xanthatin, obtained from Xanthium chinese Mill, showed antitumoral activity in vitro related to the microtubules destabilizing properties of these phytocompounds. Five binding sites for microtubule destabilizing agents have been characterized on tubulin by high-resolution X-ray crystallography: vinca domain, colchicine, pironetin, maytansine site, and more recently, the seventh site. This work aims to develop a comprehensive computational strategy to understand and eventually predict the interaction between xanthatin and 8-epi-xanthatin with the destabilizing-antimitotic binding domain of the tubulin heterodimer. In addition, we propose a putative binding site for these phytocompounds into the microtubule destabilizing binding sites on the tubulin heterodimer. Xanthanolides showed higher stability in the colchicine and pironetin binding sites, whit a greater affinity for the former. In addition, we found that xanthanolides and non-classical colchicine binding site inhibitors share a high structural similarity. The 3D structures for xanthatin and 8-epi-xanthatin were obtained using DFT with the hybrid functional B3LYP and the base 6-31G (d,p), implemented in Gaussian 09. The 3D coordinates for tubulin proteins were downloaded from PDB. The complexes tubulin-xanthanolides were predicted using a Monte-Carlo iterated search combined with the BFGS gradient-based optimizer implemented in the AutoDock Vina. The xanthanolides-tubulin complexes were energy minimized by molecular dynamics simulations at vacuum, and their stabilities were evaluated by solvated molecular dynamics simulations during 100 ns. All molecular dynamics simulations were performed using the conjugate gradient method implemented in NAMD2 and CHARMM36 forcefield.


Introduction
For decades, cancer has been the focus of many researchers. Despite all treatments available to date [1], cancer remains a very elusive disease. One of the most employed modalities to treat tumors is chemotherapy by the use of antimitotic compounds ables to disrupt the dynamic instability of microtubules (MTs) by targeting either MTs or the MTassociated proteins (MAPs) [2].
MTs are cytoskeletal polymers of αβ tubulin that polymerize and depolymerize quickly in a nucleation mechanism. α-Tubulin and β-tubulin have similar structures, with a molecular weight of around 55 kDa [3]. α-Tubulin and β-tubulin interact to form heterodimers, and several heterodimers are connected from head to tail to form protofilament. Each subunit contains a guanosine triphosphate (GTP) binding site. The α-tubulin has a stable nonexchangeable nucleotide-binding site (N-site) that only binds GTP, while the β-tubulin has an exchangeable nucleotide site (E-site) that binds both GTP or GDP. When GTP is bound to β-tubulin subunits at the plus end of a microtubule, another soluble GTP-bound heterodimer readily adds on to the end and the microtubule grows. The addition of a new dimer triggers the hydrolysis of GTP on the subunit that is no longer exposed and GDP is formed. However, if a GDP-bound (curved conformation) subunit becomes exposed at the plus end, perhaps through the departure of some subunits, it will cause a conformational change resulting in the rapid depolymerization of the microtubules [4]. Therefore, tubulins are continuously polymerized and assembled into microtubules at one end and depolymerized at the other end, forming a dynamic cycle in the process of tubulin polymerization and depolymerization [3].
These components play an essential role in a wide number of cellular functions as the cellular division [5]. The assembly between the MTs and the kinetochores is crucial for the correct formation of the mitotic spindle and the subsequent alignment and segregation of the chromosomes [6]. Drugs that target the MTs can be classified into three main groups based on their mechanism of action: MT-destabilizing agents (MDAs), MT-stabilizing agents (MSAs) [7], and more recently tubulin degradation agents (TDAs) [8][9][10]. MDAs promote depolymerization and prevent polymerization of tubulin when it is administrated in high concentrations [11], while MSAs have an opposing effect. On the other hand, TDAs destabilize the tubulin heterodimer, promoting its unfolding, thereby inducing its degradation in a proteasome-dependent pathway [9,10].
MDAs exert their mechanism of action over five binding sites in tubulin: colchicine binding site (CBS) [12], vinca domain [13], maytansine binding site [14], pironetin binding site [15], and the seventh site [10,16]. MSAs interact by two well-defined pockets in the tubulin heterodimer: taxoid binding site [17] and LAU/PEL binding site [18]. Interestingly, TDAs known to date do not have distinct binding sites; they are able to bind into the colchicine domain or the seventh site [9,10].
However, despite the wide variety of antimitotic drugs that currently exist [19,20], these are not able to completely remove the tumor and generally promote drug resistance. It justifies the continuing work on developing and designing new candidates for antimitotic drugs. Based on their use in traditional medicine, the plants have been widely studied for the search of new molecules with antitumoral activity [21].
Previous works performed by Sanchez-Lamar in 2016 [22] and Piloto-Ferrer in 2019 [23], using CHO and CT26WT cell lines, respectively, evidenced for the first time that the phytocompounds present in the vegetal extract from Xanthium chinese Mill (before known as Xanthium strumarium) were able to disrupt the dynamic properties of the mitotic spindle by MT destabilization. The chemical fractionation allows to identify as the principal responsible of this destabilizing effect the isomers xanthatin and 8-epi-xanthatin [22] (Fig. 1). However, the binding pocket of these phytocompounds in the tubulin protein remains unknown. Although there are currently experimental methodologies that allow elucidating at a molecular level the mechanism of interaction of small molecules with biomolecules, these are extremely expensive and take a long time and big amounts of reagents [24]. In the last decades with the accelerated evolution in computing media and bioinformatics, theoretical procedures had emerged to help in the investigation and understanding of physical and biological phenomena of interest. The aim of this work is to propose a putative binding site for these phytocompounds into the MDA binding sites in the tubulin heterodimer, using docking and molecular dynamics simulations.

Receptor preparation
The 3D structures of tubulin proteins were downloaded from RSCB Protein Data Bank (PDB) (www. rscb. org) [31]. Three different structures were selected to perform the analysis in the colchicine domain, due to the T7 loop (R243-L252) adopts at least three different conformations, that can close or open the pocket, according to the assembly state and presence of an exogenous molecule in the colchicine site [32]. The PDB IDs are 3ut5 (open, the T7 loop is extended and the colchicine pocket is accessible), 5ca0 (close, the pocket is blocked by the loop), and 5c8y (middle, the loop is located in middle conformation with respect to the structures mentioned above). On other hand, for the analysis in the vinca domain and maytansine binding site only one structure of tubulin was selected (PDB ID: 3ut5). Since the pironetin binding site, is normally in a close state and is opened by Pironetin binding, a structure that contains a complex Pironetin-tubulin (PDB ID: 5fnv) was used to study this site. [15,33]. All the PDB structures selected as receptors were first cleaned leaving only the tubulin heterodimer, the GTP, GDP, and the Mg 2+ ions from each subunit. Finally, the receptors were prepared for the docking using the software AutoDock Tools [29]. Because the vinca domain and the seventh site are equivalent, the results obtained for the former were assumed for both.

Docking
The complexes tubulin-xanthanolides were predicted using a Monte-Carlo (MC) iterated search combined with the BFGS gradient-based optimizer implemented in AutoDock Vina [34,35]. The conformational search space for each binding pocket was specified by a cubic box, previously defined using AutoDock Tools, each of them surrounding the destabilizing antimitotic binding pocket based on the literature evidence ( Table 1). The docking simulations were performed considering all rotatable angles of xanthanolides as flexible and tubulin as rigid. All parameters in the docking were established by default, with the exception of: energy_range = 4 kcal and num_modes = 20. In each of the studied binding pockets, the xanthanolides-tubulin complexes obtained were clustered using ligand root mean square deviation (RMSD) value of 2 Å as criteria [36]. From each cluster, one structure was selected as the representative binding conformation based on the lower value of the AutoDock Vina scoring function.

Optimization of xanthanolides-tubulin complexes
Xanthanolide-tubulin representative binding conformations obtained from docking were used as the start point. Xanthanolide parameters were obtained using the VMD [37] plugin FFTK (force field tool kit) [38]. On other hand, the parameters for non-protein components of receptors (GTP, GDP, Mg 2+ ) were obtained from the online server CGenFF (https:// cgenff. param chem. org) [39]. Molecular dynamics simulations were performed at vacuum, using the software NAMD v2.12 [40] and CHARMM36 force field [41]. An NVT ensemble was employed, and only the backbone of the amino acid residues around 10 Å from ligand was used as flexible, the rest was set as rigid. The temperature was fixed at 310 K and was controlled with a Langevin thermostat [42]. The timestep used was 2 fs, first, a minimization for 1 ps was performance followed by 500 ps of simulation. The root means square deviation (RMSD) for the ligand in all molecular dynamics simulations was calculated using as a reference the first frame.

Stability analysis of solvated xanthanolides-tubulin complexes
The final position of optimized xanthanolides-tubulin complexes was used as the start point. The systems were solvated and the charge neutralized using Na + and Cl − at a final concentration of 0.05 mol/L, using the VMD plugins Solvate and Autoionize respectively. A cubic box of solvation was generated, surrounding each complex at a distance of 20 Å from the complex surface to the borderline of the box, and using the explicit solvent model TIP3 [43]. The simulations were performed using the software NAMD v2.12 [40], the CHARMM36 force field [41], and under periodic boundary conditions. An NPT ensemble was used, the temperature was fixed at 310 K and the pressure at 1 bar, these parameters were controlled using a Langeving thermostat and barostat [44], respectively. The minimization time was 1 ps, and the production run was for 100 ns with a timestep of 2 fs. The RMSD for the ligand in all molecular dynamics simulations was calculated using as a reference the first frame.

Binding free energy calculations using the LIE-D methodology
The binding free energy calculation of the xanthanolides-tubulin complexes that remains stable after 100 ns of the simulation was calculated using the linear interaction energy methodology improved by Miranda and coworkers in 2015 [45]. The LIE-D formula (Eq. 1) takes into account the intra-ligand electrostatic interactions as suggested by Almlöf and coworkers [46]: where Δ⟨V el l−s ⟩ and Δ⟨V el l−l ⟩ terms indicate the change in electrostatic interactions when the ligand is transferred from the solution (free state) into the solvated receptor binding site (bound state). Δ⟨V vdw l−s ⟩ denotes the change in van der Waals interactions between the bound and the free state. A β value of 0.43 was calculated with the β FEP specific values for each ligand using the parameterization model E proposed by Almlöf and coworkers [46] (Table 2) and Eq. 2: where w i , β 0 , and Δβ i were calculated from explicit solvent FEP calculations of single chemical groups (Table 2). (1) This model uses w i = 1 and w i = 11, for all neutral and charge groups (anions and cations), respectively [46]. A value of α = 0.18 was employed, as it seems to be a robust value from previous works [47][48][49]. The γ coefficient value was calculated using Eq. where the parameter D define the balance (difference) between electrostatic (polar) and van der Waals (nonpolar) contributions to binding free energy in the LIE-method by the equation: The Δ⟨V el l−s ⟩, Δ⟨V el l−l ⟩ , and Δ⟨V vdw l−s ⟩ values were calculated from the generated trajectories for each ligand-protein complexes and the xanthanolides alone in solution. The energy values were calculated using the plugin NAMD Energy implemented in VMD software.

Xanthanolides-tubulin complexes obtained by docking
A total of 54 xanthanolides-tubulin representative conformations were obtained, 12 in the colchicine binding site (CBS), 17 in the vinca domain, 10 in the maytansine binding site, and 15 in the pironetin binding site (Supplementary  table 1). Two specific orientations were observed for the xanthanolides in the colchicine and pironetin binding sites. In the xanthanolides-colchicine binding site complexes, one, present the xanthanolides γ-lactamic ring pointing toward α-subunit of the tubulin heterodimer, while in the second one, it was pointing to the β-subunit ( Fig. 2A). In the pironetin binding site, one presented the γ-lactamic ring oriented to the cysteine 136 of the α-subunit (Cα136) and the aliphatic chain to the Cα4, while the other was inverted respect to the previous one (Fig. 2B). The energy range predicted by the AutoDock Vina scoring function, for the obtained complexes were around − 8.5 to − 6 kcal/mol and − 8.2 to − 5.5 kcal/ mol, for colchicine and pironetin binding sites, respectively. Among these representative conformations, poses 5, 6, 11, and 12 in the colchicine binding site, and 40 and 48 in the pironetin binding site highlight as the most stable conformations with the lower energetic values of the 54 obtained results. Additionally, it was also observed that, in the xanthanolide-colchicine binding site complexes, the energy tends to be lower as the conformation of the binding site was more closed (Supplementary table 1).
On the other hand, several representative conformations were obtained with random orientations in the vinca domain and maytansine binding site. For the study of the vinca domain, was necessary to split it on two, because according to Vinblastine binding inhibition studies [50], two different subsites have been identified in this domain: the dolastatin (Dol) [51] and vinblastine (Vlb) [52]. These subsites share a small region made up of T5 and H6-H7 loops. However, the dolastatin binding subsite reaches the helix H1 (Gβ10-Sβ25), extending the definition of the vinca domain [51]. Due to this, the docking was performed in each of the vinca domain subsites separately. The energy range predicted by the Auto-Dock Vina scoring function, for the obtained complexes were around − 5.8 to − 4.4 kcal/mol and − 6.5 to − 4.5 kcal/ mol, for vinca domain (both subsites) and maytansine binding sites, respectively (Supplementary table 1). As can be observed, the complexes obtained in these binding sites present higher energetic values with respect to the colchicine and pironetin binding sites. According to these results, the rank order of the xanthanolides affinity in each of the MDAs binding sites in tubulin is: colchicine binding site > pironetin binding site > maytansine binding site > vinca domain.

Xanthanolides-tubulin complex optimization
The molecular docking was performed using the xanthanolides as flexible but the protein as rigid. In order to consider the flexibility of the sidechain of amino acids in the binding site, a 500 ps molecular dynamic simulation was performed at vacuum. This short simulation allows a better selection of the most relevant docking poses for further analysis. In the xanthanolide-colchicine binding site, xanthanolides showed small variations in all selected structural variants of tubulin after 500 ps of MD simulation at vacuum. At the end of the simulation was observed that, when the T7 loop is in the open conformation, the xanthanolides locate close to the αβ-tubulin interphase; however, as the loop closes the pocket, xanthanolides are deeper internalized in the domain, increasing the possibility of interacting with a higher number of amino acid residues ( Supplementary  Fig. 1). This fact points to a possible entry mechanism of the xanthanolides into the colchicine binding site.
At  Fig. 2). From these were discarded for the further studies the poses 49, 46, 47, 40, and 42 according to the RMSD value obtained in the molecular dynamic simulation and the energy in the docking procedure. Additionally, both xanthanolides seem to be more stable when their γ-lactamic ring were pointing to the Cα4 than when they point to the Cα316. In the last case, xanthanolides tend to be located in the exit cleft of the domain and the RMSD values were higher ( Supplementary Fig. 2).
In the vinca domain, xanthanolides converge to similar results in each of the subsites after 500 ps of MD simulation at vacuum. In both, Vlb ( Supplementary  Fig. 3) and Dol subsites ( Supplementary Fig. 4), it is observed that even starting from different conformations in the initial time (t = 0), at the end of the simulation (t = 500 ps), all poses tended to adopt the same two representative conformations. In the first one, the xanthanolide γ-lactamic ring is directed toward the hydroxyl group (-OH) of Tyrosine 210 (Yβ210) and the Fig. 2 Representative conformation obtained for xanthatin (pink) and 8-epi-xanthatin (blue) in the colchicine binding site (A) and the pironetin binding site (B). The α-and β-subunit of tubulin are represented in dark gray and light gray ribbons, respectively aliphatic tail to the amine group (-NH) of the peptide bond of Yβ224. The second one is inverted with respect to the previous one. The first conformation showed greater representativeness than the second one for both xanthanolides in both subsites. Additionally, pose 15, which presented the highest value in the binding energy predicted by the AutoDock Vina scoring function, leave the binding site during the molecular dynamic simulation ( Supplementary Fig. 4). From each of these groups were selected for next studies the poses that show the lower RMSD value: for xanthatin, poses 18 and 19, and for 8EX, poses 22 and 25.
In the maytansine binding site, both xanthanolides also tend to adopt two different conformations equally represented at the end of the 500 ps of simulations. In one of these, the γ-lactamic ring is parallel to the indole ring of the Tryptophan 407 (Wβ407), while in the other, it is pointing to the hydrogen of the -NH group of the Wβ407 indole ring. Like in the vinca domain, some conformations leave the estimated sites by docking: poses 33 and 39, which also presented the highest values in the binding energy predicted by the AutoDock Vina scoring function ( Supplementary Fig. 5). From each of these groups were selected for next studies the poses with the lower RMSD value: for xanthatin, poses 31 and 32, and for 8EX, poses 34 and 35.

Stability evaluation of the xanthanolides-tubulin complexes
According to the obtained results by the docking procedure, the most stable complexes after 100 ns of molecular dynamic simulations were those belonging to the xanthanolides in the colchicine and pironetin binding sites. In the colchicine binding site, both xanthanolides showed minimal changes with RMSD values close or lower than 5 Å, showing a fast saturation of the curve and few fluctuations. This result suggests ligand stability in this binding site (Fig. 3A and B). However, in the pironetin binding site, even though xanthanolides remain in the domain after 100 ns, some complexes have a major structural change that can be as large as 10 Å of RMSD values. Many of these poses with high RMSD values tend to locate in the exit cleft of the domain as happened in the previous sections. Additionally, they present high fluctuations, meaning low stability of the ligand binding conformation ( Fig. 3C and D).
On the other hand, the complexes evaluated for the vinca domain and the maytansine binding site showed very low stability after 100 ns of molecular dynamic simulation. RMSD values bigger than 15 Å were obtained for all complexes, which is associated with low stability and with the dissociation of the ligand from the receptor. In all cases, the xanthanolides dissociate before 20 ns of simulation  (Fig. 4). These results match with the proposed hierarchical order to the possible stability of the xanthanolides in each of the MDAs binding sites of tubulin using the energies predicted by the AutoDock Vina scoring function for the obtained complex.
In order to propose the most probable MDA binding site to the xanthanolides on the tubulin, the LIE-D methodology was used to calculate the energy of the stable ligand-receptor complexes after 100 ns of simulation. The xanthanolide-colchicine binding site complexes showed the lowest values of energy from − 6.79 to − 5.21 kcal/mol ( Table 3). The interaction analysis reveals that these complexes are stabilized mainly by hydrophobic interactions with amino acid residues from both subunits predominating those from zone 2 of the   (Table 3), and in most cases, it tends to be located in the exit cleft of the pocket. On the other hand, the result obtained in the CBS and pironetin binding site showed stability along the molecular dynamic trajectory. In the first case, the RMSD obtained, showed minimal changes in the interaction of the ligand with the receptor, obtaining always values close or lower than 5 Å and a fast saturation of the curve, suggesting ligand stability in this domain [53]. In contrast, in the pironetin binding site, some RMSD values were higher and the curve presented big fluctuations related to low stability in the interactions between the ligands and the receptor. All previous results and the energy obtained with the LIE-D methodology suggest that the xanthanolides could interact through both, the colchicine and pironetin binding site, with greater affinity for the former.

Discussion
In a recently published work, was shown that the compound Cevipabulin binds indistinctly into the vinca domain and the seventh binding site, which are equivalents but located in different subunits [10]. In this sense, a superposition study based on the sequence identity homology shared by α-and β-tubulin, the 3D structures similarities, and the disposition of pironetin and colchicine binding sites in their respective subunits reveal that these are equivalent sites from different subunits. They share the same fold and their sequences are similar, with 34% of sequence identity and many of the observed amino acid changes are conservative (29%) [54]. All mentioned above make stronger our previous suggestion of two possible binding sites in tubulin for xanthanolides.
The higher stability observed for xanthanolides in colchicine and pironetin binding sites, against the vinca domain and maytansine binding site, could be for several reasons. First, the structural characteristics of each domain, the colchicine and pironetin binding sites are well-defined deep hydrophobic pockets [20,33,55], while the other sites are wider and more superficial, providing few hydrophobic interactions for the stability of xanthanolides. This leads to the antimitotic drugs that bind into the colchicine and pironetin binding sites being hydrophobic low molecular weight compounds [12,33], while those that bind at the vinca domain and maytansine binding site have more complexes structures and bigger sizes, being in many times small peptides [20,56]. Thus, considering that xanthanolides are compounds with a hydrophobic core with few hydrogen acceptors, two at one end (O17 and O18) and one in the other (O16), and low molecular weight, they can establish more favorable interaction in a well-defined hydrophobic pocket as the colchicine and pironetin binding site.
Additionally, pharmacophoric models that can help to understand the higher selectivity of the xanthanolides for the colchicine binding site have been developed. For the colchicine binding site, two pharmacophoric models have been proposed [57,58]. In one of them, exist seven pharmacophoric points: three hydrogen acceptors atoms (A1, A2, and A3), one hydrogen donor atom (D1), two hydrophobics centers (H1 and H2), and one planar group (R1). However, to make the molecule bioactive, it only requires at least one hydrogen acceptor atom, two hydrophobic centers, and a planar group [57]. The other pharmacophore was proposed by Wang and coworkers, 2016, in which five pharmacophoric points are proposed: three hydrophobic centers (I, II, and III) and two hydrogen bond centers (IV and V, either hydrogen bond acceptor or donor). Also, they calculate the distance to which these pharmacophoric points should be found: I-II (4-9 Å), II-III (4-8 Å), III-IV (3-5 Å), V-II (4.5-7 Å), II-IV (4.5-7 Å), and IV-I (5-6 Å) [58].
For the maytansine binding site and vinca domain, pharmacophoric models have also been proposed. The first one was suggested by Prota and coworkers in 2014 for the maytansine binding site; it consists of three key interaction points (I, II, and III) with the β-subunit of tubulin. The proposed distance between these pharmacophoric points was much higher than those raised for the colchicine binding site: I-II (10.7-12.1 Å), II-III (7.4-8.4 Å), and III-I (4.7-5 Å) [14]. The second one was proposed by Wang and coworkers in 2015; for the vinca domain, it contains two hydrophobic groups (I and II) and three hydrogen bond forming atoms.
The two hydrophobic groups I and II interact with hydrophobic pockets β and α, respectively. The three-hydrogen bond forming atoms interact with Dβ179 (one N atom) and Nα329 (one N and one O atom). There are two areas (A and B) at the opposite sides of the pharmacophore where extra chemical groups can be added to make additional interactions to increase the affinity of the ligand. As for previous pharmacophores, the possible distance between the pharmacophoric points was recommended: I-II (8.5-11.2 Å), II-N (5.8-8 Å), I-N (9-10.2 Å) [56].
Xanthanolides, xanthatin, and 8-epi-xanthatin are hydrophobic molecules with a representative core constituted by a bicyclic γ-lactamic ring which contains a seven-member ring fused with a five-member ring [59]. The first ring has an aliphatic branch while the second one has a methylene group. Additionally, they contain three oxygen atoms acting as hydrogen acceptors, enabling the interaction stabilizations by hydrogen bonds. According to these structural features, we propose preliminary pharmacophore model composed by five points: three hydrophobics centers (I, II, and III) and two hydrogen acceptors (IV and V) (Fig. 5). Points I and II are represented by the methyl (C15) in the aliphatic chain and the methylene (C13) from the five-member fused ring, respectively. Point II is a bulk hydrophobic center represented by the seven-member ring. The hydrogen acceptor center is the O16 (point IV) and O18 (point V). Distance between these points was measured (I-II: 6.70 Å, II-III: 4.06 Å, III-IV: 2.96 Å, V-II: 4.99 Å, II-IV: 5.89 Å, IV-I: 2.40 Å) and show to be similar to those proposed by Wang and coworkers in the pharmacophore model for the colchicine binding site [58] (Fig. 5). In addition, these proposed pharmacophoric points in xanthanolides are distributed in such a way that if the γ-lactamic ring is pointing toward either α or β-subunit; t hese keep a similar location and distances justifying the issue that two models equally stable were obtained in this work in the colchicine binding site. On other hand, when its isomers are compared with the pharmacophore's models proposed for the vinca domain, it does not match. This analysis of the potential pharmacophoric points of the xanthanolides has great importance for the pharmaceutical industry, leading to possible chemical modifications to improve the pharmaceutical efficiency as chemotherapeutics drugs.
Regarding the higher stability observed for the xanthanolides in the colchicine binding site than in the pironetin binding site, could be associated with several facts. The colchicine and pironetin binding sites are two well-defined hydrophobic pockets that bind small molecular weight molecules. However, despite the high similarity in their 3D structure, they present some differences in their amino acid composition [54]. These differences in the primary sequence could contribute to the lower stability observed for the xanthanolides inside the pironetin binding site. Also, the CBS is located at the interface of two unseparated subunits from the same heterodimer, providing amino acid residues from both subunits for xanthanolides interaction [12]. On the other hand, the pironetin binding site is located exclusively in the α subunit reducing the number of interactions for the stabilization of xanthanolides [60]. To date, Pironetin is the only known ligand that binds into the pironetin binding site. This pocket is usually in a closed conformation, Pironetin binding opens it with some residues shifting more than 10 Å [60]. The ethyl group of Pironetin seems to play an essential role in its action, due to the biological activity decrease in Pironetin variants where this group has been removed or replaced by other groups [61][62][63][64]. All these facts represent a challenge for binding the xanthanolides into the pironetin binding site. Also, it could explain that the xanthanolides are located at the exit of the pironetin binding site at the end of molecular dynamics simulations.
So far, our results point to the idea that the isomers xanthatin and 8-epi-xanthatin are more likely to exert their mechanism of action as MT destabilizing agents by interacting with the colchicine binding site. Based on the analysis of Pérez-Pérez et al. [65], the colchicine domain ligands were classified into classical and non-classical colchicine binding site inhibitors (CBSIs) according to their spatial orientations. The superposition of these ligands with x-ray crystal structures indicated that classical CBSIs with more globular or butterfly-like shapes occupy zones 1 and 2, mimicking the colchicine binding mode. However, non-classical CBSIs with more planar Fig. 5 Pharmacophoric points proposed for xanthanolides compared with the pharmacophore model suggested by Wang Y. and coworkers in 2016 [58]. The green and yellow spheres represent the hydrophobics (Hyd) and acceptor (Acc) pharmacophoric points proposed for xanthanolides structures tend to locate deeper into the β-subunit making use of zones 2 and 3 [20,65]. The studied xanthanolides in this work share some structural similarities with several non-classical CBSIs leading to the idea that these isomers could exert their activity by zones 2 and 3 of the colchicine binding site. However, the interaction pattern obtained in most complexes with high stability showed interactions with amino acid residues from the three zones. To date, only two CBSIs can interact with the 3 zones, ABT-751 [32] and 2-(6-ethoxy-3-(3-ethoxyphenylamino)-1-methyl-1,4-dihydroindeno[1,2-c]pyrazol-7-yloxy)acetamide [66], which interact mainly with zones 1 and 2 but also establish a single hydrogen bond with the Tyrβ202 from zone 3. Nevertheless, a convergent binding pose for xanthanolides in the CBS was not obtained. Despite all point to the possibility that these isomers could be classified as novel nonclassical CBSIs, further studies are required to propose an accurate binding conformation and classification for xanthanolides as CBSI.
Despite colchicine being a cheap chemotherapeutic option, its clinical applications are limited by its toxicity. It strongly binds to tubulin in non-cancerous cells and causes impaired protein assembly, mitotic arrest, and multiorgan dysfunction [12]. Because of this, the search for new and more efficient CBS inhibitors continues to be an attractive open research field. In a previous study, it was observed that mice treated with the chemical fraction containing the xanthanolides did not show signs of toxicity, but a significant inhibitory effect in the proliferation of the tumor [23]. These results could be related to the fact that xanthanolides have lower binding energy (− 6.79 A-5.21 kcal/mol) than colchicine (− 10 A-9 kcal/mol) [67] in the CBS. Xanthanolides could efficiently bind to tubulin in the cancer cells with an overexpression of these proteins associated with their exacerbated cell growth. It would increase the arrest of tumor cells in the G2/M phase, which led to the apoptosis of these cells and the reduction of the tumor. However, in non-cancerous cells in a quiescent state, the union of the xanthanolides to tubulin will be lower, decreasing toxicity in patients.
In summary, xanthatin and 8-epi-xanthatin isomers could bind indistinctly into the pironetin and colchicine binding sites. However, based on our results and the features of the Pironetin binding into the pironetin binding site, these isomers seem to have a greater affinity for the CBS. Additionally, the molecular structure of xanthanolides matches with the pharmacophore model proposed for the colchicine binding domain, especially with the model suggested by Wang and coworkers in 2016. The studied xanthanolides in this work share some structural similarities with several nonclassical CBSIs, which consist in a planar structure and tend to locate deeper into the β-subunit making use of zones 2 and 3 in the colchicine-binding domain.