Antiproliferative Activity and Molecular Docking of Novel Double-Modified Colchicine Derivatives

Microtubules are tubulin polymer structures, which are indispensable for cell growth and division. Its constituent protein β-tubulin has been a common drug target for various diseases including cancer. Colchicine has been used to treat gout, but it has also been an investigational anticancer agent with a known antimitotic effect on cells. However, the use of colchicine as well as many of its derivatives in long-term treatment is hampered by their high toxicity. To create more potent anticancer agents, three novel double-modified colchicine derivatives have been obtained by structural modifications in C-4 and C-10 positions. The binding affinities of these derivatives of colchicine with respect to eight different isotypes of human β-tubulin have been calculated using docking methods. In vitro cytotoxicity has been evaluated against four human tumor cell lines (A549, MCF-7, LoVo and LoVo/DX). Computer simulations predicted the binding modes of these compounds and hence the key residues involved in the interactions between tubulin and the colchicine derivatives. Two of the obtained derivatives, 4-bromothiocolchicine and 4-iodothiocolchicine, were shown to be active against three of the investigated cancer cell lines (A549, MCF-7, LoVo) with potency at nanomolar concentrations and a higher relative affinity to tumor cells over normal cells.


Introduction
Microtubules, present in all eukaryotic cells, are cylindrical polymers composed of α/β-tubulin heterodimers. They are involved in a wide range of key cellular processes, such as the maintenance of cellular morphology and the active motor transport of cellular components throughout the cytoplasm [1]. Another essential role microtubules play is the formation of mitotic spindles and force generation during mitosis with the purpose of separating chromosomes [2]. A failure within this mitotic spindle apparatus leads to mitotic arrest and eventually apoptosis. This results in cell death, which is a desirable outcome for cancer cells, but not for healthy tissues. With the objective of promoting the former and avoiding the latter effect, microtubules have become the target for a large number of antimitotic agents that act by either favoring or inhibiting microtubule polymerization by binding at specific sites on the exposed surface of α/β-tubulin heterodimers [3][4][5][6][7]

Synthesis of 3
A mixture of N-chlorosuccinimide (175 mg, 1.31 mmol) and 1 (500 mg, 1.25 mmol) in acetonitrile was stirred at RT under nitrogen atmosphere for the 72 h. Reaction time was determined by TLC. The reaction was quenched with saturated aqueous Na 2 S 2 O 3 . The whole mixture was extracted four times with CH 2 Cl 2 , and the combined organic layers were dried over MgSO 4 , filtered, and evaporated under reduced pressure. The residue was purified by CombiFlash ® (EtOAc/MeOH, increasing concentration gradient) to give 3 with yield 75% [32].

Synthesis of 5
A mixture of N-bromosuccinimide (279 mg, 1.57 mmol) and 1 (500 mg, 1.25 mmol) in acetonitrile was stirred at RT under nitrogen atmosphere for the 72 h. Reaction time was determined by TLC. The reaction was quenched with saturated aqueous Na 2 S 2 O 3 . The whole mixture was extracted four times with CH 2 Cl 2 , and the combined organic layers were dried over MgSO 4 , filtered, and evaporated under reduced pressure. The residue was purified by CombiFlash ® (EtOAc/MeOH, increasing concentration gradient) to give 5 with yield 95% [32]. 1

Synthesis of 7
A mixture of N-iodosuccinimide (560 mg, 2.49 mmol) and 1 (500 mg, 1.25 mmol) in AcOH was stirred at 70 • C under nitrogen atmosphere for the 20 h. Reaction time was determined by TLC. The reaction was quenched with saturated aqueous Na 2 S 2 O 3 . The whole mixture was extracted four times with CH 2 Cl 2 , and the combined organic layers were dried over MgSO 4 , filtered, and evaporated under reduced pressure. The residue was purified by CombiFlash ® (EtOAc/MeOH, increasing concentration gradient) to give 7 with yield 95% [32]. 1

Antiproliferative Activity of Colchicine and Its Derivatives
Four human cancer cell lines and one murine normal cell line were used to evaluate antiproliferative activity of colchicine and its derivatives: human lung adenocarcinoma (A549), human breast adenocarcinoma (MCF-7), human colon adenocarcinoma cell lines sensitive and resistant to doxorubicin (LoVo) and (LoVo/DX) respectively, and normal murine embryonic fibroblast cell line (BALB/3T3). The BALB/3T3 cell line was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA), A549 and MCF-7 cell lines-from European Collection of Authenticated Cell Cultures (Salisbury, UK), LoVo cell line was purchased from the ATCC (ATCC, Manassas, VA, USA), and LoVo/DX by courtesy of Prof. E. Borowski (Technical University of Gdańsk, Gdańsk, Poland). All the cell lines are maintained in the Institute of Immunology and Experimental Therapy (IIET), Wroclaw, Poland. Human lung adenocarcinoma cell line was cultured in mixture of OptiMEM and RPMI 1640 (1:1) medium (IIET, Wroclaw, Poland), supplemented with 5% fetal bovine serum (GE Healthcare, Logan, UT, USA) and 2 mM L-glutamine (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA). Human breast adenocarcinoma cell line was cultured in mixture of Eagle medium (IIET, Wroclaw, Poland), supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 8 µg/mL insulin and 1% amino-acids (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA). Human colon adenocarcinoma cell lines were cultured in mixture of OptiMEM and RPMI 1640 (1:1) medium (IIET, Wroclaw, Poland), supplemented with 5% fetal bovine serum (GE Healthcare, Logan UT, USA), 2 mM L-glutamine, 1 mM sodium pyruvate (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA) and 10 µg/100 mL doxorubicin for LoVo/DX (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA). Murine embryonic fibroblast cells were cultured in Dulbecco medium (Life Technologies Limited, Paisley, UK), supplemented with 10% fetal bovine serum (GE Healthcare, Logan, UT, USA) and 2 mM glutamine (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA). All culture media contained antibiotics: 100 U/mL penicillin and 100 µg/mL streptomycin (Polfa-Tarchomin, Warsaw, Poland). All cell lines were cultured during entire experiment in humid atmosphere at 37 • C and 5% CO 2 . Cells were tested for mycoplasma contamination by mycoplasma detection kit for conventional PCR: Venor GeM Classic (Minerva Biolabs GmbH, Berlin, Germany) and negative results was obtained. The procedure is repeated every year or in the case of less frequently used lines: after thawing.

The Antiproliferative Assays In Vitro
Twenty-four hours before adding the tested compounds, all cell lines were seeded in 96-well plates (Sarstedt, Nümbrecht, Germany) in appropriate media with 10 4 cells per well. All cell lines were exposed to each tested agent at four different concentrations in the range 100-0.01 µg/mL for 72 h. Cells were also exposed to the reference drug cisplatin (Teva Pharmaceuticals Polska, Warsaw, Poland) and doxorubicin (Accord Healthcare Limited, Middlesex, UK). Additionally, all cell lines were exposed to DMSO (solvent used for tested compounds) (POCh, Gliwice, Poland) at concentrations corresponding to those present in tested agents' dilutions. After 72 h sulforhodamine B assay (SRB) was performed [37].

SRB
After 72 h of incubation with the tested compounds, cells were fixed in situ by gently adding of 50 µL per well of cold 50% trichloroacetic acid TCA (POCh, Gliwice, Poland) and were incubated at 4 • C for one hour. Following, wells were washed four times with water and air dried. Next, 50 µL of 0.1% solution of sulforhodamine B (Sigma-Aldrich, Merck KGaA, Saint Louis, MO, USA) in 1% acetic acid (POCh, Gliwice, Poland) were added to each well and plates were incubated at room temperature for 0.5 h. After incubation time, unbound dye was removed by washing plates four times with 1% acetic acid whereas stain bound to cells was solubilized with 10 mM Tris base (Sigma-Aldrich, Steinheim, Germany). Absorbance of each solution was read at Synergy H4 Hybrid Multi-Mode Microplate Reader (BioTek Instruments, Inc., Winooski, VT, USA) at the 540 nm wavelength.
Results are presented as mean IC 50 (concentration of the tested compound, that inhibits cell proliferation by 50%) ± standard deviation. IC 50 values were calculated in Cheburator 0.4, Dmitry Nevozhay software (version 1.2.0 software by Dmitry Nevozhay, 2004-2014, http://www.cheburator. nevozhay.com, freely available) for each experiment [38]. Compounds at each concentration were tested in triplicates in single experiment and each experiment was repeated at least three times independently. Results are summarized in Table 1. The Resistance Index (RI) was defined as the ratio of IC 50 for a given compound calculated for resistant cell line to that measured for its parental drug sensitive cell line (Table 1).

Molecular Docking Simulations
A combination of different theoretical methods was used to explore ligand-tubulin interactions. The ligand structures were first minimized and then fully optimized based on the RHF/cc-pVDZ level of theory in GAMESS-US version 2010-10-01. Since there is no crystal structure available for human βI tubulin (TBB5_HUMAN), we obtained its sequence from UniProt (ID: Q13509). We used the tubulin structure 1SA0.pdb as a template to construct the homology model for βI tubulin using MOE2015. We then docked the small library of colchicine derivatives to the protein using the AutoDock4 program under flexible ligand and rigid receptor conditions ( Table 2). AutoDock4 software (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) is designed to predict how drug candidates bind to a receptor of a known 3D structure and consists of two main programs: AutoDock performs the docking of the ligand to a set of grids describing the target protein; AutoGrid pre-calculates these grids. The estimated Moriguchi octanol-water partition coefficient, MlogP, of the compounds were calculated by ADMET Predictor 8.0 (ADMET Predictor, Simulations Plus, Lancaster, CA, USA).

In Vitro Determination of Drug-Induced Inhibition of Human Cancer Cell Line Growth
The synthesized colchicine derivatives (2)(3)(4)(5)(6)(7)(8) and starting material (1) were evaluated for their in vitro antiproliferative effect on normal and cancer cells. Each compound was tested on four human cancer cell lines, including one cell line displaying various level of drug resistance, namely human lung adenocarcinoma (A549), human breast adenocarcinoma (MCF-7), human colon adenocarcinoma cell line (LoVo) and doxorubicin-resistant subline (LoVo/DX). The antiproliferative effect was also studied on normal murine embryonic fibroblast cell line (BALB/3T3) for better description of cytotoxic activity of the compounds studied. The mean values of IC50 ± SD of the tested compounds are collected in Table 1. To evaluate the agents' activity against the cells with MDR (multidrug resistance) phenotype, one drug resistant cancer cell line, i.e., LoVo/DX, was tested and the indexes of resistance (RI) were calculated (see Table 1). The RI values indicate how many times more resistant is the subline in comparison to its parental cell line.
All obtained derivatives with single modification at either the C-4 or C-10 position as well as double-modified compounds showed better antiproliferative activity against all tested cancer cell lines than unmodified 1 and some common chemotherapeutics such as doxorubicin and cisplatin. The IC50 values of novel 4-halothiocolchines are better than for single-modified colchicines in C-4 positions and remain at a level similar to the cytotoxicity of 2 for the A549 and MCF-7 cell lines.
As many as three of the compounds tested on the LoVo cell line (6)(7)(8), including two novel double-modified derivatives (6,8), exhibited extremely high activity (IC50 = 0.007-0.014 μM), which is even better than the activity of 2 (IC50 = 0.021 μM). During the tests on the doxorubicin-resistant subline (LoVo/DX), compounds 4 and 6 showed the best activity among all tested compounds. However, the RI values of the tested compounds indicated that colchicines did not break the drug resistance of LoVo/DX (RI = 9.64-278). Comparison between the cancer cell lines and the normal cell line (BALB/3T3) was made to define the Selectivity Index (SI) as a measure of therapeutic potential. This parameter seems to be especially important for drug-like molecules based on a scaffold of a toxic compound. The SI values showed that compounds 2, 6

In Vitro Determination of Drug-Induced Inhibition of Human Cancer Cell Line Growth
The synthesized colchicine derivatives (2-8) and starting material (1) were evaluated for their in vitro antiproliferative effect on normal and cancer cells. Each compound was tested on four human cancer cell lines, including one cell line displaying various level of drug resistance, namely human lung adenocarcinoma (A549), human breast adenocarcinoma (MCF-7), human colon adenocarcinoma cell line (LoVo) and doxorubicin-resistant subline (LoVo/DX). The antiproliferative effect was also studied on normal murine embryonic fibroblast cell line (BALB/3T3) for better description of cytotoxic activity of the compounds studied. The mean values of IC 50 ± SD of the tested compounds are collected in Table 1. To evaluate the agents' activity against the cells with MDR (multidrug resistance) phenotype, one drug resistant cancer cell line, i.e., LoVo/DX, was tested and the indexes of resistance (RI) were calculated (see Table 1). The RI values indicate how many times more resistant is the subline in comparison to its parental cell line.
All obtained derivatives with single modification at either the C-4 or C-10 position as well as double-modified compounds showed better antiproliferative activity against all tested cancer cell lines than unmodified 1 and some common chemotherapeutics such as doxorubicin and cisplatin. The IC 50 values of novel 4-halothiocolchines are better than for single-modified colchicines in C-4 positions and remain at a level similar to the cytotoxicity of 2 for the A549 and MCF-7 cell lines. Table 1. Antiproliferative activity (IC 50 ) of colchicine (1) and its derivatives (2-8) compared with antiproliferative activity of standard anticancer drugs doxorubicin and cisplatin and the calculated values of the resistance index (RI) and selectivity index (SI) of tested compounds [19,39]. The IC 50 value is defined as the concentration of a compound at which 50% growth inhibition is observed. Human lung adenocarcinoma (A549), human breast adenocarcinoma (MCF-7), human colon adenocarcinoma cell line (LoVo) and doxorubicin-resistant subline (LoVo/DX), normal murine embryonic fibroblast cell line (BALB/3T3). The SI (Selectivity Index) was calculated for each compound using the formula: SI = IC 50 for normal cell line BALB/3T3/IC 50 for respective cancerous cell line. A beneficial SI > 1.0 indicates a drug with efficacy against tumor cells greater than the toxicity against normal cells. The RI (Resistance Index) indicates how many times a resistant subline is chemoresistant relative to its parental cell line. The RI was calculated for each compound using the formula: RI = IC 50 for LoVo/DX/IC 50 for LoVo cell line. When RI is 0-2, the cells are sensitive to the compound tested, RI in the range 2-10 means that the cell shows moderate sensitivity to the drug tested, RI above 10 indicates strong drug resistance.

Compound
As many as three of the compounds tested on the LoVo cell line (6)(7)(8), including two novel double-modified derivatives (6,8), exhibited extremely high activity (IC 50 = 0.007-0.014 µM), which is even better than the activity of 2 (IC 50 = 0.021 µM). During the tests on the doxorubicin-resistant subline (LoVo/DX), compounds 4 and 6 showed the best activity among all tested compounds. However, the RI values of the tested compounds indicated that colchicines did not break the drug resistance of LoVo/DX (RI = 9.64-278). Comparison between the cancer cell lines and the normal cell line (BALB/3T3) was made to define the Selectivity Index (SI) as a measure of therapeutic potential. This parameter seems to be especially important for drug-like molecules based on a scaffold of a toxic compound. The SI values showed that compounds 2, 6 and 8 mostly targeted cancer cells, and fewer targeted normal cells (SI = 10.08-10.45, SI = 6.76-11.85, SI = 5.45-16.43 for A549, MCF-7, LoVo cancer cell lines, respectively). Also compounds 3 and 5 indicated good SI values for MCF-7 cell line (SI = 6.00, SI = 5.26, respectively), as well as compound 7 indicated good SI value for LoVo cell line (SI = 13.5).

Molecular Docking: in Silico Determination of Drug-Induced Inhibition of βI Tubulin
To further investigate the ability to inhibit tubulin aggregation by the new colchicine derivatives in cancer cell growth assays, binding energies between the new compounds and βI tubulin, one of the subunits of microtubules in the cytoskeleton structure of every eukaryotic cell, were calculated using docking methodology. The eight structures of colchicine and its derivatives described above were docked into the βI tubulin CBS and ranked according to their binding affinity ( Table 2).
Since the binding energy shows how strong the interaction between the distributed drug in the cell and β-tubulin protein can be, the partition coefficient (MlogP) values were calculated and considered for the ability of the drugs to diffuse into the cells ( Table 2). The MlogP values can be a beneficial factors in estimating and comparing the distribution of the novel drugs within the cells, organs and the body [40].
Based on our computational predictions, compounds 6, 4, 5 and 3 show the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. In the experimental part of the study, the lowest IC 50 values showed 4, 6, 8 compounds having higher activity than 3, 5, 7 (Table 1). To investigate in more detail why the computational results did not show better binding energies of compound 8 than 3 and 5, the interactions between double-modified colchicine derivatives and the CBS of βI tubulin were studied.
As shown in the diagrams representing schematic interactions of the 4 and 6 compounds with the CBS of βI tubulin Lys 352, Met 259 and Asn 258 residues interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of thiomethyl group on ring C and hydrogen of C-11 on ring C (sidechain donor) and with ring C (arene-H), respectively.
However, Lys 352, Met 259 and Val 315 residues of the CBS of βI tubulin interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of C-11 on ring C (sidechain donor) and with hydrogens of methoxyl groups on ring C (backbone donors) of the 3 and 5 compounds, respectively.

Molecular Docking: in Silico Determination of Drug-Induced Inhibition of βI Tubulin
To further investigate the ability to inhibit tubulin aggregation by the new colchicine derivatives in cancer cell growth assays, binding energies between the new compounds and βІ tubulin, one of the subunits of microtubules in the cytoskeleton structure of every eukaryotic cell, were calculated using docking methodology. The eight structures of colchicine and its derivatives described above were docked into the βІ tubulin CBS and ranked according to their binding affinity ( Table 2).
Since the binding energy shows how strong the interaction between the distributed drug in the cell and β-tubulin protein can be, the partition coefficient (MlogP) values were calculated and considered for the ability of the drugs to diffuse into the cells ( Table 2). The MlogP values can be a beneficial factors in estimating and comparing the distribution of the novel drugs within the cells, organs and the body [40].
Based on our computational predictions, compounds 6, 4, 5 and 3 show the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. In the experimental part of the study, the lowest IC50 values showed 4, 6, 8 compounds having higher activity than 3, 5, 7 (Table 1). To investigate in more detail why the computational results did not show better binding energies of compound 8 than 3 and 5, the interactions between double-modified colchicine derivatives and the CBS of βІ tubulin were studied.
As shown in the diagrams representing schematic interactions of the 4 and 6 compounds with the CBS of βІ tubulin Lys 352, Met 259 and Asn 258 residues interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of thiomethyl group on ring C and hydrogen of C-11 on ring C (sidechain donor) and with ring C (arene-H), respectively.
However, Lys 352, Met 259 and Val 315 residues of the CBS of βІ tubulin interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of C-11 on ring C (sidechain donor) and with hydrogens of methoxyl groups on ring C (backbone donors) of the 3 and 5 compounds, respectively.
The study was continued for compound 8 with the highest binding energy for better understanding over the effect of interaction of βІ tubulin's residues with novel derivatives on better activities of the derivatives. Compound 8 interacts with Ala317 residue of βІ tubulin, iodine on ring A (backbone donor) and Cys241 residue of βІ tubulin, hydrogen of methoxyl group on ring A (sidechain donor) and Ala 316 residue of βІ tubulin, the oxygen of carbonyl on ring B (sidechain acceptor).

Molecular Docking: in Silico Determination of Drug-Induced Inhibition of βI Tubulin
To further investigate the ability to inhibit tubulin aggregation by the new colchicine derivatives in cancer cell growth assays, binding energies between the new compounds and βІ tubulin, one of the subunits of microtubules in the cytoskeleton structure of every eukaryotic cell, were calculated using docking methodology. The eight structures of colchicine and its derivatives described above were docked into the βІ tubulin CBS and ranked according to their binding affinity ( Table 2).
Since the binding energy shows how strong the interaction between the distributed drug in the cell and β-tubulin protein can be, the partition coefficient (MlogP) values were calculated and considered for the ability of the drugs to diffuse into the cells ( Table 2). The MlogP values can be a beneficial factors in estimating and comparing the distribution of the novel drugs within the cells, organs and the body [40].
Based on our computational predictions, compounds 6, 4, 5 and 3 show the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. In the experimental part of the study, the lowest IC50 values showed 4, 6, 8 compounds having higher activity than 3, 5, 7 (Table 1). To investigate in more detail why the computational results did not show better binding energies of compound 8 than 3 and 5, the interactions between double-modified colchicine derivatives and the CBS of βІ tubulin were studied.
As shown in the diagrams representing schematic interactions of the 4 and 6 compounds with the CBS of βІ tubulin Lys 352, Met 259 and Asn 258 residues interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of thiomethyl group on ring C and hydrogen of C-11 on ring C (sidechain donor) and with ring C (arene-H), respectively.
However, Lys 352, Met 259 and Val 315 residues of the CBS of βІ tubulin interact with oxygen of the carbonyl of the C ring (sidechain acceptor), with hydrogen of C-11 on ring C (sidechain donor) and with hydrogens of methoxyl groups on ring C (backbone donors) of the 3 and 5 compounds, respectively.
The study was continued for compound 8 with the highest binding energy for better understanding over the effect of interaction of βІ tubulin's residues with novel derivatives on better activities of the derivatives. Compound 8 interacts with Ala317 residue of βІ tubulin, iodine on ring A (backbone donor) and Cys241 residue of βІ tubulin, hydrogen of methoxyl group on ring A (sidechain donor) and Ala 316 residue of βІ tubulin, the oxygen of carbonyl on ring B (sidechain acceptor).  Since the IC50 is a cell-based assay and the βІ tubulin isotype is not the only isotype of tubulin expressed in the cell, we decided to check interactions of the synthesized compounds with the other tubulin isotypes present in the referred cell lines. Tubulin isotypes are highly conserved in all mammals as discussed by Luduena [41]. As is commonly the case, both normal and cancer cells in humans contain the same tubulin isotypes. However, their expression levels differ and specifically βІII (TUBB3) tubulin is very narrowly distributed in normal cells while it is almost always found in cancer cells and it is often correlated with drug resistance [42][43][44]. Furthermore, βІ isotype (TUBB) is the most abundant isotype in most tumors, followed by, βІVb (TUBB4B), βІIa (TUBB2A), βV (TUBB6), and βІII (TUBB3) with 47%, 38%, 8.9%, 3.1%, and 2.2% respectively and with βІVa (TUBB4), βІIb (TUBB2B) and βVI (TUBB1) levels below 0.5% of the total β tubulin [42]. Interestingly, the binding sites for common tubulin-binding agents do not vary significantly between tubulin isotypes except for the CBS [8].
Tubulin isotype classes βІVa and βІVb comprised more than 50% of the total β tubulin in breast cancer (MCF-7) and colon cancer cells. The expression of βІ, βІII and βІV (a and b) in the MCF-7 cell line has been reported as 39.1%, 2.5% and 58.4% respectively [45]. In another study, the ratio for MCF-7 cell line was reported as 55% to 6% to 39% [46]. In colon cancer cells, the percentage expression of the referred tubulin isotypes is 61.8%, 0.2% and 38%, respectively. The ratio of the isotype expression in lung cancer cells (A549) is given as 71.9% to 1.6% to 26.5% [45]. Consequently, each of the cell lines investigated has a distinct distribution of tubulin isotypes, which affects the overall response to a cytotoxic agent whose affinity for each of these isotypes is different making the problem of computational prediction complex.
To quantify the assumption that these compounds have different binding free energies for each of the tubulin isotypes, docking simulations between the novel colchicine derivatives and βІ (UniProt ID: P07437), βІIa (UniProt ID: Q13885), βІIb (UniProt ID: Q9BVA1), βІII (UniProt ID: Q13509), βІVa (UniProt ID: P04350), βІVb (UniProt ID: P68371), βV (UniProt ID: Q9BUF5), and βVI (UniProt ID: Q9H4B7), isotypes were performed. Tubulin structure 1SA0.pdb was used as the homology model template for all tubulin isotypes using MOE2015. To visualize the results, a heat map was prepared to better illustrate the comparison of binding energies between the investigated compounds and the different tubulin isotypes using AutoDock Tableau Desktop (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) (see Table 3). Since the IC50 is a cell-based assay and the βІ tubulin isotype is not the only isotype of tubulin expressed in the cell, we decided to check interactions of the synthesized compounds with the other tubulin isotypes present in the referred cell lines. Tubulin isotypes are highly conserved in all mammals as discussed by Luduena [41]. As is commonly the case, both normal and cancer cells in humans contain the same tubulin isotypes. However, their expression levels differ and specifically βІII (TUBB3) tubulin is very narrowly distributed in normal cells while it is almost always found in cancer cells and it is often correlated with drug resistance [42][43][44]. Furthermore, βІ isotype (TUBB) is the most abundant isotype in most tumors, followed by, βІVb (TUBB4B), βІIa (TUBB2A), βV (TUBB6), and βІII (TUBB3) with 47%, 38%, 8.9%, 3.1%, and 2.2% respectively and with βІVa (TUBB4), βІIb (TUBB2B) and βVI (TUBB1) levels below 0.5% of the total β tubulin [42]. Interestingly, the binding sites for common tubulin-binding agents do not vary significantly between tubulin isotypes except for the CBS [8].
Tubulin isotype classes βІVa and βІVb comprised more than 50% of the total β tubulin in breast cancer (MCF-7) and colon cancer cells. The expression of βІ, βІII and βІV (a and b) in the MCF-7 cell line has been reported as 39.1%, 2.5% and 58.4% respectively [45]. In another study, the ratio for MCF-7 cell line was reported as 55% to 6% to 39% [46]. In colon cancer cells, the percentage expression of the referred tubulin isotypes is 61.8%, 0.2% and 38%, respectively. The ratio of the isotype expression in lung cancer cells (A549) is given as 71.9% to 1.6% to 26.5% [45]. Consequently, each of the cell lines investigated has a distinct distribution of tubulin isotypes, which affects the overall response to a cytotoxic agent whose affinity for each of these isotypes is different making the problem of computational prediction complex.
To quantify the assumption that these compounds have different binding free energies for each of the tubulin isotypes, docking simulations between the novel colchicine derivatives and βІ (UniProt ID: P07437), βІIa (UniProt ID: Q13885), βІIb (UniProt ID: Q9BVA1), βІII (UniProt ID: Q13509), βІVa (UniProt ID: P04350), βІVb (UniProt ID: P68371), βV (UniProt ID: Q9BUF5), and βVI (UniProt ID: Q9H4B7), isotypes were performed. Tubulin structure 1SA0.pdb was used as the homology model template for all tubulin isotypes using MOE2015. To visualize the results, a heat map was prepared to better illustrate the comparison of binding energies between the investigated compounds and the different tubulin isotypes using AutoDock Tableau Desktop (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) (see Table 3). Since the IC50 is a cell-based assay and the βІ tubulin isotype is not the only isotype of tubulin expressed in the cell, we decided to check interactions of the synthesized compounds with the other tubulin isotypes present in the referred cell lines. Tubulin isotypes are highly conserved in all mammals as discussed by Luduena [41]. As is commonly the case, both normal and cancer cells in humans contain the same tubulin isotypes. However, their expression levels differ and specifically βІII (TUBB3) tubulin is very narrowly distributed in normal cells while it is almost always found in cancer cells and it is often correlated with drug resistance [42][43][44]. Furthermore, βІ isotype (TUBB) is the most abundant isotype in most tumors, followed by, βІVb (TUBB4B), βІIa (TUBB2A), βV (TUBB6), and βІII (TUBB3) with 47%, 38%, 8.9%, 3.1%, and 2.2% respectively and with βІVa (TUBB4), βІIb (TUBB2B) and βVI (TUBB1) levels below 0.5% of the total β tubulin [42]. Interestingly, the binding sites for common tubulin-binding agents do not vary significantly between tubulin isotypes except for the CBS [8].
Tubulin isotype classes βІVa and βІVb comprised more than 50% of the total β tubulin in breast cancer (MCF-7) and colon cancer cells. The expression of βІ, βІII and βІV (a and b) in the MCF-7 cell line has been reported as 39.1%, 2.5% and 58.4% respectively [45]. In another study, the ratio for MCF-7 cell line was reported as 55% to 6% to 39% [46]. In colon cancer cells, the percentage expression of the referred tubulin isotypes is 61.8%, 0.2% and 38%, respectively. The ratio of the isotype expression in lung cancer cells (A549) is given as 71.9% to 1.6% to 26.5% [45]. Consequently, each of the cell lines investigated has a distinct distribution of tubulin isotypes, which affects the overall response to a cytotoxic agent whose affinity for each of these isotypes is different making the problem of computational prediction complex.
To quantify the assumption that these compounds have different binding free energies for each of the tubulin isotypes, docking simulations between the novel colchicine derivatives and βІ (UniProt ID: P07437), βІIa (UniProt ID: Q13885), βІIb (UniProt ID: Q9BVA1), βІII (UniProt ID: Q13509), βІVa (UniProt ID: P04350), βІVb (UniProt ID: P68371), βV (UniProt ID: Q9BUF5), and βVI (UniProt ID: Q9H4B7), isotypes were performed. Tubulin structure 1SA0.pdb was used as the homology model template for all tubulin isotypes using MOE2015. To visualize the results, a heat map was prepared to better illustrate the comparison of binding energies between the investigated compounds and the different tubulin isotypes using AutoDock Tableau Desktop (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) (see Table 3). Since the IC50 is a cell-based assay and the βІ tubulin isotype is not the only isotype of tubulin expressed in the cell, we decided to check interactions of the synthesized compounds with the other tubulin isotypes present in the referred cell lines. Tubulin isotypes are highly conserved in all mammals as discussed by Luduena [41]. As is commonly the case, both normal and cancer cells in humans contain the same tubulin isotypes. However, their expression levels differ and specifically βІII (TUBB3) tubulin is very narrowly distributed in normal cells while it is almost always found in cancer cells and it is often correlated with drug resistance [42][43][44]. Furthermore, βІ isotype (TUBB) is the most abundant isotype in most tumors, followed by, βІVb (TUBB4B), βІIa (TUBB2A), βV (TUBB6), and βІII (TUBB3) with 47%, 38%, 8.9%, 3.1%, and 2.2% respectively and with βІVa (TUBB4), βІIb (TUBB2B) and βVI (TUBB1) levels below 0.5% of the total β tubulin [42]. Interestingly, the binding sites for common tubulin-binding agents do not vary significantly between tubulin isotypes except for the CBS [8].
Tubulin isotype classes βІVa and βІVb comprised more than 50% of the total β tubulin in breast cancer (MCF-7) and colon cancer cells. The expression of βІ, βІII and βІV (a and b) in the MCF-7 cell line has been reported as 39.1%, 2.5% and 58.4% respectively [45]. In another study, the ratio for MCF-7 cell line was reported as 55% to 6% to 39% [46]. In colon cancer cells, the percentage expression of the referred tubulin isotypes is 61.8%, 0.2% and 38%, respectively. The ratio of the isotype expression in lung cancer cells (A549) is given as 71.9% to 1.6% to 26.5% [45]. Consequently, each of the cell lines investigated has a distinct distribution of tubulin isotypes, which affects the overall response to a cytotoxic agent whose affinity for each of these isotypes is different making the problem of computational prediction complex.
To quantify the assumption that these compounds have different binding free energies for each of the tubulin isotypes, docking simulations between the novel colchicine derivatives and βІ (UniProt ID: P07437), βІIa (UniProt ID: Q13885), βІIb (UniProt ID: Q9BVA1), βІII (UniProt ID: Q13509), βІVa (UniProt ID: P04350), βІVb (UniProt ID: P68371), βV (UniProt ID: Q9BUF5), and βVI (UniProt ID: Q9H4B7), isotypes were performed. Tubulin structure 1SA0.pdb was used as the homology model template for all tubulin isotypes using MOE2015. To visualize the results, a heat map was prepared to better illustrate the comparison of binding energies between the investigated compounds and the different tubulin isotypes using AutoDock Tableau Desktop (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) (see Table 3).

−7.33
2.00 Ala316 Ala317 Cys241 The study was continued for compound 8 with the highest binding energy for better understanding over the effect of interaction of βI tubulin's residues with novel derivatives on better activities of the derivatives. Compound 8 interacts with Ala317 residue of βI tubulin, iodine on ring A (backbone donor) and Cys241 residue of βI tubulin, hydrogen of methoxyl group on ring A (sidechain donor) and Ala 316 residue of βI tubulin, the oxygen of carbonyl on ring B (sidechain acceptor).
Since the IC 50 is a cell-based assay and the βI tubulin isotype is not the only isotype of tubulin expressed in the cell, we decided to check interactions of the synthesized compounds with the other tubulin isotypes present in the referred cell lines. Tubulin isotypes are highly conserved in all mammals as discussed by Luduena [41]. As is commonly the case, both normal and cancer cells in humans contain the same tubulin isotypes. However, their expression levels differ and specifically βIII (TUBB3) tubulin is very narrowly distributed in normal cells while it is almost always found in cancer cells and it is often correlated with drug resistance [42][43][44]. Furthermore, βI isotype (TUBB) is the most abundant isotype in most tumors, followed by, βIVb (TUBB4B), βIIa (TUBB2A), βV (TUBB6), and βIII (TUBB3) with 47%, 38%, 8.9%, 3.1%, and 2.2% respectively and with βIVa (TUBB4), βIIb (TUBB2B) and βVI (TUBB1) levels below 0.5% of the total β tubulin [42]. Interestingly, the binding sites for common tubulin-binding agents do not vary significantly between tubulin isotypes except for the CBS [8].
Tubulin isotype classes βIVa and βIVb comprised more than 50% of the total β tubulin in breast cancer (MCF-7) and colon cancer cells. The expression of βI, βIII and βIV (a and b) in the MCF-7 cell line has been reported as 39.1%, 2.5% and 58.4% respectively [45]. In another study, the ratio for MCF-7 cell line was reported as 55% to 6% to 39% [46]. In colon cancer cells, the percentage expression of the referred tubulin isotypes is 61.8%, 0.2% and 38%, respectively. The ratio of the isotype expression in lung cancer cells (A549) is given as 71.9% to 1.6% to 26.5% [45]. Consequently, each of the cell lines investigated has a distinct distribution of tubulin isotypes, which affects the overall response to a cytotoxic agent whose affinity for each of these isotypes is different making the problem of computational prediction complex.
To quantify the assumption that these compounds have different binding free energies for each of the tubulin isotypes, docking simulations between the novel colchicine derivatives and βI (UniProt ID: P07437), βIIa (UniProt ID: Q13885), βIIb (UniProt ID: Q9BVA1), βIII (UniProt ID: Q13509), βIVa (UniProt ID: P04350), βIVb (UniProt ID: P68371), βV (UniProt ID: Q9BUF5), and βVI (UniProt ID: Q9H4B7), isotypes were performed. Tubulin structure 1SA0.pdb was used as the homology model template for all tubulin isotypes using MOE2015. To visualize the results, a heat map was prepared to better illustrate the comparison of binding energies between the investigated compounds and the different tubulin isotypes using AutoDock Tableau Desktop (version 2018.2.0, Tableau Research, Standford University, Seattle, WA, USA) (see Table 3). Table 3. Heat map of binding energies interactions between the eight different colchicine derivatives and βI, βIII βIIa, βIIb, βIVa, βIVb, βV and βVI tubulin isotypes.  As can be seen in the heat map above, the binding energy between compound 8 and tubulin isotype βIIa as well as βVI are good examples of high binding energies while for compounds 3 and 5 their interaction with βІ tubulin and βIIa dominates. For compound 7 βVI, βIIa and βIII are the strongest binding tubulin isotypes. These differences might be the reasons of a discrepancy between experimental and computational data. However, even data concerning the level of tubulin isotypes expression reported for the same cell line in the literature, differ from each other. Precise levels of isotype expression are not only hard to determine but they fluctuate in the same cell line as a result of exposure to various drugs [47]. Therefore, more detailed experimental studies on the isotype expression among different cell lines are still needed to give a better insight in this issue in the future.
In terms of a mechanistic explanation, specific residues of βIIa tubulin that are involved in interactions with compound 8 are Asn 256 that interacts with C-20 on ring C (backbone donor) and with ring C (arene-H) and Cys 239 that interacts with hydrogen of methoxyl group on ring A (sidechain donor). The residues of βVI tubulin that are involved in interactions with compound 8 are Asn256, which interacts with carbon atoms of two methoxyl groups on ring A (sidechain donors), Ala248, interacts with the oxygen of carbonyl on ring C (sidechain acceptor) and Asn247 that interacts with C-20 on ring C (sidechain donor) (see Table 4).
The analysis of interactions between compounds 4, 6 and βІ and compound 8 and βІIa and βVI shows that an arene-H interaction between ring C Asn256 or Asn258 and a sidechain acceptor interaction between the oxygen of carbonyl on ring C and either Lys352 or Ala248 can result in a strong binding effect. As mentioned before, the probability of the expression of βIIa in most tumors is approximately 9% versus less than 0.5% for each isotype in the group of βIVa, βVI and βIIb [42].
To the best of the authors' knowledge of the literature, ligands binding to alpha and β tubulin are exclusive, except of course of ATP, and therefore, we do not expect any cross-interactions of our compounds with alpha tubulin [48]. Concerning interactions with ABC transporters, it is quite possible that our compounds are substrates for these multidrug resistance enzymes, but this is common to many chemotherapy drugs, including taxanes. Therefore, in order to inhibit this interaction, it would most advantageous to use our compounds in combination with some of their modulators, e.g., verapamil [49].
As can be seen in the heat map above, the binding energy between compound 8 and tubulin isotype βIIa as well as βVI are good examples of high binding energies while for compounds 3 and 5 their interaction with βI tubulin and βIIa dominates. For compound 7 βVI, βIIa and βIII are the strongest binding tubulin isotypes. These differences might be the reasons of a discrepancy between experimental and computational data. However, even data concerning the level of tubulin isotypes expression reported for the same cell line in the literature, differ from each other. Precise levels of isotype expression are not only hard to determine but they fluctuate in the same cell line as a result of exposure to various drugs [47]. Therefore, more detailed experimental studies on the isotype expression among different cell lines are still needed to give a better insight in this issue in the future.
In terms of a mechanistic explanation, specific residues of βIIa tubulin that are involved in interactions with compound 8 are Asn 256 that interacts with C-20 on ring C (backbone donor) and with ring C (arene-H) and Cys 239 that interacts with hydrogen of methoxyl group on ring A (sidechain donor). The residues of βVI tubulin that are involved in interactions with compound 8 are Asn256, which interacts with carbon atoms of two methoxyl groups on ring A (sidechain donors), Ala248, interacts with the oxygen of carbonyl on ring C (sidechain acceptor) and Asn247 that interacts with C-20 on ring C (sidechain donor) (see Table 4).
The analysis of interactions between compounds 4, 6 and βI and compound 8 and βIIa and βVI shows that an arene-H interaction between ring C Asn256 or Asn258 and a sidechain acceptor interaction between the oxygen of carbonyl on ring C and either Lys352 or Ala248 can result in a strong binding effect. As mentioned before, the probability of the expression of βIIa in most tumors is approximately 9% versus less than 0.5% for each isotype in the group of βIVa, βVI and βIIb [42].
To the best of the authors' knowledge of the literature, ligands binding to alpha and β tubulin are exclusive, except of course of ATP, and therefore, we do not expect any cross-interactions of our compounds with alpha tubulin [48]. Concerning interactions with ABC transporters, it is quite possible that our compounds are substrates for these multidrug resistance enzymes, but this is common to many chemotherapy drugs, including taxanes. Therefore, in order to inhibit this interaction, it would most advantageous to use our compounds in combination with some of their modulators, e.g., verapamil [49].

Discussion
We synthesized three novel double-modified 4-halothiocolchicines (4, 6, 8) and evaluated their biological activity according to the in vitro antiproliferative tests as well as the molecular docking. For a better comparison, also the activity of single-modified colchicine derivatives (2, 3, 5, 7) as well as colchicine itself (1) was evaluated on four human cancer cell lines and normal murine embryonic fibroblast cell line. The results of our study clearly showed that the antiproliferative activity of novel 4-halothiocolchines (4, 6, 8) is better than the activity of 4-halocolchicines (3, 5, 7) and remain at a level similar to the cytotoxicity of 2 for the A549, MCF-7 and LoVo cell lines. Furthermore, the cytotoxicity of compounds 4, 6 and 8 is higher than cytotoxicity of unmodified colchicine (1) and commonly used chemotherapeutics such as doxorubicin and cisplatin.
The introduction of thiomethyl group in C-10 position significantly increased the cytotoxicity in comparison to single-modified 4-halo derivatives (3, 5, 7) as well as allowed to reduce the toxicity for 4-bromo and 4-iodo derivatives. Compounds 4-bromothiocolchine (6) and 4-iodothiocolchicine (8) proved to be less toxic to normal murine fibroblast cells than the currently used anticancer drugs, such as cisplatin and doxorubicin, which is confirmed by their high SI values. The appropriate modification of colchicine molecule and synthesis of its analogs might overcome the toxicity, which is a major challenge in designing a potential colchicine-based drug candidate. However, it is still challenging to draw clear conclusions from the molecular-level calculations. Compounds 6, 4, 5 and 3 showed the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. The results only partially correlate with in vitro determined IC50 values. This may be explained by several additional effects taking place in living cells compared to the computational simulations that focus only on the binding mode of the compounds to the target. Specifically, off-target interactions involving efflux pumps with different affinities for the individual compounds may explain the observed partial correlation between IC50 values and binding free energies. Additionally, differences in the solubility values and membrane permeability may have to be accounted for when ranking the various compounds in biological assays and comparing them to computational predictions based on binding affinity for the target alone. We have partially addressed this issue by performing docking

Discussion
We synthesized three novel double-modified 4-halothiocolchicines (4, 6, 8) and evaluated their biological activity according to the in vitro antiproliferative tests as well as the molecular docking. For a better comparison, also the activity of single-modified colchicine derivatives (2, 3, 5, 7) as well as colchicine itself (1) was evaluated on four human cancer cell lines and normal murine embryonic fibroblast cell line. The results of our study clearly showed that the antiproliferative activity of novel 4-halothiocolchines (4, 6, 8) is better than the activity of 4-halocolchicines (3, 5, 7) and remain at a level similar to the cytotoxicity of 2 for the A549, MCF-7 and LoVo cell lines. Furthermore, the cytotoxicity of compounds 4, 6 and 8 is higher than cytotoxicity of unmodified colchicine (1) and commonly used chemotherapeutics such as doxorubicin and cisplatin.
The introduction of thiomethyl group in C-10 position significantly increased the cytotoxicity in comparison to single-modified 4-halo derivatives (3, 5, 7) as well as allowed to reduce the toxicity for 4-bromo and 4-iodo derivatives. Compounds 4-bromothiocolchine (6) and 4-iodothiocolchicine (8) proved to be less toxic to normal murine fibroblast cells than the currently used anticancer drugs, such as cisplatin and doxorubicin, which is confirmed by their high SI values. The appropriate modification of colchicine molecule and synthesis of its analogs might overcome the toxicity, which is a major challenge in designing a potential colchicine-based drug candidate. However, it is still challenging to draw clear conclusions from the molecular-level calculations. Compounds 6, 4, 5 and 3 showed the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. The results only partially correlate with in vitro determined IC50 values. This may be explained by several additional effects taking place in living cells compared to the computational simulations that focus only on the binding mode of the compounds to the target. Specifically, off-target interactions involving efflux pumps with different affinities for the individual compounds may explain the observed partial correlation between IC50 values and binding free energies. Additionally, differences in the solubility values and membrane permeability may have to be accounted for when ranking the various compounds in biological assays and comparing them to computational predictions based on binding affinity for the target alone. We have partially addressed this issue by performing docking −8. 30 1.99 Asn256 Asn247 Ala248

Discussion
We synthesized three novel double-modified 4-halothiocolchicines (4, 6, 8) and evaluated their biological activity according to the in vitro antiproliferative tests as well as the molecular docking. For a better comparison, also the activity of single-modified colchicine derivatives (2, 3, 5, 7) as well as colchicine itself (1) was evaluated on four human cancer cell lines and normal murine embryonic fibroblast cell line. The results of our study clearly showed that the antiproliferative activity of novel 4-halothiocolchines (4, 6, 8) is better than the activity of 4-halocolchicines (3, 5, 7) and remain at a level similar to the cytotoxicity of 2 for the A549, MCF-7 and LoVo cell lines. Furthermore, the cytotoxicity of compounds 4, 6 and 8 is higher than cytotoxicity of unmodified colchicine (1) and commonly used chemotherapeutics such as doxorubicin and cisplatin.
The introduction of thiomethyl group in C-10 position significantly increased the cytotoxicity in comparison to single-modified 4-halo derivatives (3, 5, 7) as well as allowed to reduce the toxicity for 4-bromo and 4-iodo derivatives. Compounds 4-bromothiocolchine (6) and 4-iodothiocolchicine (8) proved to be less toxic to normal murine fibroblast cells than the currently used anticancer drugs, such as cisplatin and doxorubicin, which is confirmed by their high SI values. The appropriate modification of colchicine molecule and synthesis of its analogs might overcome the toxicity, which is a major challenge in designing a potential colchicine-based drug candidate. However, it is still challenging to draw clear conclusions from the molecular-level calculations. Compounds 6, 4, 5 and 3 showed the lowest binding energies of −8.6, −8.6, −8.4 and −8.3 kcal/mol, respectively. The results only partially correlate with in vitro determined IC 50 values. This may be explained by several additional effects taking place in living cells compared to the computational simulations that focus only on the binding mode of the compounds to the target. Specifically, off-target interactions involving efflux pumps with different affinities for the individual compounds may explain the observed partial correlation between IC 50 values and binding free energies. Additionally, differences in the solubility values and membrane permeability may have to be accounted for when ranking the various compounds in biological assays and comparing them to computational predictions based on binding affinity for the target alone. We have partially addressed this issue by performing docking simulations for the remaining tubulin isotypes, several of them may be expressed in cancer cells in a manner different than in normal cells.
We have demonstrated that a higher affinity for βVI tubulin of the compounds investigated may explain the differences in their biological activities. Our studies clearly show the potential of the obtained double-modified compounds. In particular, 4-halothiocolchicines are worthwhile for a continuation of the search for strong and broad-spectrum anticancer agents. Inspired by these preliminary results we plan subsequent modifications in C-7 position to obtain a series of triple-modified derivatives. Further evaluation should help to find more detailed structure-activity relationships of microtubule-targeting drugs and CBS inhibitors, which can help in rational drug design in the future.