Docking, Synthesis and Antiproliferative Activity of N-Acylhydrazone Derivatives Designed as Combretastatin A4 Analogues

Cancer is the second most common cause of death in the USA. Among the known classes of anticancer agents, the microtubule-targeted antimitotic drugs are considered to be one of the most important. They are usually classified into microtubule-destabilizing (e.g., Vinca alkaloids) and microtubule-stabilizing (e.g., paclitaxel) agents. Combretastatin A4 (CA-4), which is a natural stilbene isolated from Combretum caffrum, is a microtubule-destabilizing agent that binds to the colchicine domain on β-tubulin and exhibits a lower toxicity profile than paclitaxel or the Vinca alkaloids. In this paper, we describe the docking study, synthesis, antiproliferative activity and selectivity index of the N-acylhydrazone derivatives (5a–r) designed as CA-4 analogues. The essential structural requirements for molecular recognition by the colchicine binding site of β-tubulin were recognized, and several compounds with moderate to high antiproliferative potency (IC50 values ≤18 µM and ≥4 nM) were identified. Among these active compounds, LASSBio-1586 (5b) emerged as a simple antitumor drug candidate, which is capable of inhibiting microtubule polymerization and possesses a broad in vitro and in vivo antiproliferative profile, as well as a better selectivity index than the prototype CA-4, indicating improved selective cytotoxicity toward cancer cells.


Introduction
Microtubules (MTs) are cytoskeletal polymers formed by the polymerization of á-and b-tubulin heterodimers, which is followed by GTP hydrolysis; the polymerization occurs through two important steps: nucleation and elongation. MTs are found within all dividing eukaryotic cells, as well as in most differentiated cell types, and play crucial roles in cell division, cell motility, cellular transport, the maintenance of cell polarity, and cell signaling [1].
Microtubules are labile polymers that display two types of dynamic behaviors, which are called ''treadmilling'' and ''dynamic'' instability. The latter, is characterized by the alternating growing and shortening phases of the microtubule ends. The transition from a growing phase to a shortening phase is called a catastrophe, while a transition from a shortening phase to a growing phase is known as a rescue. Because microtubule dynamics play an important role in various cellular functions, such as mitosis, they are a potential target for development of anticancer drugs [1][2][3][4].
Microtubule-targeting antimitotic drugs are usually classified into two main groups. One group, which is composed of microtubule-destabilizing agents, inhibits microtubule polymerization and includes compounds such as the Vinca alkaloids, vincristine (1) and vinblastine (2) (Figure 1); these two compounds were the first anti-microtubule agents approved to treat cancer. The second group encompasses the microtubule-stabilizing agents; these compounds stimulate microtubule polymerization and include paclitaxel, which is used to treat breast and ovarian cancer, non-small-cell lung cancer and Kaposi's sarcoma [4].
While vinblastine binds close to the exchangeable GTP site on the b-tubulin in a region called the Vinca-binding domain, paclitaxel (3, Figure 1) binds to the inner surface of the microtubules in a deep hydrophobic pocket on the btubulin; this site is called the paclitaxel binding site [4][5].
During the development of orally bioavailable anti-microtubule agents that overcome the neurotoxicity and development of resistance commonly observed with the Vinca alkaloids, paclitaxel and their analogues, combretastatin A4 (CA-4, Figure 1) was discovered and is currently considered a promising lead-compound. This stilbene natural product, which was isolated from Combretum caffrum, binds to the colchicine domain on b-tubulin and exhibits a low toxicity profile [6]. Despite its potent antiproliferative activity, CA-4 (4) failed to exhibit anticancer efficacy in animal models because it has low water solubility, poor oral bioavailability, a short half-life and a double bond that isomerizes (Z to E) in vivo; this isomerization causes a loss of affinity for btubulin and consequently a loss of cytotoxic activity [7][8][9][10][11].
This paper describes the docking studies, synthesis and assessment of antiproliferative activity and selectivity index of Nacylhydrazone derivatives (5a-r) designed as CA-4 analogues.
The initial design conception of the N-acylhydrazone derivatives (5a-r) is depicted in Figure 2. The most important structural modification was the replacement of ethylene linker between the aromatic subunits A and B with a more stable N-acylhydrazone (NAH) scaffold, generating compound 5a. To design a congeneric series (5b-r), several modifications were introduced in the substitution of aromatic subunit B based on docking studies with the colchicine binding site of the b-tubulin protein.

Results and Discussion
All designed compounds were predicted to favorably interact with the DAMA-colchicine binding site in b-tubulin (PDB code: 1sa0) [12]. In the best-ranked solutions, there were few polar interactions, and the complementarity between the ligand and the receptor protein involved extensive, nonspecific interactions with hydrophobic groups. These results were in accordance with the DAMA-colchicine interaction mode observed in the co-crystallized structure: there is only one polar interaction, which occurred between the Cys241 SH group and one of the methoxy groups on the ligand (data not shown). Previously, the proximity between these groups was explored to establish a cross-link between the colchicine derivatives substituted at this methoxy position and Cys241 [13]. Combretastatin A4 (CA-4) was also predicted to interact primarily with the hydrophobic groups; its trimethoxy ring (ring A) occupied a similar position to the corresponding colchicine ring, and its second ring (ring B) formed two hydrogen bonds, which were between its phenolic hydroxyl group and Thr179 peptide carbonyl group, as well as between the adjacent methoxy group and Ser178 side chain [7,10,11]. Similar studies performed with the E-isomer CA4 show the loss of interactions with residues Ser178 and Thr179, which may somehow explain the inactivity of this isomer (see Figure S2 in supporting material). Additionally, its N-acylhydrazone analogue, which was LASSBio-1593 (5a), interacted with Ser178 through a methoxy group on ring A, and its isovaline ring (ring B) formed two hydrogen bonds, one with Val238 and the other with Tyr202 ( Figure 3).
Based on the docking studies with compound 5a, several modifications were enacted on the 4-methoxy-3-hydroxy-phenyl moiety (ring B, Figure 2) to vary the oxygenated pattern (5c-j) and explore more lipophilic substituents (5b, 5l-r), while making allowances for the hydrophobic nature of colchicine binding pocket ( Table 1). The modification of the linker between rings A (i.e., 3,4,5-trimethoxyphenyl) and B (i.e., 3-hydroxy,4-methoxyphenyl) resulted in the introduction of an N-acylhydrazone (NAH) subunit to replace the ethylene bridge (CH = CH). As expected, this type of modification caused significant conformational changes and altered the spatial arrangement of rings A and B during molecular recognition by b-tubulin ( Figure 2). These findings are supported by data from the literature describing the anti-tubulin activity of E-chalcones (e.g., 6, Figure 1) [14][15][16]. Additionally, the introduction of halogens substituents at position 4 of the B ring took advantage of the metabolic protection that might be exerted by these substituents, preventing aromatic hydroxylation at C4 catalyzed by the CYP450 enzymatic complex [17].
To identify the most energetically favorable pose (i.e., pose prediction), each pose of the N-acylhydrazone derivatives 5a-r within the colchicine binding site of b-tubulin was evaluated (i.e., scored) based on their complementarity to the target with respect   to their shape and properties, such as electrostatics. It is noteworthy that score is the most adequate way of selecting the best pose, since the scores are assigned according to the interaction mode of a ligand with the binding site, as measured by fitness function. The fitness function was selected after redocking experiments with colchicine in the binding site of b-tubulin (PDB code: 1sa0). The RMSD between the experimental structure and the top scored pose, determined after redocking experiments with the four fitness functions available in GOLD 5.0.1 program (i.e. Chemscore, Goldscore, ASP and ChemPLP), revealed that Chemscore was the fitness function with the best performance in this study (RMSD = 1.0606).
Giving a good score to a compound indicates that it exhibited good binding with the protein, and the results were compared to the data obtained with CA4 ( Table 1). As depicted in Table 1, five compounds (5d, 5k, 5l, 5m and 5n) were predicted to display better binding than CA4. In this group, the most favorable complementary interaction was observed with compound 5d (LASSBio-1588), which forms a hydrogen bond between the hydroxyl group on ring B (i.e., 4-hydroxyphenyl) with Val 662. However, compounds 5k, 5l, 5m and 5n display complementary interactions using the lipophilic nature of ring B to exploit the hydrophobic pocket composed by residues Leu242, Val238, and Leu255 at the colchicine site of b-tubulin (data not shown). These data agree with the work of Dorléans and coworkers: ligands of the colchicine binding site establish few polar interactions within the protein-ligand complex, and van der Waals interactions are more relevant during molecular recognition [18]. The worst scores were observed with compounds 5g, 5h, 5i and 5p, which possessed polar groups on ring B that could not act as hydrogen bond donors. The score values determined during the docking studies and some physicochemical properties (cLogP, cLogD, MR and the aqueous solubility) for compounds 5a-r are summarized in Table 1 (see also Figure S3 in supporting material).
The N-acylhydrazones (5a-r) were obtained at a two-step linear route ( Figure 4) [19], using methyl 3,4,5-trimethoxybenzoate ester (7) as the starting material. While exploring a hydrazinolysis reaction, ester 7 was refluxed with hydrazine hydrate 80% in ethanol, providing the 3,4,5-trimetoxybenzohydrazide (8) in 93% yield. The hydrazide (8) was condensed with the appropriate aldehydes, which were selected in accordance with the molecular design depicted in Figure 1, in the presence of ethanol and catalytic hydrochloric acid to furnish the CA-4 analogues 5a-r in high yields.
Compounds 5a-r were characterized by 1 H NMR, 13 C NMR and IR spectroscopy and their purity was determined by HPLC, with a reverse-phase column at different systems of mobile phase. All N-acylhydrazone derivatives (5a-r) were obtained as a single diastereoisomer (Z or E), as indicated by the analysis of the 1 H and 13 C NMR spectra; no duplicate signals attributed to the hydrogen or carbon atom of the imine (N = CH) were observed. The stereochemistry of the imine double bond was subsequently assigned based on our previous results [23] and the X-ray crystallographic studies performed with 5b (LASSBio-1586).
A single crystal of compound 5b (LASSBio-1586) was obtained and subjected to X-ray diffraction; the ORTEP [20,37] view is shown in Figure 5. Crystallographic analysis confirmed that the configuration about the C2 = N2 double bond [distance 1.273(3) Å ] was E and revealed a nearly flat conformation of the benzoylhydrazide moiety, which was described by the least Table 1. Scores estimated by molecular docking (ChemScore fitness function) for colchicine binding site of b-tubulin, cLogP, cLogD 7.0 , molar refractivity and the aqueous solubility of CA-4 and its N-acylhydrazone analogues 5a-r.  The parallel arrays are bound by weak van der Waals interactions between methyl group C15 and the O2 oxygen atom from a neighboring molecule, demonstrating the availability of this group for intermolecular interactions once the methoxy group in the para position is rotated to the opposite side. Crystallographic data of compound 5b (excluding structure factors) can be seen in supporting information. Crystallographic data of compound 5b (excluding structure factors) can be seen in supporting information.
The antiproliferative activity of compounds 5a-r was determined based on an MTT assay [21] and using CA-4 as standard against the tumor cell lines: HL-60 (human leukemia), SF-295 (human glioblastoma), MDA-MB435 (melanoma), PC3M (prostate cancer), OVCAR-8 (ovaries adenocarcinoma), NCI-H258M (pulmonary bronchio-alveolar carcinoma) and HCT-8 (adenocarcinoma ileocecal) ( Table 2). To determine the selectivity index of compounds 5a-r, their antiproliferative profile was also evaluated toward human lymphocytes ( Table 2). Table 2, all compounds except for derivatives 5i, 5j, 5k and 5n exhibited moderate to high antiproliferative potency with IC 50 values #18 mM and $4 nM. These results are in agreement with Jin and co-workers [22], who described the antiproliferative activity of some NAH containing the trimethoxyphenyl subunit against PC3, A431 and BGC823 tumor cells for the first time. The N-acylhydrazones with hydrophobic substituents on ring B (i.e., 5l, 5m, 5o, 5p, 5q and 5r) were more potent, which was predicted by the score values obtained from the docking studies. The in silico study failed to predict the cytotoxic activity of compound 5n and 5d, which scored as better binders than CA-4. The inactivity of compound 5n (IC 50 .25 mM) suggested that there were steric constraints in the recognition between the ligand    and the active site of b-tubulin because compounds with bulkier groups (for MR values see Table 1) attached to the imine (i.e.5i, 5j, 5k and 5n) displayed the worst activities. Moreover, compounds 5i, 5k and 5n bind differently from CA4 and 5b at the colchicine binding site, with no interaction with residues Ser178 and/or Thr 179 (see Figure S4 in supporting material). The addition of a 2chromone subunit caused the loss of antiproliferative potency, while the inclusion of oxygenated substituents at the phenyl ring (ring B; 5a, 5c, 5d, 5e, 5f, 5g, 5h) did not significantly interfere with the cytotoxic potency compared to compound 5b; however, these compounds were still significantly less active than CA-4 ( Table 2).

As shown in
To investigate the selective cytotoxic activity of the Nacylhydrazones derivatives (5a-r), their antiproliferative potency was also assessed toward human lymphocytes and the results were compared to the data from CA-4 ( Table 3). The selectivity index (SI), which was the IC 50 for human lymphocytes/IC 50 for cancer cell lines after treatment with CA-4 and N-acylhydrazones (5a-r), was calculated, as depicted in Table 3. Excluding the compounds that were inactive or slightly cytotoxic (5i, 5j, 5k and 5n), the more lipophilic (cLogP $3.15 #4.37, Table 1) compounds (5b, 5l, 5m, 5o, 5p, 5q, 5r) exhibited cytotoxic potency against human lymphocytes similar to the lead compound, CA-4. Notably, CA-4 was proven to be a non-selective cytotoxic agent, with higher antiproliferative potency against human lymphocytes versus tumor cell lines, except for HL-60 and OVCAR-8 (Table 3). In contrast, LASSBio-1586 (5b) exhibited a cytotoxic selectivity index from 2.4 to 42 times greater than CA-4 (Table 3; SI values for 5e versus SI values for CA-4). The best comparative selectivity indices (CA-4 vs 5b) were obtained from the SF-295 (SI = 13), MDA-MB435 (SI = 42) and NCI-H258M (SI = 9.5) tumor cell lines, and the worst results were found for OVCAR-8 (SI = 0.5).
Considering the IC 50 (#0.8 mM and $0.064 mM, Table 2) and the SI values (Table 3), LASSBio-1586 (5b) was selected as the most promising compound, and its ability to inhibit tubulin polymerization was investigated. The tubulin polymerization assay was performed by CEREPH employing a single concentration of 5b (C = 30 mM), using vinblastine as positive control. In this assay, LASSBio-1586 (5b) inhibited 91% of the tubulin polymerization, validating the rational design employed in the molecular design of the derivatives 5a-r (data not shown; available in the supplementary information, Figure S1).
To establish the minimum structural requirements essential for the anti-tubulin activity of LASSBio-1586 (5b), some molecular modifications were introduced to its structure, leading to the design of compounds 9-12 ( Figure 6). The N-acylhydrazone derivatives 9 and 10 were synthesized using the same methodology employed to obtain compounds 5a-r [19]. The homologous compound 11 was prepared in good yield via chemoselective alkylation of the sp 3 nitrogen in the N-acylhydrazone functionality using methyl iodide and potassium carbonate in acetone [23]. Semicarbazone 12 was synthesized in three linear steps in 25% overall yield, as illustrated in Figure 7 [24].
The in vitro antiproliferative activity of compounds 9-12 was assessed against HL-60, SF296, HCT-8 and MDA-MB435 tumor cells and compared with the data from LASSBio-1586 (5b) and CA-4 ( Table 4). As displayed in Table 4, the elimination of the methoxy groups from the trimethoxyphenyl subunit (ring A) present in LASSBio-1586 (5b) caused the loss of cytotoxic activity, as depicted by compound 9, suggesting that this subunit was a pharmacophore. Similarly, retroisostere 10 was inactive, validating the role of the trimethoxyphenyl moiety as a pharmacophore when linked to the carbonyl group of the NAH functionality. The homologous compound 11 was well tolerated, exhibiting a slight increase in cytotoxic potency against HL-60 and HCT-8 tumor cell lines relative to compound 5b. However, the aza-homologous 12 was inactive, suggesting that the semicarbazone unit was not suitable to replace the ethylene linker in CA4 or the NAH in 5b. The greater conformational freedom introduced by the NH group may have compromised the bioactive conformation, altering the optimal spatial positioning between the aromatic rings necessary for molecular recognition with the b-tubulin binding site. To support these hypotheses, compounds 9-12 were subjected to docking studies and the best poses with the colchicine binding site of b-tubulin were analyzed ( Figure 8). As presented in Figure 8, compounds 9 and 10 lost the hydrogen bond with Ser178 observed during the molecular interaction between compound 5b and the colchicine binding pocket of b-tubulin protein. Similarly, semicarbazone 12 does not interact electrostatically with Ser178 and adopts a specific and unfavorable orientation within the active site of b-tubulin.
Considering the overall cytotoxic profile of LASSBio-1586 (5b) and its confirmed ability to inhibit microtubule polymerization, the antitumor activity was evaluated. The Hollow Fiber Assay (HFA) was selected because it is employed by the National Cancer Institute (NCI) as the standard model for the evaluation of new antiproliferative drugs before assessment via the in vivo-grown human tumor xenograft screen [25][26][27][28].
The hollow fibers were well tolerated by the animals, and no signs of rejection were detected. The treatments with LASSBio-1586 (5b) and 5-FU did not affect the health of the mice beyond acceptable limits and no deaths occurred.

Conclusions
Based on the results of docking studies, a series of Nacylhydrazone derivatives were used as structural analogues of CA-4. These studies identified the major structural requirements essential for molecular recognition by the colchicine binding site of b-tubulin. Of the active compounds, LASSBio-1586 (5b) emerged as a simple antitumor drug candidate and was capable of inhibiting microtubule polymerization; this compound also possessed broad in vitro and in vivo antiproliferative profile and a better selectivity index than the lead compound, CA-4, which indicated that 5b displayed improved selective cytotoxicity toward cancer cells.

Ethics Statement
Procedures are in accordance with guidelines for the welfare of animals in experimental neoplasia [40] and with national and international standard on the care and use of experimental laboratory animals [41] and were approved by the local Ethical      were determined by HPLC (.95%) using the Shimadzu -LC20AD apparatus, a Kromasil 100-5C18 (4,6 mm6250 mm) column and the SPD-M20A detector (Diode Array) at 254 nm for quantification of analyte in a 1 mL/min constant flux. The injector was programmed to inject a volume of 20 -mL. The mobile phases used were: CH 3 CN:H 2 O 1:1; 6:4 and 7:3.
Ultraviolet spectroscopy was performed using Femto spectrophotometer. The wavelength used in solubility assay was determined by the -l max characteristic of each compound. Spectra were analyzed in Femtoscan software. Mass spectrometry was obtained by positive ionization at Bruker AmaZon SL and data analyzed in Compass 1.3.SR2 software.
General Procedure for the preparation of 3,4,5trimethoxybenzohydrazide (8) To a solution of methyl 3,4,5-trimethoxybenzoate (7)    General Procedure for the preparation of 3,4,5trimethoxybenzoyl-arylhydrazones (5a-r) To a solution of 8 (0.2 g, 0.884 mmol) in absolute ethanol (7 mL) containing one drop of 37% hydrochloric acid, was added 0.884 mmol of corresponding aldehyde derivative. The mixture was stirred at room temperature until TLC indicated the end of reaction (0.5-4 h). Then the mixture was poured into ice and the precipitate was filtered out and dried. Yields and characterization pattern are described below:   The data are reported as the mean 6 S.E.M., n = 6-7 animals/group, which were treated for 4 days intraperitoneally. 2 The negative control group received 5% DMSO.   Benzohydrazide A solution of benzoic acid (2.0 g, 16.4 mmol) in 50 mL of methanol, containing 5 drops of sulfuric acid, were refluxed with a Dean-Stark apparatus for 5 hours to obtain the corresponding methyl ester. Then, 328 mmol of hydrazine hydrate 80% were added to the reaction mixture and kept under reflux for 2 hours to obtain benzohydrazide in a one-pot methodology in 94% yield as a white solid, m.p. 112-114uC. The melting point data is in agreement with previous reports [33]. Synthesis of N-(3,4,5-trimetoxyphenyl) hydrazine carboxamide (15) To a suspension of 14 (0.6 g, 1.98 mmol) in 15 mL of dry toluene, was added 29.7 mmol of hydrazide hydrate 64% and the mixture was stirred at room temperature until conversion of carbamate to correspondent semicarbazide. The product was filtered under vacuum, washed with n-hexane and obtained as a brown solid in 90% yield, m.p. .250uC; I.R.(KBr) cm 21

X-ray Crystallography
A colorless prismatic single crystal of the compound LASSBio-1586, suitable for x-ray study, was obtained by slow evaporation of a solution of methanol-dimethylformamide (2:1) at room temperature 295(2) K. Data collection was performed using the Kappa Apex II Duo diffractometer operating with Cu-Ka radiation at 100 K. 8336 data points were collected of what 2687 are symmetry independent (R int = 0.044). The molecule crystallizes in the Pca2 1 space group, having Z = 4. Structure solution was obtained using Direct Methods implemented in SHELXS [36] and the model refinement was performed with full matrix least squares on F 2 using SHELXL [36], with final residuals R1 = 0.037, wR2 = 0.105 for 2461 observed data with I.2s(I), and R1 = 0.046, wR2 = 0.111 for all data. The data completeness allowed for a qualitative decision of the chirality, however because only low weight atoms are present, the Flack parameter has a relatively large standard deviation, being 0.04 (19). The crystal packing is stabilized by an intermolecular hydrogen bond of type N1-H1…O1 i , building a linear chain though (100). Hydrogen bond geometry is given in Table 6. The programs ORTEP-3 [20], SHELXS/SHELXL [36] were used within WinGX 37 software package.

Solubility Assay
The solubility assay was performed considering the absorptivity of compounds in ultraviolet spectroscopy as described by Schneider and coworkers [38]. The assay wavelength was determined by the l max characteristic of each compound. Saturated aqueous solutions were prepared (0.01 mg/mL) and were kept under stirring for 2 hours at 37uC. The supernatant was filtered in 0,45 mm filters and transferred to a quartz cuvette (10 mm) to spectra acquisition.
Solubility was determined by linear regression using as graph plots, solutions prepared by dilutions of the original solution in methanol. The data were obtained in triplicates and the mean values were used to the graph plots. The correlation coefficient (R 2 ) values were between 0.9972 and 0.9999.

Tubulin Polymerization Assay
The tubulin polymerization assay was performed by CEREPH, as described by Bonne and co-workers [39].

Hollow Fiber Assay
A total of 26 female BALB/c nude (nu/nu) mice aging 6-8 weeks were obtained from the animal facilities of State University of São Paulo (USP), Faculty of Medicine, São Paulo (SP), Brazil. They were kept in well-ventilated and sterile cages (Alesco, São Paulo) under standard conditions of light (12 h with alternative day and night cycles) and temperature (2261uC) and were housed with access to commercial sterile rodent stock diet (Nutrilabor, São Paulo, Brazil) and water ad libitum. As previously mentioned procedures are in accordance with guidelines for the welfare of animals in experimental neoplasia [40] and with national and international standard on the care and use of experimental laboratory animals [41] and were approved by the local Ethical Committee on Animal Research (Process No. 102/2007).

HF Preparation, Surgery Deployment and Determination of the Antiproliferative Capacity
Polyvinylidene fluoride (PVDF) HFs with a 1-mm internal diameter and a molecular weight cutoff point of 500 kDa were used (Spectrum Laboratories, Houston, TX). The fibers were cut Table 6. Intermolecular hydrogen bond geometry. into pieces 12-15 cm long, washed 26 with sterile distilled water and kept in sterile conditions. Before use, under sterile conditions, the fibers were incubated in complete RPMI with 20% fetal bovine serum (FBS) overnight (packaging time). Cell viability was assessed by trypan blue exclusion assay. Then, a cell suspension of 7.0610 6 cell/mL at 4uC was injected into the fiber, with the ends thereof immediately heat-sealed. The fibers were cut into 2 cm each, transferred to petri plates and incubated in complete RPMI medium for 24 h prior to implantation in mice. Each cell was injected into one fiber of a different color (HCT-116, yellow fibers; SF-295, blue fibers).

V Antiproliferative Assay
Tumor cell proliferation was quantified through the ability of living cells to reduce the yellow dye 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT, Sigma Aldrich) to a purple formazan product [21]. For this purpose, the fibers removed from animals were incubated with MTT 1 mg/mL in 6-well plates during 4 h at 37uC, 5% CO 2 and 95% humidity. The MTT solution was aspirated; fibers were washed with saline solution containing protamine sulphate 2.5% and incubated in protamine solution overnight at 4uC. Fibers were transferred to 24 well plates, cut into 2 or 3 pieces and put to dry. The formazan was dissolved in 500 mL of DMSO, aliquots (150 uL) were transferred to 96 well plates and absorbance was measured at 595 nm (DTX-880, Beckman Coulter).

Statistical Analysis
In order to determine differences between groups, data (mean 6 S.E.M) were compared by one-way analysis of variance (ANOVA) followed by Student Newman-Keuls test (P,0.05).

Molecular Modeling
Compounds were constructed and submitted to a conformational analysis by molecular mechanics (MMFF method) with the Spartan 8.0 software (Wavefunction Inc.; Licence number: DQAIR/HASPUSB). The most stable conformer of each structure was reoptimized with the AM1 semiempirical molecular orbital method [42] and saved as mol2 files for docking studies into the colchicine binding site of the b-tubulin crystallographic structure available in the Protein Data Bank with code 1sa0. This structure was chosen because of the b-tubulin conformation induced by the co-crystallized colchicine, which prevents curved btubulin from adopting a straight structure, inhibiting assembly. Docking studies were implemented with the GOLD 5.0.1 program (CCDC), which employs a genetic algorithm for docking flexible ligands into protein binding sites and ranks the resulting poses according to their scores determined by available scoring functions. Hydrogen atoms were added to the protein structure according to the tautomeric and ionized states inferred by the program. The colchicine structure was removed for the docking studies, which were performed with the ChemScore scoring function, which contains specific energy terms for hydrogen bonding and lipophilic interactions [43,44]. The data and poses were analyzed on Pymol program. Licences numbers: Pymol (8588); Gold (G/414/2006).