Pyrrolo[2′,3′:3,4]cyclohepta[1,2-d][1,2]oxazoles, a New Class of Antimitotic Agents Active against Multiple Malignant Cell Types

A new class of pyrrolo[2′,3′:3,4]cyclohepta[1,2-d][1,2]oxazoles was synthesized for the treatment of hyperproliferative pathologies, including neoplasms. The new compounds were screened in the 60 human cancer cell lines of the NCI drug screen and showed potent activity with GI50 values reaching the nanomolar level, with mean graph midpoints of 0.08–0.41 μM. All compounds were further tested on six lymphoma cell lines, and eight showed potent growth inhibitory effects with IC50 values lower than 500 nM. Mechanism of action studies showed the ability of the new [1,2]oxazoles to arrest cells in the G2/M phase in a concentration dependent manner and to induce apoptosis through the mitochondrial pathway. The most active compounds inhibited tubulin polymerization, with IC50 values of 1.9–8.2 μM, and appeared to bind to the colchicine site. The G2/M arrest was accompanied by apoptosis, mitochondrial depolarization, generation of reactive oxygen species, and PARP cleavage.


■ INTRODUCTION
Microtubules are intracellular polymers involved in the regulation of a large number of cellular processes, including proliferation, division, determination and maintenance of cellular shape, motility, and intracellular transport. 1 They are highly dynamic structures composed of multiple heterodimers of αand β-tubulin, and they undergo alternating polymerization and depolymerization phases. 2 Disrupting this dynamic equilibrium interferes with cell division and leads to cell death. Tubulin and its associated structures represent an attractive target in the treatment of cancer. 3 Over the past 40 years, a large number of natural and synthetic compounds interfering with microtubule dynamics through interactions with multiple binding sites on tubulin have been described. 4 On the basis of their effects on microtubule dynamics, they can be classified as either microtubule-stabilizing agents or microtubule-destabilizing agents. 5 Among natural derivatives, taxanes, e.g., paclitaxel and docetaxel, belong to the first group of compounds, while vinca alkaloids, e.g., vinflunine, vinorelbine, vincristine, and colchicine, belong to the second group. 6 Despite the large number of new promising drug candidates, no molecules binding at the colchicine site have been approved thus far for the treatment of cancer, leaving the drug discovery process still open. 7,8 Since its discovery, combretastatin A-4 (CA-4) is still considered a promising lead compound binding at the colchicine site (Chart 1). 9 It inhibits tubulin polymerization with low IC 50 values, 10 and it presents a potent activity against multiple cancer cell lines, including cells bearing a multidrug resistance (MDR) phenotype. 11 In addition to its antimitotic activity, CA-4 can interfere with tumor vasculature, essential for solid tumor survival, leading to necrosis of tumor tissues. 12 Nevertheless, due to its poor water solubility, low bioavailability, and rapid clearance, CA-4 exhibits poor activity in vivo, thus leading to the synthesis of different water-soluble prodrugs including CA-4 phosphate disodium (CA-4P) (Chart 1). In different preclinical models, CA-4P reduces blood flow and causes tumor cell death due to changes in the morphology of immature endothelial cells resulting from interference with tubulin polymerization. 13 As the cisconfiguration of the olefinic double bond is essential for the antiproliferative activity of CA-4, this bond has been fixed through its incorporation into fiveor six-membered heterocycle rings. 12,14,15 The 4,5-diarylisoxazoles showed potent antitumor activity in inducing cell cycle arrest at the G2/M phase of the cell cycle and potent antitubulin activity (Chart 1). 16,17 KRIBB3 (Chart 1), belonging to the same class, displayed antiproliferative activity through inhibition of microtubule polymerization and spindle assembly checkpoint activation. In in vivo models, KRIBB3 caused a 50−70% reduction of tumor growth at a dose of 50−100 mg/kg. 18,19 Isoxazoles or [1,2]oxazoles represent the core structure of many drug candidates. Due to its ability to form multiple noncovalent interactions with a wide number of proteins, this moiety confers different biological activities, such as antitumor, antinflammatory, antidepressant, antiviral, antibacterial, and antitubercolosis activities. 20−27 A series of 5-(1H-indol-5-yl)-3phenylisoxazoles have anticancer activity. 28 Several small molecules containing the indole moiety have also been described as potent tubulin polymerization inhibitors. 29−32 Our research group has devoted much effort to the synthesis and evaluation of the biological properties of fused tricyclic systems incorporating the pyrrole ring. 33−41 Since the [1,2]oxazole system is found as a pharmacophore moiety of several compounds with promising antitumor properties, we started a program investigating different classes of pyrazoleand pyrrole-fused systems of types 1, 2, and 4, incorporating the [1,2]oxazole unit ( Figure 1). 42 −45 In particular, from the class of [1,2]oxazole of type 2, ethyl 8-(3,5-dimethoxybenzyl)-5,8-dihydro-4H- [1,2]oxazolo [4,5-g]indole-7-carboxylate 3 emerged for its in vitro nanomolar growth inhibitory effects across the National Cancer Institute (NCI) cancer cell line panel, with mean graph midpoints (MG_MIDs) of 0.25 μM on the full panel and a GI 50 range of 0.03−31.1 μM.
The potent antitumor activity made the class of compounds worth further evaluation, encouraging the synthesis of new [1,2]oxazolo derivatives with the aim of obtaining more potent antiproliferative agents. For a better insight into the structure− activity relationship (SAR) of the tricyclic scaffold, containing also the pyrrole moiety, we started a drug discovery program aimed at understanding the optimal structural requirements of this class of small molecules. Thus, we first identified [1,2]oxazolo [5,4-e]isoindole system 4, which highlighted the potential of this group of compounds as tubulin polymerization inhibitors. 44 This class of compounds also displayed potent growth inhibitory activity on the NCI panel (GI 50 = 0.01− 27.00 μM). 44 Moreover, some derivatives significantly impaired the growth of human cancer cell lines of different histological origin, including experimental models of diffuse malignant peritoneal mesothelioma (DMPM), without interfering with normal cell proliferation. Their antiproliferative activity was found to derive from their ability to impair microtubule assembly during mitosis, with a consequent cell cycle arrest at the G2/M phase and induction of caspasedependent apoptosis. In addition, selected derivatives, at welltolerated doses, significantly reduced tumor volume in a DMPM xenograft model. 44,45 Bearing in mind the polycyclic structure of colchicine, which includes two cyclohepta rings, and that this structural feature is recurrent in other examples reported in the literature 46,47 as potent inhibitors of tubulin assembly, we planned the expansion of the cyclohexyl central ring by one member while maintaining the [1,2]oxazole and pyrrole moieties. Thus, the new tricyclic derivatives pyrrolo[2′,3′:3,4]cyclohepta [1,2d] [1,2]oxazoles 5 ( Figure 1) were synthesized in order to investigate the effects of this structural modification on the biological properties of these compounds. This ring system was unexplored so far as a chemical entity, and because of the close correlation with the parent structure of type 2, in this set of derivatives we decided to retain some structural features, specifically the carboxyester and methoxy-substituted benzyl groups, that had emerged as crucial for biological activity.

■ CHEMISTRY
The synthetic strategy optimized by us to obtain the title ring system is outlined in Scheme 1. We started from cyclohepta-[b]pyrrol-8-one ketones of type 6−8, 48,49 as the α position to the carbonyl is appropriate for the introduction of the second electrophilic site, essential for the subsequent cyclization with dinucleophiles.
Reaction of intermediates 25−42 and 43−50 with hydroxylamine hydrochloride, as a 1,3-dinucleophile, and a stoichiometric amount of acetic acid in refluxing ethanol furnished [1,2]oxazole derivatives 51−68 in 60−90% yields (Table 1). Selected [1,2]oxazoles were then subjected to smooth chlorination with N-chlorosuccinimide to afford the corresponding chloro [1,2]oxazoles. In particular, derivatives 69−75, belonging to the ethoxycarbonyl series, were obtained in good yield (60−75%). For [1,2]oxazoles 51−59, which bear two pyrrole positions for potential chlorination, a mixture of the 7and 8-halo substituted derivatives was detected by NMR analysis, and it was not possible to isolate either component as a pure compound. Only in the case of the N-methyl derivative 52 was the 8-chloro substituted derivative 76 (50%) recovered as a pure compound from the reaction mixture.

■ RESULTS AND DISCUSSION
Antiproliferative Activity in the NCI Panel. All the synthesized compounds 51−76 were tested at a 10 −5 M concentration for their antitumor activity on the full NCI-60 panel comprising cancer cell lines derived from nine human cancer cell types (leukemia, non-small-cell lung, colon, central nervous system, melanoma, ovarian, renal, prostate, and breast). 50 On the basis of these results, six compounds (62,63,66,67,70,75) were selected for further screening on the same panel at five concentrations at 10-fold dilutions (10 −4 − 10 −8 M). Almost all compounds showed antiproliferative activity against all tested human tumor cell lines, with nM to μM GI 50 values (Table 2).
From a SAR point of view, the presence of an ethoxycarbonyl group at position 8 was crucial for activity. The most potent compound was 66, which has a 3,5dimethoxybenzyl substitutent at the pyrrole nitrogen and a mean graph midpoint (MG_MID) of 0.08 μM on the full NCI panel. From analysis of the GI 50 values listed in Table 3, 66 was particularly effective against the melanoma (GI 50 = 0.09− 0.01 μM), prostate (GI 50 = 0.04 μM), and renal (GI 50 = 0.07− 0.02 μM) cancer subpanels (Figures S1 and S2), maintaining nanomolar activity against all the tested cell lines. The calculated MG_MID value for each subpanel was 0.04 μM, much lower than the overall cell line MG_MID value. Notably, the best activity was observed for the MDA-MB-435 cell line of the melanoma subpanel, with a GI 50 of 10 nM. Moreover, the colon and CNS cancers had mean values of 0.06 μM, again lower than the average mean value.
Compound 67, a 3,4,5-trimethoxybenzyl substituted derivative, was the second best in potency and demonstrated high selectivity against the leukemia (GI 50 Figures S3 and S4). Isoxazole 67, even if it was 1 order of magnitude less potent than the dimethoxy substituted analogue 66, reached nanomolar GI 50 values in each subpanel, and it also had a 10 nM GI 50 against the NCI-H522 non-small-cell lung cancer cells.
Screening Results in Lymphoma Models. All compounds were further tested at the concentration of 1 μM on four cell lines derived from distinct lymphoma histotypes, plus two with secondary resistance to the PI3Kδ inhibitor idelalisib 51 or to the BTK inhibitor ibrutinib. 52 After a 72 h incubation, compounds 57, 66, 67, 71, 74, and 75 showed potent growth inhibitory effects against all tested cell lines, with the percentage of proliferating cells reduced to 9−60% of the untreated cells (Table 4). For comparison, the same experiments were conducted using compound 3, but the response to it was minimal, and thus it was not considered further.
Compounds showing some activity, plus compound 70 based on the NCI panel data, were tested with a wider range of concentrations. Some presented potent growth growth inhibitory effects on some or all of the lymphoma cell lines, with IC 50 values lower than 500 nM ( Table 5).
Effects of Test Compounds in Human Peripheral Blood Lymphocytes (PBLs). To obtain an initial idea of whether the compounds described here had activity against normal cells, 63, 66, and 75 were examined for cytotoxicty against PBLs from healthy donors. As shown in Table 6, these three compounds were practically devoid of any activity both in quiescent and in lymphocytes induced to proliferate by the mitogenic stimulus phytohematoaglutinin (PHA). In all cases, we obtained a GI 50 > 100 μM, demonstrating low toxicity for these healthy human cells.
In this context, we point out that in other studies other molecules that bind in the colchicine site were shown to have low toxicity toward lymphocytes from healthy subjects. 53−56 Although at present the reason for this low toxicity is unclear, it is nevertheless interesting that even healthy lymphocytes induced to actively replicate with a mitogenic stimulus respond in the same way as quiescent lymphocytes.
Tubulin Assays. To assess if pyrrolocyclohepta[1,2]oxazoles were able to bind to tubulin, seven compounds (Table 7) were tested for their antitubulin activity in comparison with reference compound CA-4, which potently inhibits both tubulin assembly and colchicine binding to tubulin. 57 Moreover, compound 3 was also evaluated as a comparison between the two scaffolds.
The colchicine assay was performed on compounds that yielded IC 50 values of <6 μM in the assembly assay. Reaction mixtures in the assembly assay contained 9 μM (0.9 mg/mL)  tubulin in the assembly assay, and in the colchicine assay they contained 0.5 μM tubulin, 5.0 μM [ 3 H]colchicine, and 5.0 μM inhibitor.
In the assembly assay, three compounds had IC 50 values of >6 μM. These were compounds 3, 56, and 63. In addition, we examined several other compounds shown in Table 1, and they were uniformly minimally active in the tubulin assembly assay. The five other compounds (57,58,66,67, and 75) were more active in the assembly assay, with 66 and 75 the most active, with IC 50 values of 2.6 and 1.9 μM, respectively. A value of 1.2 μM was obtained for CA-4. The five pyrrolocyclohepta[1,2]oxazoles most active as assembly inhibitors inhibited colchicine binding by 25−62% versus 97% for CA-4. Overall, the most powerful compound was 66, which inhibited tubulin polymerization with an IC 50 of 2.6 μM and displayed 62% inhibition of colchicine binding. No compound was as active as CA-4 in any assay.
The effects of compounds 66 and 75 on tubulin assembly are shown in Figure 2, panels A and B, respectively. These data were obtained in computer-driven recording spectrophotometers equipped with electronic temperature controllers that rapidly change the temperature in the reaction mixtures in the cuvettes. The assembly reaction was measured by following turbidity development at 350 nM. After a minute's equilibration at 0°C, the temperature was jumped to 30°C and assembly was followed for 20 min. At 21 min, the temperature was jumped backward to 0°C, and the reaction mixtures were followed for another 8 min. Several compound concentrations were evaluated in each experimental sequence, and the IC 50 for inhibition of turbidity development was defined as the compound concentration, obtained by interpolation, that inhibited the extent of turbidity development by 50% after 20 min at 30°C. The 30−0°C transition was included to distinguish inhibition of microtubule assembly from aberrant assembly reactions induced by numerous compounds. Typically, the aberrant assembly reaction products either are cold stable or have different temperature stability properties as compared to microtubules. Molecular Modeling. Compound 3 and all the compounds belonging to the new class of pyrrolo[2′,3′:3,4]cyclohepta [1,2-d] [1,2]oxazoles 5 were docked into the colchicine and vinblastine binding sites, by selecting for each of them the pose with the best G-Score (kcal/mol). A better affinity for the colchicine site (Table S2) was observed for all ligands, further confirming their specificity for this binding pocket. Moreover, most of the newly synthesized compounds had a better affinity than the parent compound 3. To further investigate the binding mode of the best active compounds (57, 58, 63, 66, 67, and 75) in the biological assays with respect to 3, molecular modeling studies were performed on the 3N2G model, which displays two additional neighboring pockets (zones 2 and 3) in the tubulin colchicine domain. As was the case with their parent compound 3, compounds 57, 58, 63, 66, 67, and 75 had a better G-Score toward the main site (zone 1) of the colchicine domain (Table  S3). As shown in Figure S5, our compounds had unfavorable steric contacts with an additional hydrophobic pocket of the β subunit, formed by residues E200, L255, A316, A317, A354, C241, and T179.
On the other hand, the best docking poses of active compounds with tubulin structure 4O2B, containing zones 1 and 2 of the colchicine site, showed strong hydrophobic    interactions with β-tubulin residues L248, A250, A354, I318, A316, and L255 ( Figure 3). In particular, 66 and 75, which are the compounds with the best biological activity, displayed a binding geometry similar to that of colchicine in zones 1 and 2 of the pocket, by directing their methoxybenzyl groups toward the C241 residue ( Figure 3D and Figure 3F). Moreover, 75 also established a halogen bond between its chlorine and the backbone of V181 and an H-bond between the oxazole moiety and the backbone of N249. Conversely, compound 3 and the less active compounds (57, 58, 63, and 67) showed a different binding orientation, with the tricyclic portion steered toward residue C241 ( Figure 3A,B,C,E,G). In particular, for compound 3 we observed a π−cation between its pyrrole and β-tubulin K352, while hydrophobic interactions were much weaker than with the other derivatives.
The best docking poses of 57, 58, 63, 66, 67, 75, and 3 against the 4O2B model were submitted to explicit water solvent molecular dynamics (MD) simulations, with the aims to add depth to our analysis and to investigate the possibility of induced-fit phenomena in the tubulin recognition process of our ligands. As a reference, the X-ray model of 4O2B, containing colchicine in its binding pocket, was included in similar calculations. In the Supporting Information we reported the geometric behavior of all MD simulations and the analysis of their most representative structures, by computing the related binding free energy and the global number of contacts (Table S4).
With the respect to its docking pose ( Figure 3G), the most representative MD structure of 3 showed the establishment of three H-bonds ( Figure 4G). In particular, the two methoxy groups interact with β-tubulin N101 and N249, while the carbonyl group establishes an H-bond with S318. However, the formation of these hydrogen bonds does not allow stability of the bonding mode and does not bring about an energy gain, due to the lower ability to establish hydrophobic interactions.
Regarding 66, the most representative MD structure showed that the tubulin molecule adjusted its residues to allow establishment of an H-bond between its methoxy group and the side chain of C241 and to permit a π−cation between its oxazole ring and K254 ( Figure 3D). Further hydrophobic interactions with β-tubulin K254, A250, L255, A316, and A354 stabilized the complex. Likewise, 75 showed a binding mode similar to that of 66, by engaging an H-bond between its 4- Reaction mixtures (0.25 mL, final volume) contained 0.8 M monosodium glutamate (adjusted to pH 6.6 in a 2 M stock solution), 0.9 μM (0.9 mg/mL) tubulin, 4% dimethyl sulfoxide, compounds at the indicated concentrations, and following a 15 min preincubation in 0.24 mL, 0.2 mM GTP (added in a 10 μL volume). The reaction mixtures, following the preincubation, were kept on ice and transferred to cuvettes held at 0°C in a recording spectrophotometer. After baselines were established, the reactions were initiated. At 1 min, the electronic temperature controller automatically increased the temperatures in the cuvettes to 30°C, and at 21 min, the temperatures in the cuvettes were returned to 0°C (the temperature transitions take about 30 and 60 s, respectively).  Compound 66 Induced Alteration of Cell Cycle Checkpoint Proteins. We studied the effects of 66 on the expression of various checkpoint proteins that play roles in cell cycle regulation. Cells that enter mitosis do so through the involvement of cyclin B1 complexed to cdc2. This complex is activated through the dephosphorylation of phospho-cdc2, which is a cdc25c-dependent process that ultimately leads to the phosphorylation of cyclin B1. This phosphorylated enzyme triggers cells to enter mitosis. 58,59 Figure 6 demonstrates a substantial increase of cyclin B1 expression after a 24 h treatment with 0.5 μM 66. In contrast, total cdc25c expression was strongly reduced, and in good agreement, the expression of phosphorylated cdc2 was strongly decreased after both 24 and 48 h. Dephosphorylation of this protein is needed to activate the cdc2/cyclin B complex, and this effect is stimulated by cdc25c. 58,59 These data demonstrate that cdc2/cyclin B1 complexes were not activated, thus blocking cells from exiting mitosis and leading to apoptotic cell death.  Compound 66 Induced Apoptosis through the Mitochondrial Pathway. In the initial stages of induction of apoptosis, the mitochondrial transmembrane potential (Δψ mt ) is altered and leads to to a reduction of Δψ mt and release of cytochrome c into the cytoplasm. 60,61 Moreover, this effect occurs with many antimitotic agents and in a variety of cell lines. 62−64 As shown in Figure 8  One consequence of mitochondrial depolarization caused by the release of cytochrome c into the cytoplasm is the increase in reactive oxygen species (ROS). 65 Therefore, we wanted to evaluate whether ROS production increased following treatment with compound 66. To do this, we used the fluorescent probe 2,7-dichlorodihydrofluorescein diacetate (H 2 -DCFDA), which is oxidized to the fluorescent compound dichlorofluorescein (DCF) upon ROS production. The results of the cytofluorimetric analysis are presented in Figure 8 (panel B), which demonstrates that 66 induced the production of ROS in HeLa cells after a 48 h treatment at 0.5 μM, in agreement with the reduction of Δψ mt . Note that the increase in ROS is only detectable after mitochondrial depolarization, indicating that ROS production results from mitochondrial damage.  (Figure 9). Similarly, expression of Xiap, a member of the family of inhibitors of apoptosis proteins, was reduced (at 24 h) and diappeared (at 48 h) after HeLa cell treatment with 66 ( Figure 9). The functions of this protein are to inhibit the activity of caspase-3, caspase-7, and caspase-9 through a direct interaction with these enzymes. Following this interaction, the entire apoptotic process is inhibited. 69 Thus, treatment of   We show here that expanding the central ring to seven members, in part to mimic the seven-member rings of colchicine, resulted in enhanced antiproliferative activities in multiple cell lines. This was based on increased antitubulin activity, which in turn caused cell cycle arrest at G2/M, with resultant apoptosis. Molecular modeling rationalized the improvement in activity by central ring expansion, probably caused by an improvement in binding affinity for the colchicine binding pocket because of a greater contribution of the lipophilic energy components.
Among these compounds, five derivatives (62, 63, 66, 67, and 75) showed promising antiproliferative effects, and in particular, 66 and 67, bearing a methoxysubstituted N-benzyl moiety and an ethoxycarbonyl group, reached nanomolar growth inhibitory effects against solid and liquid tumor cells and submicromolar activity against lymphoma cell lines. Their mechanism of action is probably through inhibition of tubulin assembly by binding in the colchicine site, and this mechanism was particularly marked for 66, which inhibited tubulin polymerization with an IC 50 of 2.6 μM and inhibited colchicine binding by 62% under the conditions examined.
Investigation of the mechanism of action showed the ability of the new [1,2]oxazoles to impair cell cycle progression and induce apoptosis through the mitochondrial pathway. The most active compound 66 was able to arrest HeLa cells in the G2/M phase of the cell cycle in a concentration dependent manner. This effect was accompanied by apoptosis, mitochondrial depolarization, generation of ROS, and activation of PARP cleavage. These results indicate that the cellular actions of these agents involved mitotic arrest, due to interference with the functions of the mitotic spindle, and an apoptotic cell death. Taken together, the biological results collected so far indicate that our class of [1,2]oxazoles might find an important place in the set of molecules of interest for the development of pharmaceutical strategies against cancer. Further evolution of this class in terms of ADMET profile will be considered to establish the best trade-off between biological activity and drug-like properties for further preclinical studies.

■ EXPERIMENTAL SECTION
Chemistry. Synthesis and Characterization. MW irradiation was performed using a CEM Discover Labmate apparatus. All melting points were taken on a Buchi melting point M-560 apparatus. IR spectra were determined in bromoform with a Shimadzu FT/IR 8400S spectrophotometer. 1 (22−24). To a solution of 7, 8 (9 mmol) in anhydrous DMF (17 mL), NaH (0.24 g, 10 mmol) was added at 0°C, and the reaction mixture was stirred at room temperature for 1.5 h. Then the suitable alkyl or aralkyl halide (13.5 mmol) was added at 0°C, and the reaction mixture was stirred at room temperature until the reaction was complete (TLC). Then the reaction mixture was poured onto crushed ice. The precipitate was removed by filtration and dried. If there was no precipitate, the solution was extracted with dichloromethane (3 × 50 mL). The organic layer was dried over Na 2 SO 4 , and the solvent was removed under reduced pressure. The crude product was purified by column chromatography, with dichloromethane as eluting solvent.  In all cases, the reaction mixtures were poured onto crushed ice. The precipitate was removed by filtration and dried. If there was no precipitate, the solution was extracted with ethyl acetate (3 × 30 mL  Journal of Medicinal Chemistry pubs.acs.org/jmc Article in anhydrous toluene (40 mL) was added, and the reaction mixture was stirred at room temperature for 1.5 h. Then a solution of ethyl formate (1.09 mL, 13.5 mmol) in anhydrous toluene (12 mL) was added at 0°C, and the reaction mixture was stirred until the reaction was complete (1.5−4 h). The solvent was removed under reduced pressure, and water (50 mL) was added to the residue. The aqueous phase was acidified with 3 N HCl and extracted with dichloromethane (2 × 60 mL). The organic phase was dried over Na 2 SO 4 , and the solvent was removed under reduced pressure. The crude product was purified by column chromatography with dichloromethane as eluting solvent.  65 mmol) was added, and the reaction mixture was heated at reflux for 1 h. Then the reaction mixture was poured onto crushed ice. The precipitate was removed by filtration and dried. If there was no precipitate, the solution was extracted with dichloromethane (3 × 20 mL). The organic layer was dried over Na 2 SO 4 , and the solvent was removed under reduced pressure. The crude product was purified by column chromatography with dichloromethane/ethyl acetate 95:5 as eluting solvent.
9-(Phenylsulfonyl) -4,5,6,9-  Cell Proliferation Analysis. The antiproliferative activity of all compounds was assessed by using the 3-(4.5-dimethylthiazolyl-2)-2,5diphenyltetrazolium bromide (MTT) test. Cells were seeded in 96well plates (nontissue culture treated) at a density of 1 × 10 5 cells/mL and treated with a single concentration of 1 μM for 72 h. Selected compounds that reached proliferation inhibition below 60% were further tested in order to calculate IC 50 values. In this case, cells were treated in triplicate with serially diluted compounds in the appropriate tissue culture medium at a range of 40−10 000 nM. Cells were incubated for 72 h at 37°C, 5% CO 2 . Wells containing medium only were included on each plate and served as blanks for absorbance readings. MTT (Sigma, Buchs, Switzerland) was prepared as a 5 mg/ mL stock solution in phosphate buffered saline (PBS) and filtersterilized. MTT solution (22 μL) was added to each well, and tissue culture plates were incubated at 37°C for 4 h. Cells were then lysed with 25% sodium dodecyl sulfate lysis buffer, and absorbance was read at 570 nm using a Beckman Coulter-AD340 plate reader.
Evaluation of Cytotoxicity in PBLs. PBLs were obtained from human peripheral blood (leucocyte rich plasma-buffy coats) from healthy volunteers using the Lymphoprep (Fresenius KABI Norge AS) gradient density centrifugation.
Buffy coats were obtained from the Blood Transfusion Service, Azienda Ospedaliera of Padova and provided at this institution for research purposes. Therefore, no further informed consent was needed. In addition, buffy coats were provided without identifiers. The experimental procedures were carried out in strict accordance with approved guidelines.
After extensive washing, cells were resuspended (1.0 × 10 6 cells/ mL) in RPMI-1640 with 10% fetal bovine serum and incubated overnight. For cytotoxicity evaluations in proliferating PBL cultures, nonadherent cells were resuspended at 5 × 10 5 cells/mL in growth medium, containing 2.5 μg/mL PHA (Irvine Scientific). Different concentrations of the test compounds were added, and viability was determined 72 h later by the MTT test. For cytotoxicity evaluations in resting PBL cultures, nonadherent cells were resuspended (5 × 10 5 cells/mL) and treated for 72 h with the test compounds.
Tubulin Studies. Electrophoretically pure bovine brain tubulin was obtained as described previously. 70 Analysis of effects on tubulin polymerization was performed by turbidimetry at 350 nm in recording spectrophotometers equipped with electronic temperature controllers as described in detail elsewhere. 71 The tubulin used in these studies was more active than that used in ref 71, and so the concentration of tubulin was reduced from 10 to 9 μM (1.0 to 0.9 mg/mL) and the concentration of GTP from 0.4 to 0.2 mM. This was done to obtain an IC 50 for CA-4, the reference compound, similar to that obtained in ref 71. The binding of [ 3 H]colchicine to tubulin was perfomed as described in detail previously 72 except that the tubulin concentration was reduced from 0.1 mg/mL to 0.05 mg/mL and only one, instead of two, DEAE-cellulose filter was used for each reaction mixture.
Flow Cytometric Analysis of Cell Cycle Distribution. 5 ×10 5 HeLa cells were treated with different concentrations of the test compounds for 24 h. After the incubation period, the cells were collected, centrifuged, and fixed with ice-cold ethanol (70%). The cells were treated with lysis buffer containing RNase A and 0.1% Triton X-100 and stained with PI. Samples were analyzed on a Cytomic FC500 flow cytometer (Beckman Coulter). DNA histograms were analyzed using MultiCycle for Windows (Phoenix Flow Systems).
Apoptosis Assay. Cell death was determined by flow cytometry of cells double stained with annexin V/FITC and PI. The Coulter Cytomics FC500 (Beckman Coulter) was used to measure the surface exposure of phosphatidylserine on apoptotic cells according to the manufacturer's instructions (Annexin-V Fluos, Roche Diagnostics).
Assessment of Mitochondrial Potential and ROS. The mitochondrial membrane potential was measured with the lipophilic cationic dye 5,5′,6,6′ tetrachlo-1,1′,3,3′-tetraethylbenzimidazolcarbocyanine (JC-1) (Molecular Probes), as described. 73 The method is based on the ability of this fluorescent probe to enter selectively into mitochondria since it changes reversibly its color from green to red as membrane potential increases. This property is due to the reversible formation of JC-1 aggregates upon membrane polarization that causes a shift in the emitted light from 530 nm (i.e., emission of JC-1 monomeric form) to 590 nm (emission of JC-1-aggregate) when excited at 490 nm.
The production of ROS was measured by flow cytometry using H 2 DCFDA (Molecular Probes), as previously described. 59 Briefly, after different times of treatment, cells were collected by centrifugation and resuspended in PBS containing H 2 DCFDA at the concentration of 0.1 μM. The cells were then incubated for 30 min at 37°C, centrifuged, and resuspended in PBS. The fluorescence was directly recorded with the flow cytometer, using as excitation wavelength 488 nm and emission at 530 nm.
Western Blot Analysis. HeLa cells were incubated in the presence of the test compound and, after different times, were collected, centrifuged, and washed two times with ice cold PBS. The pellet was resuspended in lysis buffer. After the cells were lysed on ice for 30 min, lysates were centrifuged at 15 000g at 4°C for 10 min. The protein concentration in the supernatant was determined using the BCA protein assay reagents (Pierce, Italy). Equal amounts of protein (10 μg) were resolved using sodium dodecyl sulfate−polyacrylamide gel electrophoresis (Criterion Precast, BioRad, Italy) and transferred to a PVDF Hybond-P membrane (GE Healthcare). Membranes were blocked with a bovine serum albumin solution (5% in Tween PBS 1×), and the membranes were gently rotated overnight at 4°C in the albumin solution. Membranes were then incubated with primary antibodies against PARP cleaved fragment, cdc25c, cyclin B, p-cdc2 Tyr15 , XIAP, and Mcl-1 (all from Cell Signaling) or GAPDH (Sigma-Aldrich) for 2 h at room temperature. Membranes were next incubated with peroxidase labeled secondary antibodies for 1 h. All membranes were visualized using ECL Select (GE Healthcare), and images were acquired using an Uvitec-Alliance imaging system (Uvitec, Cambridge, U.K.). To ensure equal protein loading, each membrane was stripped and reprobed with anti-GAPDH antibody. To obtain relative quantitative data, ImageJ software (NIH, USA) was used for scanning densitometry analysis of Western blots.
Molecular Modeling. Docking Studies. All molecular modeling simulations were carried out using the Schrodinger Suite version 2018. 74 In particular, the LigPrep tool 75 was used to model the 3D structure of each ligand, to calculate and to energy minimize their protonation state at pH 7.4 using OPLS_2005 as force field. 76 First, docking studies of new derivatives were performed by using three different crystal structures of tubulin, characterized by two dimers of α−β tubulin heterodimers, downloaded from the Protein Data Bank (PDB). 77 Models having PDB codes 4O2B and 1Z2B were selected as colchicine-bound 78 and vinblastine-bound 79 cocrystal structures, respectively. According to literature data, the tubulin colchicine domain consists of the main site, where colchicine binds (zone 1), and two additional neighboring pockets (zones 2 and 3). 80 To better discriminate the most likely binding area of the new derivatives, in addition to the 4O2B structure, representing the colchicine-like binding site area (zones 1 and 2), the model with PDB code 3N2G, 80 cocrystallized with the inhibitor G2N, was also used in docking studies to represent the binding zones 2 and 3. Each X-ray model was preprocessed using the Protein Preparation Wizard tool and the OPLS_2005 force field, in order to add hydrogen atoms, to assign partial charges and to build missing atoms, side chains, and loops. The nucleotides (GTP and GDP) and the metals (Mg 2+ and Zn 2+ ) were retained during the docking calculations, while all water molecules were removed. Thus, the docking grids were prepared using as centroid the cocrystallized ligands (i.e., colchicine for 1SA0, vinblastine for 1Z2B, and G2N for 3N2G), while box size and position were generated automatically. Docking studies were performed by using the software Glide version 7.8 81 and by applying the Glide Extra-Precision (XP) protocol, selected after redocking analysis (for details, see Supporting Information paragraph "Redocking Analysis" and Table S1). Ten poses per ligand were taken into account, and the default docking scoring function was used for selecting the best binding mode for each ligand. To perform the following computational studies, for each tubulin crystal structure only one α-tubulin and one β-tubulin structure were selected, specifically the C and D chains for 4O2B, the B and C chains for 1Z2B, and the A and B chains for 3N2G, with respect to the redocking analysis.
Molecular Dynamics Simulations (MDs). To better characterize the binding mode of our best active compounds (57,58,63,66,67,75, and 3), their complexes with the 4O2B model were subjected to molecular dynamics simulations (MDs). The Desmond package 82 was used for MDs, employing OPLS_2005 as force field in an explicit solvent (TIP3 water model). 83 The best docking pose for each single compound was taken as initial coordinates for the MDs. An orthorhombic water box was built for the solvation of the system, ensuring a buffer distance of approximately 10 Å between each box side and the complex atoms. The system was neutralized by adding K + counterions, and it was minimized and pre-equilibrated using the default relaxation routine implemented in Desmond. Simulation time was set to 20 ns, under NPT conditions at 1 atm and 300 K, with a recording interval equal to 40 ps. The time step was set to 2 fs. MD analyses were performed using the Simulation Event Analysis tool of Desmond, while visualization of each protein−ligand complex was carried out using Maestro. For each compound, average ligand RMSD value was calculated on their heavy atoms by first aligning the complex on the protein backbone of the reference structure. Moreover, by use of the Desmond Trajectory Clustering tool, the best representative structure of the whole MDs was generated in order to examine the possibility of induced-fit binding events of our compounds. Finally, these selected structures were submitted to the calculation of the ΔG bind value by using the MM/GBSA method as implemented in the Prime module 84 from Maestro using the default settings.
Statistical Analysis. The differences between different treatments were analyzed, using the two-sided Student's t test. P values lower than 0.05 were considered significant.  Table S1 listing RMSD (Å) and G-Score (kcal/mol) values, obtained by means of Glide Extra-Precision (XP), for each chain in the dimer used in the crystallographic structures used in computational simulations; Table S2 listing G-Score (kcal/mol) values for the best poses of all compounds complexed with both the 4O2B and 1Z2B crystallographic structures; Table S3 listing G-Score values, expressed as kcal/mol, of the best active compounds 57, 58, 63, 66, 67, 75, and 3 against both the 4O2B and 3N2G crystallographic structures; Figure  S5 showing the best-docked poses of 57, 58, 63, 66, 67, 75, and 3 against the crystal structure of tubulin with the PDB code 3N2G, depicting zones 2 and 3 of the colchicine site; geometric and energetic analysis of molecular dynamics simulations (MDs); Table S4  Author Contributions ◇ V.S., R.R., and M.B. contributed equally.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
This work was financially supported by the Ministero dell'Istruzione, dell'Universitàe della Ricerca (MIUR). This research was supported in part by the Developmental Therapeutics Program in the Division of Cancer Treatment and Diagnosis of the National Cancer Institute, which includes federal funds under Contract HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. The authors also thank the Developmental Therapeutic Program of the National Cancer Institute for performing cytoxicity studies with selected compounds in the 60 cancer cell line screen. The authors also acknowledge the Italian Association for Cancer Research (AIRC) research project "Small molecule-based targeting of lncRNAs 3D structure: a translational platform for the treatment of multiple myeloma" (Code 21588), the PRIN 2017 research project "Novel anticancer agents endowed with multi-targeting mechanism of action" (Code 201744BN5T), and the PRIN 2017 research project "Selective mGlu3 metabotropic glutamate receptor ligands as new potential therapeutic agents in experimental models of parkinsonism" (Code 2017XZ7A37) funded by the Italian MIUR.