Novel Piperazine-based Compounds Inhibit Microtubule Dynamics and Sensitize Colon Cancer Cells to Tumor Necrosis Factor-induced Apoptosis*

Background: Mitotic spindles are important targets of cancer chemotherapeutic agents. Results: A new class of tubulin-targeting agents is identified that effectively sensitizes colon cancer cells to ligand-induced apoptosis. Conclusion: AK301 is a novel, piperazine-based compound that induces mitotic arrest and increases ligand-dependent apoptosis. Significance: Mitotically active compounds that stimulate ligand-induced apoptotic signaling might be useful for augmenting immune-based cancer therapies. We recently identified a series of mitotically acting piperazine-based compounds that potently increase the sensitivity of colon cancer cells to apoptotic ligands. Here we describe a structure-activity relationship study on this compound class and identify a highly active derivative ((4-(3-chlorophenyl)piperazin-1-yl)(2-ethoxyphenyl)methanone), referred to as AK301, the activity of which is governed by the positioning of functional groups on the phenyl and benzoyl rings. AK301 induced mitotic arrest in HT29 human colon cancer cells with an ED50 of ≈115 nm. Although AK301 inhibited growth of normal lung fibroblast cells, mitotic arrest was more pronounced in the colon cancer cells (50% versus 10%). Cells arrested by AK301 showed the formation of multiple microtubule organizing centers with Aurora kinase A and γ-tubulin. Employing in vitro and in vivo assays, tubulin polymerization was found to be slowed (but not abolished) by AK301. In silico molecular docking suggests that AK301 binds to the colchicine-binding domain on β-tubulin, but in a novel orientation. Cells arrested by AK301 expressed elevated levels of TNFR1 on their surface and more readily activated caspases-8, -9, and -3 in the presence of TNF. Relative to other microtubule destabilizers, AK301 was the most active TNF-sensitizing agent and also stimulated Fas- and TRAIL-induced apoptosis. In summary, we report a new class of mitosis-targeting agents that effectively sensitizes cancer cells to apoptotic ligands. These compounds should help illuminate the role of microtubules in regulating apoptotic ligand sensitivity and may ultimately be useful for developing agents that augment the anti-cancer activities of the immune response.

We recently identified a series of mitotically acting piperazine-based compounds that potently increase the sensitivity of colon cancer cells to apoptotic ligands. Here we describe a structure-activity relationship study on this compound class and identify a highly active derivative ((4-(3-chlorophenyl)piperazin-1-yl)(2-ethoxyphenyl)methanone), referred to as AK301, the activity of which is governed by the positioning of functional groups on the phenyl and benzoyl rings. AK301 induced mitotic arrest in HT29 human colon cancer cells with an ED 50 of ≈115 nM. Although AK301 inhibited growth of normal lung fibroblast cells, mitotic arrest was more pronounced in the colon cancer cells (50% versus 10%). Cells arrested by AK301 showed the formation of multiple microtubule organizing centers with Aurora kinase A and ␥-tubulin. Employing in vitro and in vivo assays, tubulin polymerization was found to be slowed (but not abolished) by AK301. In silico molecular docking suggests that AK301 binds to the colchicine-binding domain on ␤-tubulin, but in a novel orientation. Cells arrested by AK301 expressed elevated levels of TNFR1 on their surface and more readily activated caspases-8, -9, and -3 in the presence of TNF. Relative to other microtubule destabilizers, AK301 was the most active TNF-sensitizing agent and also stimulated Fas-and TRAIL-induced apoptosis. In summary, we report a new class of mitosistargeting agents that effectively sensitizes cancer cells to apoptotic ligands. These compounds should help illuminate the role of microtubules in regulating apoptotic ligand sensitivity and may ultimately be useful for developing agents that augment the anti-cancer activities of the immune response.
Eukaryotic cell division involves replication of DNA in S phase followed by equal segregation of mitotic chromosomes during anaphase (1). Cell cycle checkpoints have evolved to ensure faithful DNA replication and chromosomal division. Cells that harbor defective cell cycle checkpoint regulators can result in genetic instability and aneuploidy, ultimately leading to tumor development (2). Mitosis orchestrates multiple cellular changes and depends on many intricate signaling pathways, despite being the shortest phase of the cell cycle (3). Signaling pathways, including kinases and several checkpoint proteins, spatiotemporally regulate dynamic chromosomal rearrangements and reorganization. It is because of this intricacy that mitosis is considered the most sensitive phase of the cell cycle (4,5). Damage to cellular processes that affect mitosis can activate spindle assembly checkpoint, which delays progression into anaphase (6). Prolonged arrest in mitosis makes the cells more sensitive to cellular insults, which has likely made mitosis a desirable target for chemotherapy (4,(7)(8)(9). Aneuploidies and other genomic and chromosomal abnormalities can induce cellular stress on cancer cells and make them highly sensitive to agents that disrupt mitosis (10).
Microtubules form spindle fibers during mitosis that are critical for chromosomal alignment and segregation (11,12). Previous findings suggest that agents that target the mitotic spindle can make highly effective chemotherapeutic drugs. Successful use of several vinca alkaloids, taxanes, and other natural compounds for the treatment of human cancers has validated the effectiveness of microtubule-targeting drugs (13)(14)(15). Several other mitotic proteins have also emerged as potential targets of chemotherapy. These targets include kinases, motor proteins, proteasome inhibitors, and inhibitors of chromatin reorganizing proteins. Some of these newly developed compounds may provide clinical benefits over some of the presently used drugs (4).
One of the primary challenges of cancer chemotherapeutics is the targeting of cancer cells while sparing normal cells of the surrounding tissue (16). The use of vaccines and immune stimulants to specifically target tumors has generated promising results. For colon cancer, complementing traditional chemo-therapy with IL-2 and granulocyte/macrophage colony-stimulating factor was shown to significantly increase patient survival (17). However, immune stimulants can sometimes result in modest cell killing activity. Cell killing by the activated immune response includes direct cell killing by cytotoxic T cells and NK cells, as well as cell killing apoptotic ligands, such as TNF.
We previously reported several novel synthetic small molecules that dramatically increase colon cancer cell death by TNF and other death ligands, while being unable to induce apoptosis on their own (18). Interestingly, many of these compounds also induced mitotic arrest. To gain insight into the mechanisms of action of these compounds, we studied the structure-activity relationship of a particularly promising class of piperazinebased compounds. Here we report a structure-activity relationship study of this class of compounds and identify a highly active derivative, AK301. Furthermore, we show that AK301 hampers tubulin polymerization, triggers the formation of multiple microtubule organizing centers (MTOCs), 2 and increases the surface expression of TNFR1. Molecular docking studies indicate that AK301 binds to ␤-tubulin near the colchicinebinding site, but in a novel orientation. Lastly, AK301 was found to be more effective in sensitizing cancer cells to TNF-induced apoptosis than other known microtubule-destabilizing agents. We propose that AK301 and its derivatives represent a novel class of microtubule-targeting compounds that will be useful for studying the relationship between microtubule dynamics and apoptosis sensitivity. This class of compounds may also have beneficial therapeutic properties because of their ability to sensitize cancer cells to ligand-induced apoptosis.

EXPERIMENTAL PROCEDURES
Cell Culture-The HT29 and HCT116 colon cancer and WI38 fibroblast cell lines were obtained from the American Type Culture Collection. HT29 and WI38 cell lines were cultured in McCoy's 5A medium and minimum Eagle's medium, respectively, with 10% fetal bovine serum, nonessential amino acids, and antibiotic/antimycotic (Invitrogen). The compounds tested were obtained from the ChemBridge DIVERSet TM library (San Diego, CA). Drug treatments were performed ϳ24 h after passage for 18 h, unless otherwise indicated. TNF was obtained from Pierce, TRAIL was obtained from R&D Systems, and ␣-Fas antibody (clone CH11) was obtained from Millipore.
Flow Cytometry-HT29 and WI38 were analyzed for DNA content by ethanol fixation and staining with propidium iodide as described previously (19). Floating and adherent cells were combined and analyzed by flow cytometry. Adherent cells were harvested using trypsin-EDTA, centrifuged together with the floating cells at 100 ϫ g for 5 min, and resuspended in 1 ml of cold saline with 6 mM glucose and 0.5 mM EDTA. Cells were then fixed by adding 3 ml of cold 100% ethanol while gently vortexing and stored at Ϫ20°C for at least 2 h. Cells were then pelleted and washed once with PBS containing 5 mM EDTA. Pelleted cells were stained with 30 g/ml propidium iodide (Molecular Probes, Invitrogen) and 0.3 mg/ml RNase A (Sigma-Aldrich) in 1 ml PBS solution for 40 min in dark at room temperature. The stained cells were filtered through 35-m cell strainer tubes (BD Biosciences) prior to analysis on FACSCalibur flow cytometry (BD Biosciences) using Cell Quest software (BD Biosciences). The data were analyzed using FlowJo (version 9.6.2 for Mac; TreeStar Inc., Ashland, OR).
Cell Viability Assay-Cell viability was assessed using trypan blue exclusion assay. After treatment, the cells were incubated with trypan blue at room temperature. Viable/dye excluding cells were then counted using a hemocytometer.
Immunofluorescence Microscopy-Cells cultured on coverslips were fixed with 4% paraformaldehyde or 100% ice-cold methanol and then permeabilized with 0.5% Triton X-100 in PBS. Cells were blocked in 5% serum (in PBS) and then incubated for 1 h at room temperature on the shaker with the primary antibody (in 5% serum) against phospho-histone H3 Ser-28 (sc-12927; Santa Cruz Biotechnology), ␤-tubulin (E7 monoclonal antibody; Developmental Studies Hybridoma Bank), or Aurora kinase A (BD Biosciences). ␥-Tubulin antibody (Abcam) incubation was performed overnight at 4°C. Appropriate secondary antibodies (Jackson ImmunoResearch) were used for 45 min of incubation. Nuclei were visualized using DAPI (5 g/ml in PBS; DI306; Invitrogen). Coverslips were mounted on slides using ProLong Gold Antifade Reagent (Invitrogen). Images were acquired using a Nikon A1R confocal microscope (version 2.11; Nikon Instruments Inc.) and NIS-Elements Advanced Research Software (version 4.13.01, build 916; Nikon Instruments Inc.). Quantification of immunostaining was performed using ImageJ image analysis software as described previously (20). Following background subtraction and image stacking, both DAPI and immunofluorescence images were merged. Image brightness and contrast was modified with Adobe Photoshop software CS6 (Adobe Systems).
In Vitro Tubulin Polymerization Assay-The HTS-tubulin polymerization assay kit (BK004P; Cytoskeleton, Inc., Denver, CO) was used as per manufacturer instructions. The reaction assay contained 100 l of 4 mg/ml tubulin in G-PEM buffer (80 mM PIPES, pH 6.9, 0.5 mM EGTA, 2 mM MgCl 2 , and 1 mM GTP). 10 l of 10ϫ compounds were prewarmed to 37°C in a half area 96-well plate (distilled H 2 O was used as control). The polymerization was carried out at 37°C, and light scattering was recorded at 340 nm every minute for 60 min using Spectramax M2 absorbance plate reader (Molecular Devices, Sunnyvale, CA).
Whole Cell Microtubule Analysis-Microtubules in whole cells were analyzed by flow cytometry as described previously (21). Cells were cultured in 24-well plates for 24 -36 h and treated with the colchicine, AK301, or AK302 for 16 h. The medium was collected, and the cells were harvested by trypsin EDTA treatment and pelleted by centrifugation at 600 ϫ g for 5 min. Cell pellets were resuspended and fixed with 0.5% glutaraldehyde under permeabilizing conditions in microtubule stabilizing buffer (80 mM PIPES, pH 6.8, 1 mM MgCl 2 , 5 mM EDTA, and 0.5% Triton X-100) for 10 min. Glutaraldehyde was quenched with 700 l of 1 mg/ml NaBH 4 in PBS. Cells were pelleted by centrifugation at 1000 ϫ g for 7 min. Cells were blocked with 5% donkey serum and immunostained with ␤-tu-bulin (E7 monoclonal) antibody for 1 h at room temperature, followed by secondary staining with Alexa Fluor 488 donkey anti-mouse antibody (Invitrogen) for 1 h. Finally, cells were pelleted by centrifugation and treated with 0.3 mg/ml of RNase A and 50 g/ml of propidium iodide solution in PBS. The cells were analyzed by flow cytometry. All steps in this protocol were carried out at room temperature.
In Silico Molecular Docking-Structural representations of the ligand molecules (AK3, AK301, AK302, AK303, and AK304) were drawn using Accelrys Draw (version 4.1; Accelrys, Inc.) in MOL2 format and converted to Protein Data Bank format using Accelrys Discovery Studio Client (version 3.5). Individual Protein Data Bank files were modified in AutoDock using MGLTools 1.5.6 (Scripps Institute). Crystal structures of tubulin complexed with colchicine, paclitaxel, and vinblastine (Protein Data Bank codes 1SA0 (22), 1TUB (23), and 4EB6 (24), respectively) were obtained from the Protein Data Bank. Water molecules, ligands, and other heteroatoms were removed from the protein molecules using Accelrys Discovery studio client (version 3.5; Accelrys, Inc.). Addition of hydrogen atoms to the protein was performed using MGLTools (version 1.5.6) for AutoDock. For each known ligand type, grid maps were generated that corresponded to their respective known binding sites on tubulin.
AutoDock 4.2 and Vina 1.1.2 were used for initial docking studies. Generally, the docking parameters were left to the default settings. However, the grid spacing was changed from 0.375 to 1.0. The size of the grid was 30 ϫ 30 ϫ 30 Å. The internal scoring function was used to assess receptor-ligand interactions in five independent runs.
Additionally, Molegro Virtual Docker software version 6.0 was used to perform computer simulated docking analysis to confirm the least energy poses acquired using AutoDock Vina. Charges for both tubulin and the ligands were calculated by Molegro Virtual Docker and assigned to their respective models. Moreover, probable explicit hydrogens were added to tubulin as well as the ligands, possible missing bonds were assigned, and side chain minimization was performed. Finally, flexible torsions were manually applied to the ligands. Because tubulin is a relatively large protein, a molecular surface was created using Molegro Virtual Docker workspace grid points, and the top three cavities were identified using the expanded van der Waals method for the molecular surface with volumes ranging from 5 to 10,000 cubic units and default settings. To perform docking, the cavity containing the colchicine-binding domain was used with a radius of 20 to cover the entire cavity. A MolDock score with a grid resolution of 0.30 Å was used as a scoring function for the simulation (25). Ten runs were performed for each of the ligands using MolDock simplex evolution algorithm for fast and accurate docking and scoring. 1500 iterations were performed for each of the runs to achieve least minimized energy poses for the ligands. During the virtual FIGURE 1. Structural changes in the side groups of AK3 dictate its potency in inducing mitotic arrest. A, structural analogs of AK3 ((4-(3-chlorophenyl)piperazin-1-yl)(2-methoxyphenyl)methanone) with modifications to benzoyl group (AK301, 2-ethoxyphenyl; AK303, 3-methoxyphenyl; and AK304, 4-methoxyphenyl) and phenyl group (AK302, 4-chlorophenyl). B, HT29 cells were treated with 5 M of AK3 analogs for 18 h followed by an assessment of G 2 /M arrest by flow cytometry. The data suggest that AK3, AK301, and AK303 induce G 2 /M phase arrest, whereas AK302 and AK304 are inactive at this concentration (*, p Ͻ 0.0001). Ctrl, control. C, dose-response analysis of AK3-, AK301-, and AK303-induced mitotic arrest by flow cytometry. AK3, AK301, and AK303 were titrated at different concentrations between 0 and 500 nM and quantified G 2 /M arrest by flow cytometry. The results are reported in triplicate as means Ϯ S.E. AK303 induced significantly less mitotic arrest compared with AK3-or AK301-treated cells. AK301 was found to be the most potent compound. D, dose-response curves showing reduced cell growth of AK3-and AK301-treated cells with increasing concentration (in micromol). Cell viability of AK3-or AK301-treated cells was assessed by trypan blue assay in triplicate, and the results are reported as means Ϯ S.E. screening process, internal electrostatic interaction and hydrogen bond between ligand and protein were permitted. Energy minimization and H-bond optimization was applied to each of the poses post-run.
Caspase-3 Assay-Caspase-3 activity was determined as described previously (26). Cells were collected, centrifuged at full speed, and washed once with PBS. Pelleted cells were lysed by two rounds of freeze-thaw in lysis buffer containing 10 mM Tris-HCl, pH 7.5, 0.1 M NaCl, 1 mM EDTA, and 0.01% Triton X-100 and centrifuged at 10,000 ϫ g for 5 min. The assays were performed on a 96-well plate by mixing 50 l of lysis supernatant with 50 l of 2ϫ reaction mix (10 mM PIPES, pH 7.4, 2 mM EDTA, 0.1% CHAPS, 10 mM DTT) containing 200 nM of the fluorogenic substrate acetyl-Asp-Glu-Val-Asp-7-amino-4methylcoumarin (DEVD-AMC; Enzo Life Sciences). The fluorescence was quantified at the start of the reaction and after 30 min. Protein concentrations were determined using CBQCA protein quantitation kit (Invitrogen). Caspase activity was determined by dividing the change in fluorescence by total protein content of the reaction mixture.
Cell Surface TNFR1 Analysis-Cells were treated in a 24-well plate as described above. Treated live cells were transferred to ice, washed once with PBS, and blocked with 5% donkey serum. Cells were immunostained with TNFR1 antibody overnight at 4°C, followed by staining with Alexa Fluor 488 antibody for 1 h at room temperature. Cells were fixed with 4% paraformaldehyde, washed, and harvested by trypsin EDTA. Cells were pelleted by centrifugation and analyzed by flow cytometry.
Statistical Analyses-One-way analysis of variance was used for comparing more than two groups. Tukey's post hoc test was employed to determine the significance of differences between multiple groups, with p Ͻ 0.05 considered significant. Two-way analysis of variance was used for more than two independent variables, and Bonferroni correction was used for multiple comparisons (p Ͻ 0.05).

Functional Groups of AK3 Mediate Mitotic Arrest in HT29
Cells-In previous studies, we identified the AK3 compound from the ChemBridge DIVERSet TM library that induced mitotic arrest in colon cancer cells at low micromolar concentrations (0.5-1 M) (18). To assess properties of the chlorophenyl and methoxybenzoyl functional groups of AK3, four different AK3 analogs were analyzed for their ability to induce mitotic arrest in HT29 colon cancer cells at 5 M (Fig. 1A). The analogs comprise positional isomers of chloride and methoxy groups on the phenyl (AK302) and benzoyl (AK303 and AK304) rings, respectively, as well as the substitution of an ethoxy (AK301) for methoxy on the benzoyl ring. Flow cytometric analysis of propidium iodide-stained HT29 colon cancer cells showed that a greater percentage of AK3-, AK301-, and AK303treated cells arrested in the G 2 /M phase of the cell cycle compared with untreated or AK302-or AK304-treated cells (p Ͻ 0.0001) (Fig. 1B). A titration of the most active compounds among the analogs (AK3, AK301, and AK303) and an assess-ment of the G 2 /M arrest by flow cytometry is shown in Fig. 1C. AK301 induced G 2 /M arrest more efficiently than AK3 or AK303 with a half-maximal effective concentration (EC 50 ) value of 115 nM. Further, HT29 cells showed decreased growth in the presence of AK3 or AK301, with dose response curves showing a higher potency of AK301 (Fig. 1D). Together, these results indicate that the position of the functional groups on  AK3 derivatives play an important role in determining their activity in inducing cell cycle arrest.
AK301 Induces an Irreversible Mitotic Arrest in HT29 Cells-To determine whether AK301-treated cells arrested in mitosis, HT29 cells were analyzed for presence of the mitosis marker phospho-histone H3 Ser-28. Fig. 2 (A and B) shows that both AK3 and AK301 increased phospho-histone H3 Ser-28 staining, consistent with mitotic arrest. To further assess whether this state of mitotic arrest was reversible, AK301 was washed out of the growth medium after 18 h of treatment, and cell growth was observed at 24, 48, and 72 h following drug washout. As shown in Fig. 2B, AK301-treated cells did not recuperate after 72 h of post-drug washout and showed significantly less growth compared with the untreated cells (p Ͻ 0.0001).

AK301 Affects Microtubule Function and Results in Multiple
MTOCs-To determine the effect of AK301 on spindle formation in the arrested cells, we performed immunostaining for ␤-tubulin. As shown in Fig. 3A, AK301 induced the formation of multiple MTOCs, as is the case for AK3-treated HT29 cells. To further determine whether the MTOC assemblies were stable or resulted from spontaneous microtubule assembly, HT29 cells were stained for Aurora A kinase and ␥-tubulin. Aurora  kinase A is responsible for recruitment of ␥-tubulin to established centrosomes, and ␥-tubulin serves to nucleate and orient microtubules. As shown in Fig. 3B, multiple Aurora kinase A and ␥-tubulin centers were observed following AK301 treatment of HT29 cells. These data suggest that AK301 leads to the formation of multiple MTOCs, including deposition of Aurora kinase A and ␥-tubulin at the spindle poles. We performed a dose-response analysis of ␤-tubulin immunostaining to assess the effects of AK301 on the formation of multiple MTOCs. As shown in Fig. 4A, AK301 significantly induced spindle impairment or multipolarity compared with the inactive AK302 and AK304 at concentrations as low as 50 nM. We also quantified the number of Aurora A kinase and ␥-tubulin foci and found that AK301increased the number of these foci as well (Fig. 4, B  and C).
AK301 Induces G 2 Arrest in WI38 Lung Fibroblast Cells-To assess the effects of AK301 on normal cells, we performed a dose response analysis for growth inhibition on WI38 lung fibroblast cells. As shown in Fig. 5A, WI38 cells treated with AK3 or AK301 showed decrease in growth, similar to HT29 cells (Fig. 1D). Comparison of the cell cycle data between AK301-treated HT29 and WI38 cells showed that AK301 induced a G 2 /M arrest in both cell lines (Fig. 5B). However, quantification of mitosis using the mitosis-specific phosphohistone H3 staining revealed that AK301 induced significantly higher mitotic arrest in HT29 cells compared with WI38 cells (Fig. 5C). Analysis of AK301-treated WI38 cells that were arrested in mitosis showed that they lacked multiple MTOCs and instead showed a more extensive microtubule breakdown (Fig. 5D, arrowheads). AK301 altered microtubules of interphase cells; untreated WI38 cells showed elongated and spindle-shaped microtubules, whereas AK301-treated cells displayed a meshlike network of microtubules. Together, these results suggest that AK301 can have cell-specific effects and that these effects may be mediated by changes in microtubule dynamics.
AK301 Affects the Rate of Tubulin Polymerization in Vitro-Given the formation of multiple MTOCs in AK301-treated HT29 cells (Fig. 3, A and B) and the alteration of microtubule structures in WI38 cells, we assessed the effect of AK301 on the rate of tubulin polymerization, in an in vitro tubulin polymerization assay. As shown in Fig. 6A, addition of AK301 slowed the rate of tubulin polymerization compared with the control sample. In contrast, AK302, identified as the inactive analog of AK3, did not alter the rate of polymerization. Colchicine, a known inhibitor of microtubule polymerization, acts by binding the ␤-subunit of tubulin dimer and prevents their addition to the growing polymer. Microtubule formation was almost completely inhibited in the presence of colchicine, compared with AK301 (Fig. 6B).
To assess the effect of AK301 on microtubule dynamics in vivo, we performed quantitative whole cell microtubule analysis as described previously (21). Fig. 7A outlines this procedure that quantifies the degree of tubulin polymerization by flow cytometry. Fig. 7B shows the mean fluorescence intensity of mitotic cells treated with colchicine (500 nM), AK301 (250 nM), and its inactive analog, AK302 (1 M). Colchicine shows the lowest microtubule fluorescence intensity, consistent with microtubule destabilization by colchicine. However, AK301 treatment shows a staining intensity intermediate to that of control and colchicine. AK302 did not induce significant reduction in staining. These data, along with the in vitro experiments, suggest that AK301 induces a partial breakdown of microtubules.
In Silico Molecular Modeling-We performed molecular docking studies to assess tubulin as the potential target of AK301. We employed automated docking using AutoDock Vina with tubulin as the target receptor and assessed SAR of the analogs to identify in silico conformations (29). Colchicine, paclitaxel, and vinblastine were docked to their respective sites on tubulin to check for accuracy of molecular docking predictions (supplemental Figs. S1 and S2) (22)(23)(24). AK3 and its analogs were docked to ␤-tubulin in the colchicine-, paclitaxel-, and vinblastine-binding sites. AutoDock Vina reported multiple conformations and corresponding binding affinities for each of the compounds in five independent trials. AK3 and its active analogs, AK301 and AK303, docked in the same positions as indicated by superimposition of central piperazine core and the flanking functional groups of the small molecules (supplemental Fig. S1). The lowest energies for different compounds in known binding sites are shown in Table 1. Control binding affinities correspond to the energies obtained by docking colchicine, paclitaxel, vinblastine, and vincristine into their respective binding sites (vinblastine and vincristine were docked into the high affinity sites on tubulin). Active AK3 analogs showed compatibility for the colchicine site. In addition,  the order of the predicted binding affinity matched the potency of AK3 analogs, as predicted by SAR studies. Fig. 8A shows an in silico model of AK301 binding to tubulin in the colchicine-binding domain. Although AK301 docked into the colchicine-binding site, it assumed a different, novel orientation relative to colchicine. This orientation allows for hydrogen bond interactions between the oxygen atom of the ethoxy group or the carbonyl group and the Asn-101 residue (3.0 and 2.9 Å, respectively). Further, this places the hydrophobic chloride proximal to hydrophobic tubulin residues: Leu-255 (3.9 Å) and Ile-378 (3.5 Å) (Fig. 8B). In contrast, moving the chloride group from 3Ј C to 4Ј C of the halophenyl ring to generate the less active AK302 compound changes the conformation of the molecule in the tubulin-binding site. The changed conformation of AK302 decreased hydrophobic interactions with Leu-255 (6.7 Å) and Ile-378 (9.6 Å) and increased the distance for hydrogen bond interactions (6.6 Å for methoxy group and 5.8 Å for the carbonyl group) at the other end of the molecule (Fig. 8C). Similar changes in conformation were also observed when the position of the methoxy group was moved from 2Ј C to 4Ј C, as is the case with AK304 compound. In general, the docking scores obtained correlated with the predicted in vitro activity and fit the SAR of AK3 analogs, with AK301 being the most potent among the analogs (Table 1). To confirm AK301 binding in the colchicine-binding domain of tubulin, we also performed molecular docking simulations in Molegro Virtual Docker (25). We performed 10 runs for AK301, targeting the largest cavity in tubulin, containing the colchicine-binding domain of tubulin. MolDock scoring function was used to assess the binding of the lowest energy poses. AK301, as predicted by AutoDock Vina, docked to the colchicine-binding domain of tubulin with the least energy poses superimposing each other (Fig. 8D). The specific functional groups and their positions may therefore play an important role in determining the potency of AK3 analogs in inducing mitotic arrest in colon cancer cells.
TNF-dependent Induction of Apoptosis in AK301-treated Cells-AK3 and other piperazine-based compounds were originally identified by their ability to acutely sensitize colon cancer cells to ligand-induced apoptosis (18). To determine whether AK301 could also induce cell death in combination with TNF, we performed a dose-response analysis of AK301 in the presence and absence of TNF (Fig. 9A). Analysis of the subdiploid cell population indicated that AK301, on its own, did not induce apoptosis in HT29 colon cancer cells. However, in the presence of TNF, AK301-treated cells underwent significant apoptosis (EC 50 of 172 nM) at concentrations ϳ5 times lower than those reported for AK3 (18).
To determine the association between mitotic arrest and TNF-induced apoptosis, the compounds shown in Fig. 1A were tested for their ability to induce apoptosis in the presence of TNF. Fig. 9B shows that only the compounds that induced a FIGURE 8. A, novel orientation of AK301 in the colchicine-binding site of tubulin as modeled by AutoDock. AK301 (shown in green) is predicted to bind to tubulin in the colchicine-binding domain (colchicine is shown in red), but in a different orientation. B, residues surrounding AK301 allow for strong hydrogen bond interactions with Asn-101 of tubulin. Consequently, the chloride group is positioned in tubulin surrounded by hydrophobic residues (Leu-255 and Ile-378). C, changing the position of the chloride in AK3 from the 3Ј C to 4Ј C generates the inactive compound AK302. This change disrupts the hydrogen bond interactions of the methoxy and carbonyl groups on the methoxy-substituted benzoyl ring. Decreased hydrogen bond interactions result in lower binding affinity of AK302 to tubulin (Table 1). D, least energy poses of AK301, as predicted by AutoDock Vina (gray) and Molegro Virtual Docker (green), docked to ␤-tubulin. The least energy pose of AK301, predicted by Molegro Virtual Docker (MVD), had a similar orientation to that predicted by AD Vina. mitotic arrest could induce apoptosis in combination with TNF. This finding suggests an association between mitotic arrest and TNF sensitivity. To further assess the correlation between the two events, cell cycle analysis of HT29 cells was performed. As shown in Fig. 9C, most of the cells in the control population are in the G 1 phase of cell cycle. With TNF treatment alone, no significant shifts are observed. However, with the addition of AK301, cells shifted from G 1 to G 2 /M. Finally, co-treatment of HT29 cells with AK301 and TNF resulted in a decrease in G 2 /M population and a subsequent appearance of subdiploid population. These data suggest that the subdiploid population is likely derived from the G 2 /M arrested cells.
Relationship between Mitotic Arrest and Cancer Cell Apoptosis-To further examine the relationship between mitotic arrest induced by AK301 and its analogs and apoptosis, we performed a dose-response analysis of AK301, AK302, and AK304 on mitotic arrest and caspase-3 activation. As shown in Fig. 10A, AK301 induced mitotic arrest in a dose-dependent manner, whereas AK302 and AK304 were inactive even at concentrations of 5 M. Analysis of dose-dependent caspase-3 activity (using DEVD-AMC fluorogenic substrate) for these compounds in the presence of TNF is shown in Fig. 10B. The data indicate trends in caspase-3 activation similar to those observed for mitotic arrest. These results demonstrate a close relationship between mitotic arrest by AK301 and TNF-induced apoptosis.
There are a number of known microtubule destabilizers that are capable of inducing mitotic arrest (30 -32). To determine how these compounds compare with AK301 in sensitizing cells to TNF, HT29 cells were treated with two different concentrations each of AK301, colchicine, nocodazole, and a vinca alkaloid-vincristine in the absence or presence of TNF for 18 h. Cell extracts were prepared for capase-3 enzymatic assay. As shown in Fig. 10C, none of the compounds induced caspase-3 activation on their own. However, upon co-treatment of HT29 cells with the drugs and TNF, significant caspase-3 activation was observed (Fig. 10D). Interestingly, AK301 induced the highest levels of caspase-3, even at 125 nM (p Ͻ 0.0001). Together with the results of previous experiments, this suggests that AK301 is a novel, potent inhibitor of microtubules capable of inducing arrest on its own and increasing cell sensitivity to TNF-induced apoptosis.
Generality of AK301 Activity-To assess the general effects of AK301 sensitization, we determined the effect of AK301 on caspase-3 activation by TRAIL and Fas. As shown in Fig. 11A, AK301 enhanced TRAIL-induced caspase-3 activity. HT29 cells were more sensitive to Fas ligation, but AK301 further enhanced Fas-induced caspase-3 activity, as shown in Fig. 11B. AK301 was also tested on HCT116 cells, another human colon cancer cell line. HCT116 cells were more sensitive to TNFinduced apoptosis than HT29 cells. Despite their sensitivity, AK301 significantly enhanced TNF-induced caspase-3 activa- tion over TNF background (#, p Ͻ 0.01). These data suggest that AK301 is broadly active on different cancer cell lines and with different death ligands, such as TRAIL and Fas.
Increased TNFR1 Cell Surface Expression Mediates Enhanced Caspase-8 and -9 Activation-TNF is coupled to caspase-8 through TNFR1, which in turn can activate procaspase-3 (33). Moreover, activated caspase-8 can interact with the intrinsic death pathway by activating caspase-9 (34). We analyzed the presence of caspase-8 and -9 in AK301-and AK301/TNFtreated cells by immunoblotting. As shown in Fig. 12A, AK301 or TNF treatment alone did not induce caspase-8 or caspase-9 activation. However, cells co-treated with AK301 and TNF showed significant increases in the activation of both of these caspases.
To determine whether caspase-8 was activated by increased TNF-TNFR1 coupling at the cell surface, we assessed the levels of TNFR1 in the membrane fraction of HT29 cells (an Nonidet P-40 extract) relative to a whole cell extract (RIPA). As shown in Fig. 12B, cells treated with AK301 showed an increase in the appearance of TNFR1 in the membrane fraction. Moreover, we observed an increase in TNFR1 in the RIPA extract after treatment with AK301 or TNF alone, which suggests that AK301 may act as a trigger for TNFR1 production. We quantitatively analyzed cell surface expression of TNFR1 in HT29 cells after treatment with AK301 analogs to determine whether TNFR1 surface expression was specific to AK301. As shown in Fig. 12C, HT29 cells treated with AK301 showed an increase in TNFR1 cell surface staining, whereas the other inactive analogs did not. Together, these data suggest that AK301 induces an increase in cell surface expression of TNFR1. This increase in surface expression may facilitate TNF binding and trigger an apoptotic cascade.

DISCUSSION
We previously identified a class of small molecules that induced mitotic arrest in colon cancer cells and sensitized these cells to apoptosis in the presence of death ligands, such as TNF and FasL (18). Here, we performed a detailed structure-activity relationship study on a piperazine-based compound (AK3) that was initially found to be highly effective at sensitizing colon cancer cells to apoptosis. Specifically, we identified 4-(3-chlorophenyl)piperazin-1-yl(2-ethoxyphenyl)methanone, referred to as AK301, which can induce mitotic arrest in colon cancer cells with an EC 50 of ϳ115 nM. This derivative is ϳ5-fold more potent than the original AK3 compound identified in the initial screen. AK301 also increased the sensitivity of HT29 and HCT116 human colon cancer cell lines to TNF-induced apoptosis at relatively low concentrations. AK301 was also capable of inducing cell death in the presence of TRAIL and FasL. Our SAR studies also revealed a number of related compounds that were inactive for both mitotic arrest and TNF sensitization (AK302 and AK304). Together, these compounds indicate a close relationship between mitotic arrest and sensitivity to TNF-induced apoptosis. In addition, these molecules were employed to determine a potential cellular target leading to mitotic arrest and apoptosis.
Characterization of the mitotic arrest state of AK301-treated cells indicated multipolar spindle assembly. The formation of the multipolar spindles was accompanied by appearance of multiple ␥-tubulin and Aurora kinase A staining loci, which is consistent with disruption of centrosome regulation and bipolar spindle formation. Most cancer cells have over-replicated centrosomes (35,36), which are clustered at the poles during mitosis (37); disruption of centrosome clustering by disruption of microtubule spindles by AK301 may prevent the concerted segregation of supernumerary centrosomes and lead to spindle multipolarity (38). These complex, multipolar structures are apparently difficult to resolve because cell division is significantly inhibited even after the removal of AK301. The degree of AK301-induced mitotic arrest is cell type-dependent. Proliferation of the WI38 lung fibroblast cell line was reduced by AK301, but cells arrested more frequently in G 2 than in mitotic phase. Although the reason for this difference in arrest phase is not known, it may be related to the presence of functional cell cycle checkpoints in nontransformed WI38 cells. Previous studies have shown that the CHFR (checkpoint with forkhead and ring finger domains) protein is a critical component in the cellular response to mitotic stress (including stress induced by microtubule disruption) (39,40). Cells expressing a functional CHFR protein can delay entry into mitosis, thereby preventing catastrophic events during mitosis (41,42). On the other hand, studies have shown that HT29 cells and other cancer cells down-regulate CHFR expression (through promoter hypermethylation) and are more likely to enter mitosis and not recover (43,44). The lack of functional mitotic checkpoints likely contributes to the effectiveness of mitosis-targeting chemotherapeutic agents and may explain the different responses of HT29 and WI38 cells to AK301. However, microtubule disruption was observed in both arrested HT29 cells and WI38 interphase cells treated with AK301. This finding suggests that AK301 may target microtubules. This target is further supported by in vitro and in vivo tubulin polymerization studies and by in silico docking of AK301 to tubulin. It remains possible that AK301 interacts with other cellular proteins to achieve its effect of the cell cycle arrest and apoptosis, but all our doseresponse and structure-activity studies point to microtubules as being an important target.
Microtubules are filamentous polymers of the cytoskeleton, composed of repeating ␣/␤-tubulin heterodimers, responsible for determining cell shape, motility, intracellular transport, and cell division (3). Microtubules have long been known as the drivers of chromosome migration and chromosome segregation (45). Microtubules become highly dynamic during mitosis and generate bipolar spindles that capture the sister chromatids and align them at the equatorial plate (46,47). With proper chromosome alignment and cellular signaling, cells enter into anaphase and complete cell division (48). The importance of microtubules in mitosis has made them a fruitful target for cancer therapies. However, it is clear that not all tubulin disruptors are equally useful as cancer therapies. This may be due in part to their influence on apoptosis pathways. Here we show that among microtubule disruptors, AK301 is particularly potent at . AK301-treated HT29 cells were co-treated with TRAIL (20 or 40 ng/ml) or anti-Fas antibody (10 g/ml) and analyzed for caspase-3 activation. Both TRAIL and Fas ligation significantly enhanced caspase-3 activation, despite HT29 cells being more sensitive to Fas ligation (*, p Ͻ 0.0001). C, TNF was titrated onto HCT116 cells in the presence or absence of 500 nM AK301. AK301 significantly increased caspase-3 activity as observed by DEVD-AMC cleavage (#, p Ͻ 0.01), even though HCT116 cells are inherently more sensitive to TNF-induced cell death. All results are reported in triplicate as means Ϯ S.E.
sensitizing cancer cells to TNF and other apoptotic ligands. The interaction of microtubules with apoptosis is not limited to disrupting agents because paclitaxel can also sensitize cancer cells to TNF (49). The mechanism by which apoptotic signaling is enhanced by microtubule targeting agents is not clear, but further study of this effect could improve our understanding of apoptosis regulation and may lead to the generation of more effective microtubule targeting agents. Our present data point to an increase in the expression of TNFR1 on the surface of cancer cells.
Molecular docking was employed to assess the potential of AK301 binding to tubulin dimers. For this analysis, we focused on the microtubule binding sites for colchicine, paclitaxel, and vinblastine. Molecular docking defines energy-optimized ligand orientations formed between the drug and its receptors (50). Molecular docking predictions of AK301 and its derivatives showed relatively high affinity for the colchicine-binding region of tubulin but docked in a different orientation than colchicine. Further analysis of these in silico complexes supported the significance of this binding position; we found that the longer chain ethoxy group of AK301 favored strong hydrogen bond interactions and positioned the chlorophenyl ring in a hydrophobic pocket. In summary, this theoretical structural analysis predicted the affinity of AK301 for tubulin.
We propose that AK301 represents a novel class of mitotic inhibitors capable of inducing mitotic arrest on their own and inducing apoptosis in combination with TNF with high efficiency. How mitotic arrest leads to ligand-dependent cell death is not fully understood. We previously showed that mitotically arrested cells have increased cell surface expression of TNFR1 (18). Increased TNF-TNFR1 interactions at the cell surface (or following TNF internalization) may increase the formation of death-inducing signaling complex and caspase-8 activation (51,52), which has been observed in arrested cells. Interestingly, AK301 was the most potent TNF-sensitizing agent tested in these studies, relative to other well studied microtubule inhibitors (colchicine, nocodazole, and vincristine). How AK301 achieves such a high degree of TNF sensitization is not clear. Based on our tubulin polymerization assay, AK301 reduces the rate of tubulin polymerization, but does not prevent it completely (like colchicine). We speculate that AK301 interferes with tubulin polymerization, but just enough such that the cells can continue to deliver and present TNFR1 and/or Fas on the cell surface. However, it should be noted that AK301 might possibly interact with a microtubule-related target or an upstream target that affects tubulin polymerization. The activity of the AK301 class of compounds, both as effective mitotic inhibitors and as apoptotic ligand-sensitizing agents, suggests that they may be well suited for cancer treatment, particularly when used on cancers with a high inflammatory cell infiltrate or following treatment with an immune stimulant. For basic research applications, this class of compounds should help illu- FIGURE 12. Increased TNFR1 cell surface expression and caspase-8 activation in AK301-treated cells. A, AK301 enhances TNF-induced activation of caspase-8 and caspase-9. Cells were treated with AK301 for 16 -18 h in the presence or absence of TNF. Immunoblot analysis of cell lysates with antibodies against full-length and cleaved caspase-8 and caspase-9 showed cleaved caspase-8 and caspase-9 in cells co-treated with AK301 and TNF. B, AK301 induces increased cell surface expression of TNFR1 on HT29 cells. Immunoblot analysis of an Nonidet P-40 fraction (membrane proteins) and RIPA fraction (total protein/cell lysate) showed an increase in TNFR1 in the membrane fraction post-AK301 treatment. C, increase in cell surface TNFR1 expression by flow cytometry. HT29 cells treated with AK301 and its less active analogs were stained for TNFR1 and then fixed for flow cytometry. The data indicate increased cell surface expression of TNFR1 in AK301-treated cells, but not with the inactive analogs of AK301. Cntl, control. minate how microtubules are employed to regulate apoptosis sensitivity.