To screen compound libraries for molecules inhibiting binding of the outer kinetochore Ndc80 complex (Ndc80, Nuf2, Spc24, Spc25) to MTs, we used a fluorescence microscopy-based in vitro approach (Fig. 1). First, the recombinant human Ndc80 construct, used to crystallize the complex [12], was produced in Escherichia coli using a bicistronic plasmid from which the Nuf2-Spc24 and Ndc80-Spc25 peptides were generated (Fig. 1A). Following their intracellular assembly, the complex was purified from E. coli cell extract based on the GST tag at the N-terminus of Nuf2. The complex was released in solution with PreScission Protease (its recognition sequence was introduced between the GST tag and the NUF2 coding sequence) and was separated from contaminants by gel filtration chromatography (Fig. 1A). The high degree of purity of the preparation was confirmed by SDS-PAGE analysis and coomassie staining, which identified only both peptides (Fig. 1B). Next, the Ndc80 complex was fluorescently labeled with the Alexa Fluor 488 C₅-maleimide (next named the Ndc80⁴⁸⁸ complex) and was separated from unreacted fluorophore by gel filtration chromatography. The final preparation was highly pure as evidenced by gel filtration analysis and detection at 519 nm (emission maximum of the Alexa Fluor 488 dye) and at 280 nm (intrinsic protein fluorescence) (Fig. 1B). Next, the Ndc80⁴⁸⁸ complex (0.1 µmol/l) was incubated with 10,200 compounds (comprising the Chembridge DIVERset and a compound collection from the Univ. Rome) at a starting concentration of 50 µmol/l. Binding (or lack thereof) of the complex to rhodamine-labeled taxol-stabilized MTs (0.1 µmol/l) was scored by wide-field fluorescence microscopy (Fig. 1C; images in Fig. 2A). One compound, named “compound B” (6-furan-2-yl-3-methyl-4-oxo-4,5,6,7,-tetrahydro-1H-indole-2-carboxylic acid tetrahydro-furan-2-ylmethyl ester; DIVERset, Chembridge) was found to be active (Fig. 2A,B). Total internal reflection fluorescence microscopy was used to establish the IC₅₀ value of the compound, which was measured to be ∼20 µmol/l (Fig. 2C).

A compound inhibiting the Ndc80⁴⁸⁸ complex-MT interaction could act either against the complex or against the MTs. To probe the specificity of compound B toward the Ndc80⁴⁸⁸ complex, we examined its ability to inhibit MT binding of recombinant CLIP-170⁴⁸⁸, a MT plus-end tracking protein. We found that compound B prevented CLIP-170⁴⁸⁸ from binding to MTs (Fig. 2D), suggesting that compound B acted at the MT level. Of note, MT poison nocodazole did not prevent the Ndc80⁴⁸⁸ complex from binding to the taxol-stabilized MTs (Fig. 2E), suggesting that compound B interacted in a unique manner with the stabilized MT polymer, and indirectly, with binding of the Ndc80⁴⁸⁸ complex to the MTs.

To study whether compound B affected mitosis, HeLa cells were synchronously released from G1/S (thymidine block) into growth media containing 0.1–50 µmol/l compound B. Time-lapse videomicroscopy showed an accumulation of mitotic cells in the presence of the compound, while the mock-treated cells progressed through mitosis (Fig. 3A, upper and lower panels). At compound B concentrations of 5 µmol/l and above, the cells arrested robustly in metaphase (∼15 h) and then underwent cell death, as diagnosed by cell shrinkage (Fig. 3B). The observed mitotic delay came from mitotic checkpoint activity as confocal immunofluorescence (IF) imaging showed that SAC protein Mad1 accumulated at kinetochores in cells treated with compound B (as with nocodazole, which was used as a positive control, Fig. 3C). The IF analysis further revealed that sister chromatids (DAPI) and kinetochores (CREST serum) were not aligned on the metaphase plate. This phenotype is indicative of chromatids being unable to bind to spindle MTs and/or of spindle defects, as observed with nocodazole (Fig. 3C). To determine whether compound B affected kinetochore-spindle attachment or interfered with spindle integrity, we examined by confocal IF imaging the localization of chromosomes and kinetochores, and the state of the spindle (anti-α-tubulin) in cells synchronously released from a G1/S arrest into medium containing 10 µmol/l of compound B. All cells lacked a mitotic spindle, as with nocodazole (Fig. 3D), supporting the idea that compound B acts at the MT level, probably by inhibiting tubulin assembly. Because drugs that inhibit tubulin polymerization also destabilize MTs [3], we next probed whether compound B destabilized metaphase spindles. We arrested HeLa cells in metaphase using 10 µmol/l of proteosome inhibitor MG132. The cells, all of which contained a mitotic spindle, were then treated with 1% DMSO or 10 µmol/l compound B. IF imaging showed that compound B depolymerized the spindle (Fig. 3D, bottom set of panels). Thus, compound B prevents tubulin assembly and destabilizes spindle MTs in cells.

To probe whether the activity of compound B is reversible or not, we synchronously released G1/S arrested HeLa cells into fresh medium containing 5 µmol/l compound B (nocodazole was used as the positive control). The cells efficiently arrested in metaphase due to absence of a mitotic spindle (Fig. 3E, left column of images in each set). Compound B and nocodazole were then washed out and the cells were released in MG132 containing medium. Within 3 h, all cells had arrested with a mitotic spindle suggesting that our compound does not covalently bind to tubulin, allowing for full reversibility of its intracellular activity (Fig. 3E, right column of images in each set).

During our screen we found MTs to be stable in the presence of compound B (Fig. 2, panel A and panels C to E), suggesting that taxol prevented MT depolymerization by the compound. To test this hypothesis, we treated HeLa cells with 0.5 µmol/l taxol, thereby arresting them in metaphase with stabilized spindles. Following the addition of either 10 µmol/l compound B or 1% DMSO, the spindles did not depolymerize (Fig. 4A). Hence, the effects of taxol are likely dominant over those of compound B since our previous experiments had shown that compound B alone efficiently depolymerized metaphase spindles at 10 µmol/l (Fig. 3, panels C and D). To quantify the taxol-mediated inhibition of MT disassembly by compound B, we measured in vitro the depolymerization of taxol (20 µmol/l)-stabilized MTs following incubation with 50 µmol/l compound B (the screen condition). Nocodazole and maytansine (both at 50 µmol/l) were included as controls. At 50 µmol/l, compound B depolymerized 3% of the taxol-stabilized MT population, while nocodazole and maytansine depolymerized 11% and 31% of the taxol-treated MTs, respectively (Fig. 4B). The limited ability of compound B to depolymerize taxol-stabilized MTs explains why we did not identify compound B as a MT-depolymerizing drug in our screen. This finding further supports our hypothesis that the interaction of compound B with the stabilized MTs must have prevented the Ndc80⁴⁸⁸ complex from establishing contact with the MT surface.

To quantify the antitubulin activity of compound B, we added 1–50 µmol/l compound B to 10 µmol/l αβ-tubulin. Tubulin polymerization in GTP-containing buffer was tracked turbidimetrically at 350 nm [13], [14], with combretastatin A-4 and nocodazole acting as controls. Compound B inhibited tubulin polymerization with an IC₅₀ value of 5.5 µmol/l, whereas combretastatin A-4 and nocodazole were 5-fold more effective (IC₅₀ values of 1.2 µmol/l and 0.96 µmol/l, respectively; Fig. 5A). Because compound B also destabilized spindle MTs in cells (Fig. 3D), we also quantified in vitro the MT depolymerization activity of the compound. For this purpose, MTs were assembled from 10 µmol/l αβ-tubulin; 5 min into the reaction compound B was added in concentrations up to 200 µmol/l. MT depolymerization was tracked as a decrease in turbidity at 350 nm (Fig. 5B). Compound B affected MT stability only at molar ratios exceeding 20∶1, while nocodazole induced MT depolymerization at a drug:tubulin ratio of 1 (Fig. 5B). Thus, compound B inhibited tubulin polymerization at a concentration at least 40-fold lower than that needed to disassemble MTs. In contrast, a 10-fold difference was observed for nocodazole (Fig. 5B versus Fig. 5A). The measured discrepancies in the antitubulin activities of compound B are typical for tubulin destabilizing drugs [3].

Agents that inhibit tubulin polymerization and destabilize MTs include the colchicinoids, combretastatins, benzimidazole carbamates, vinca alkaloids, and maytansinoids [3]. These drugs act primarily at β-tubulin in the αβ-heterodimer, albeit at different sites. Colchicinoids, combretastatins and benzimidazole carbamates (e.g. nocodazole) localize to the colchicine site near the C-terminus of β-tubulin at the αβ-tubulin intradimer interface [15], [16], whereas vinca alkaloids and maytansinoids localize to the N-terminal GTPase domain of β-tubulin [17], [18]. To determine whether compound B (1–50 µmol/l) acts at the colchicine-binding site, we assayed whether it inhibited binding of [³H]colchicine (5 µmol/l) to tubulin (1 µmol/l), in comparison with combretastatin A-4 and nocodazole (Fig. 5C). Compound B significantly inhibited [³H]colchicine binding to tubulin, indicating that it acts at the colchicine site. Its binding affinity, however, was ∼3-fold lower than that of nocodazole (compare values at 5 µmol/l and 20 µmol/l, Fig. 5C).

Compound B (Fig. 2B) represents a novel antitubulin chemotype. Consequently, its overt lack of structural similarity with known colchicine site agents made it difficult to understand which part of compound B was responsible for its antitubulin activity. To solve this problem, we performed a computer-assisted search of commercial libraries comprising ∼3,000,000 compounds for molecules containing either a 6-furan-2-yl-3-methyl-4-oxo-4,5,6,7-tetrahydro-1H-indole structure (encircled in blue, Fig. S1) or the C2 carboxylate-based tetrahydrofurfuryl side group (encircled in red, Fig. S1). We identified 104 compounds containing a tetrahydrofurfuryl group; none were active in the Ndc80⁴⁸⁸ complex-MT binding assay. In contrast, 17 compounds containing the 6-furan-2-yl-3-methyl-4-oxo-4,5,6,7-tetrahydro-1H-indole fragment inhibited to varying extent the binding of the Ndc80⁴⁸⁸ complex to MTs, suggesting that this region harbors the antitubulin activity of compound B. Four of the 17 compounds (named A6, D7, E7, and A8) were ∼20-fold more active than compound B, as measured in the Ndc80⁴⁸⁸ complex-MT binding assay (IC₅₀ values of ∼1 µmol/l; Fig. 6A). Their enhanced activity was also confirmed turbidimetrically: the IC₅₀ values of compounds A6, D7, E7 and A8 for inhibiting tubulin assembly were 2.3, 3.0, 4.0, and 4.2-fold lower, respectively, than that of compound B. The activity of the most potent molecule, A8, was similar to that of combretastatin A-4 and nocodazole (Fig. 6B). The capacity of these four molecules to destabilize MTs in vitro was also superior to that of compound B; the ability of compound A8 to depolymerize MTs was 12-fold higher than that of compound B and was 10% higher than that of nocodazole (Fig. 6C). The high antitubulin activity of compounds D7, A6, A8 and E7 likely results from an improved ability of the compounds to localize to the colchicine site. Indeed, compounds D7, A6, A8 and E7 were 3-fold more active than compound B as inhibitors of [³H]colchicine-tubulin binding (Fig. 6D), concordant with their inhibitory effects on tubulin assembly (Fig. 6B).

Next, we determined whether the elevated antitubulin activities of compounds A6, A8, D7 and E7 (versus compound B), and the discrepancies in activity between the compounds themselves were reflected in their effects on mitosis. G1/S synchronized HeLa cells were released into growth medium containing 100 or 500 nmol/l of these compounds or compound B; the duration of the mitotic delay by the compounds and the percentage of cells undergoing subsequent cell death were scored. Except for A6, the new compounds were also more active than compound B in this cell-based assay (Fig. 6E). Possibly, compound A6, which has a bulky side group similar to that of compound B (Fig. 6A) has a reduced uptake or lower affinity for the colchicine site in cells. The four compounds triggered mitotic delays to an extent that directly reflects their activities in vitro (compare Fig. 6E and Fig. 6, panels B to D). The observation that the compounds caused mitotic delays of varying duration also suggests that their side chains are important for their biological activity and are unlikely to be hydrolyzed in human cells. At 500 nmol/l, the most potent molecules (D7, E7, and A8) elicited a robust mitotic block, followed by cell death (Fig. 6E). Of note, each of our compounds acted in cells with an efficacy exceeding those measured in vitro (compare Fig. 6E and Fig. 5). This discrepancy may be due to increased compound:tubulin ratios in cells. Alternatively, the compounds may be modified intracellularly, resulting in enhanced activity. Because compound B and the four second-generation compounds are 6-furan-2-yl-3-methyl-4-oxo-4,5,6,7-tetrahydro-1H-indole-2-carboxylic acid derivatives, we have named these tubulin inhibitors “furan-metoticas”.

After confirming that our most active compound A8, inhibited spindle formation, depolymerized metaphase spindles, acted in a reversible manner, and triggered a mitotic arrest and death of HeLa cells (Fig. 7), we examined the efficacy of the compound on the growth of the A2780 (ovarian carcinoma) and HCT-116 (colorectal carcinoma) tumor cell lines. RPE, a telomerase-immortalized cell line derived from retinal pigment epithelium, served as the non-cancer control. Antitubulin agents colchicine and nocodazole were used as controls for drug activity. The RPE, A2780, and HCT-116 cells were efficiently killed by compound A8 and the control drugs. Importantly, the efficacy of compound A8 (IC₅₀: 25 nmol/l) was similar to that of the established tubulin inhibitors (Fig. 8).

Recombinant human Ndc80 complex (Ndc80^Bonsai complex) was produced in Escherichia coli using a bicistronic expression plasmid (Fig. 1A, [12]). Following transcription from the TAC promoter/operator by induction with 0.4 mmol/l IPTG (20°C, 15 h), GST-Nuf2-Spc24 and Ndc80-Spc25 polypeptides were generated, which dimerized intracellulary into the GST-Ndc80 complex. The complex was isolated from E. coli cell extract following incubation with glutathione beads. The complex was released from the beads with PreScission Protease (GE Healthcare Life Sciences) and contaminants were removed by gel filtration chromatography (Superdex 200, GE Healthcare Life Sciences). On average 4 mg of Ndc80 complex were obtained from 1 liter of cell culture. Next, 1 mg Ndc80 complex was labeled with 0.1 mg Alexa Fluor 488 C₅-maleimide fluorophore (Invitrogen) at a 1.5-2 stoichiometry. The fluorescent complex (Ndc80⁴⁸⁸ complex) was separated from unreacted fluorophore by gel filtration chromatography (Fig. 1B) and was then used to screen for compounds inhibiting Ndc80⁴⁸⁸ complex-MT binding (Fig. 1C). The Ndc80⁴⁸⁸ complex is fully functional as it is recruited to and acts at kinetochores when injected into prometaphase HeLa cells [12].

MTs were polymerized (30 min, 37°C) from a 1∶10 mixture of rhodamine-labeled:unlabeled bovine brain αβ-tubulin (Cytoskeleton) in BRB-80 buffer (80 mmol/l PIPES pH 6.8, 1 mmol/l MgCl₂, 1 mmol/l EGTA) and were then stabilized with 20 µmol/l taxol. The Ndc80⁴⁸⁸ complex (0.1 µmol/l) was incubated with 100 µmol/l compound (30 min, 24°C) in 96-well polypropylene plates (Eppendorf). The 10,000 compound DIVERset library (Chembridge) and 200 compounds from the Department of Medicinal Chemistry and Technologies, University of Rome La Sapienza were used in the screen. Next, the MTs (100 nmol/l) were added to the compound-Ndc80⁴⁸⁸ complex solution in a 1∶1 v/v ratio. After incubation (10 min, 24°C), the reaction mixtures were transferred to 96-well glass-bottom plates (Greiner Bio-one). Each well was imaged at 519 nm and 590 nm with an IX81 Olympus inverted wide-field microscope (Olympus). Total internal reflection fluorescence imaging was performed with an Olympus IX71 inverted wide-field microscope (Olympus). ImageJ was used for image processing (NIH, USA).

Polymerization of 10 µmol/l αβ-tubulin in BRB-80 buffer was monitored turbidimetrically at 350 nm for 15 min in a temperature-controlled spectrophotometer (37°C; SpectraMax Plus³⁸⁴, Molecular Devices). To probe the effect of a compound on MT stability the molecule was added 5 min after initiation of tubulin assembly. The decrease in protein turbidity (350 nm) resulting from MT depolymerization was followed for 15 min at 37°C [13]. Quantitative comparisons of drug effects on tubulin assembly were also examined in 0.8 M glutamate with a 15 min drug-tubulin preincubation prior to the addition of GTP [14]. Tubulin assembly was monitored for 20 min at 30°C in Beckman DU7400/7500 spectrophotometers, and the compound concentration inhibiting tubulin assembly by 50% was determined.

Tubulin binding of [³H]colchicine was measured by the DEAE-cellulose filter method [20]. Reaction mixtures contained 1.0 µmol/l αβ-tubulin, 1.0 mol/l monosodium glutamate, 0.1 mol/l glucose-1-phosphate, 1.0 mmol/l MgCl₂, 1.0 mmol/l GTP, 0.5 mg/ml BSA, 5% DMSO, 5 µmol/l [³H]colchicine, and inhibitor at 1, 5, 20 or 50 µmol/l. These reaction conditions strongly stabilize the colchicine-binding activity of tubulin [21].

Taxol-stabilized MTs were prepared and diluted as described for the Ndc80⁴⁸⁸ binding experiments. Varying compound concentrations in 10 µl of DMSO were added to 490 µl of diluted MTs at 37°C. Following incubation (30 min, 37°C), 450 µl of each mixture was centrifuged (10 min; 30,000 rpm; 37°C) in a Beckman Optima TLX centrifuge (TLA120.1 rotor). Pellets were dissolved in 25 µl of 8 mol/l urea, and the protein content of 20 µl was compared with that from pellets formed without compound.

Viability analysis in the presence of antitubulin drugs was performed on hTERT RPE-1 (ATCC), HCT-116 (ATCC), and A2780 (EACC) cell lines. Cell lines were authenticated by PCR [22]. Absence of eight commonly encountered mycoplasma species was confirmed by PCR [23]. Experiments were performed with freshly thawed cells, which were seeded in 96-well MICROTEST tissue culture plates (Falcon) at ∼500 cells/well and grown at 37°C. After 6 h, nocodazole, colchicine or compound A8 were added (1–1000 nmol/l). The cells were incubated for 3 d at 37°C and then assayed for viability (CellTiter 96 Assay, Promega).

G1/S synchronization with 2.5 mmol/l thymidine was carried out for 16 h. IF imaging (Leica TCS SP2 confocal microscope) was performed on cells fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 and treated with 4% BSA. The cells were incubated with appropriate antibodies diluted in 4% BSA in PBS. Primary antibodies used in IF imaging were mouse monoclonal antiNdc80 antibody (1∶1000, Genetex Clone 9G3.23), CREST serum (1∶50, Antibodies Incorporation) and mouse monoclonal anti-α-tubulin antibody (1∶2000, Sigma-Aldrich Clone B512). Secondary antibodies used were Alexa Fluor 488 (1∶150), Cy3 (1∶400) and Cy5 (1∶50) (Jackson Immunoresearch).