Rat brain tubulin was purified from microtubule protein previously extracted from frozen rat brains, as described previously (Sackett et al., 1991; Montecinos-Franjola et al., 2019). Tubulin from chicken brain and from chicken red blood cells was purified from frozen whole brains and from washed red blood cells (#33131-1, Pel-Freez Biologicals, Rogers, AZ) (https://www.pelfreez-bio.com/products/animal-serum-plasma-complement-and-ancillary-products/whole-blood-and-red-blood-cells/), respectively, as described previously (Sackett, 1995; Montecinos-Franjola et al., 2019). Bovine brain tubulin protein was from Cytoskeleton, Inc., Boulder, CO (#HTS02-A) or from Sigma-Aldrich, St. Louis, MO (#T4925). Purified proteins were stored in PM buffer (0.1M Pipes-KOH, pH 7.0, 1 mM MgCl₂) at −80°C. The estimation of protein concentration was made using the Bradford assay (Bio-Rad) with BSA as the calibration standard (#23209, ThermoFisher Scientific, Waltham, MA).

MDL-(E)-1-(2,5-dimethoxyphenyl)-3-[4-(dimethylamino)phenyl]-2-methylprop-2-en-1-one (MDL) was not commercially available and therefore was synthesized according to the literature (Ducki et al., 1998) with the following minor modifications. In the second step, the oxidation of 1-(2,5-dimethoxyphenyl)-1-propanol using pyridinium dichromate, one equivalent of acetic anhydride was used per equivalent of starting material. The product from this reaction was stirred in diethyl ether with 2 g of Florisil per mmol of starting material and then filtered to remove excess pyridinium dichromate. The structure of the product was confirmed by 1H NMR. An extinction coefficient was measured in DMSO (ε_{385 nm} = 2.69 × 10⁴ M⁻¹cm⁻¹).

Drugs and test compounds were obtained from multiple sources listed in Supplementary Table S1. Concentrated stock solutions were prepared in water, in PM buffer or in DMSO as indicated in the same table, and stored at −20°C. The concentration was determined spectrophotometrically for those compounds with known extinction coefficients.

Tubulin was polymerized at a concentration of 2 µM in PM buffer (0.1 M Pipes pH 7, 1 mM MgCl₂) + 0.1 mM GTP in a volume of 60 µL. The reaction mixture was transferred to a 10-mm pathlength microcuvette (Hellma United States Inc., Plainview, NY) and placed in the cuvette holder of a SpectraMax M2 Multimode Microplate Reader (Molecular Devices, San Jose, CA) equipped with temperature control set at 37°C for these experiments. Drugs at 20 µM were added directly to the polymerization mixture and incubated for 15 min previous to starting the recordings. Tubulin polymerization kinetics was followed by recording the absorbance at 350 nm. After recording the baseline, tubulin polymerization was induced by the addition of 2 µM paclitaxel (taxol). The final concentration of DMSO in the mixtures was 4%–5% v/v.

The fluorescence emission of MDL was recorded at 550 nm in the SpectraMax M2 Multimode Microplate Reader with excitation at 400 nm. Tubulin samples at 1 µM were prepared in PM buffer in 60 µL volume, incubated for 30 min, and then transferred to a 10-mm pathlength microcuvette which later was placed in the cuvette holder of the SpectraMax M2 set at 25°C. The kinetics of MDL binding and its displacement by competitor drugs was detected by recording the fluorescence emission at 550 nm every 5 s. MDL was used at 10 µM and podophyllotoxin at 100 µM. For fluorescence spectra recordings, the drugs were added at 50 µM final prior to addition of MDL and the DMSO was adjusted to 5% v/v.

The equilibrium dissociation constants for the binding of MDL to tubulin were determined from the concentration-dependent changes in the fluorescence emission at 550 nm upon serial dilution of the MDL. The starting mix contained 1 µM tubulin and 10 µM MDL all in PM buffer (starting mix). The dilution mix contained 1 µM tubulin in PM buffer and was prepared in a larger volume (dilution mix). The concentration of DMSO was adjusted to 5% v/v in both cases. The dilution series was prepared by mixing the “starting mix” and the “dilution mix” at a 1:3 ratio resulting in a decrease of the MDL concentration to 1/3 of the original, or 3.3 µM, while keeping the tubulin concentration constant at 1 µM. This new sample was then mixed at 1:3 ratio with the “dilution mix” for obtaining the next sample in the dilution series at 1.1 µM MDL. Subsequent samples in the dilution series were prepared in the same fashion until reaching a low concentration <1 nM MDL. Next, 60 µL of the dilution series were transferred to a 96-well black half-area flat bottom fluorescence microplates (#3642 Corning, NY) including the blanks. The loaded plates were incubated for at least 30 min and then placed in the SpectraMax M2 Multimode Microplate Reader for recording the fluorescence signals with temperature control set at 25°C.

For competition experiments with colchicine-site drugs, the samples were prepared in a similar fashion with some changes. For full dilution series experiments (e.g., podophyllotoxin and benzimidazoles), the starting mix contained 1 µM tubulin and 5 µM MDL all in PM buffer (starting mix). The dilution mix contained 1 µM tubulin, 5 µM MDL and 100 µM drugs, all in PM buffer, and was prepared in a larger volume (dilution mix). The concentration of DMSO was adjusted to 8% v/v in both cases to promote the solubility of the ligands. The dilution series was prepared by mixing equal amounts (1:1) of each mix resulting in a 50% decrease of the initial drug concentration at each step of the dilutions series, while keeping constant the concentrations of tubulin at 1 µM and of MDL at 5 µM. For the single concentration drug screening experiments, the samples contained 1 µM tubulin (unless otherwise specified), 5 µM MDL and the drugs were added at a final concentration of 50 µM directly to the mixtures. Next, 60 µL of the samples were transferred to 96-well microplates. For competition experiments, the loaded plates were incubated for at least 1 h due to the slower kinetics observed for MDL release from the colchicine binding site after addition of the competing drugs. The MDL fluorescence was recorded in the SpectraMax M2 Mulitmode Microplate Reader with temperature control set at 25°C.

The dissociation constant for the MDL-tubulin interaction was determined by fluorescence intensity measurements as previously described for MDL and other colchicine binding-site drugs (Peyrot et al., 1992; Sharma et al., 2010). The raw fluorescence signal was blank-subtracted and plotted as a function of the drug concentration. The data was analyzed by non-linear regression using the following one-component binding model that relates the fluorescence intensity observed (I_obs) with the apparent equilibrium dissociation constant K_d,MDL as: K d , MDL =   [ Tub ] · [ MDL ] [ Tub − MDL ] (1) I obs =   I free + ( [ MDL ] K d , MDL · I bound ) 1 + [ MDL ] K d , MDL (2) Where (Tub) and (MDL) are the concentrations of free tubulin and MDL at equilibrium, I_free and I_bound are fluorescence signals of the free MDL in solution and of MDL bound to tubulin under saturating concentrations, respectively. Weights were assigned to each data point based on the reciprocal of the standard deviation of each data point (averages of three to four measurements). For comparing plots of the various tubulins, or the different drugs, the data was converted to fraction bound with the following relationship: f b =   I obs − I free I bound − I free (3)

In the competition experiments, for instance with podophyllotoxin, the value of IC₅₀ was determined graphically from plots of f_b vs. (podophyllotoxin) and the value of the apparent dissociation constant, K_d,podo was calculated with the following equation (Yung-Chi and Prusoff, 1973; Craig, 1993): K d , podo =   IC 50 1 + [ MDL ] K d , MDL (4)

For single point competition experiments, the F/F_max ratios were calculated from the fluorescence intensity values recorded in the presence of the competitor drug, F, and the fluorescence intensity of MDL in the absence of competitor, F_max. The percentage inhibition was calculated from the F/Fmax ratios: % inhibition = ( 1 − F F m a x ) × 100 (5)

All of the competition titrations yielded curves consistent with simple single-site binding, and therefore we make the assumption that all of the tested compounds do as well. With this assumption, single point data (50 µM test compound) was fit with a single-site binding isotherm by non-linear regression. Then the mid-point of that curve (IC_50%) was used to calculate the K_d(app) using Eq. 4, as above. The curve-fitting procedure did not converge for % inhibition below 1%, therefore the apparent dissociation constant could not be determined in those cases (nd). For further details and examples, see Supplementary Material.

Competition binding assays are often used as a method to quantify binding to a target protein. Such assays can be performed using a tritiated ligand that is known to bind to the target, but spectroscopic methods are more convenient. Colchicine is often used for competition binding studies on tubulin, but it does not bind well to CeTb so it cannot be used for the purposes of this study. Another molecule that is a candidate for a competition binding assay is known in the literature as MDL-27048 (Figure 1). The fluorescence of MDL-27048 (hereafter referred to as MDL) is very weak when free in solution and greatly enhanced upon binding to the colchicine site of mammalian brain tubulin (BTb) (Peyrot et al., 1992). However, it was unknown whether MDL would interact with CeTb. Colchicine-site drugs inhibit the polymerization of purified tubulin into microtubules. In assays using tubulin from mammalian brain, MDL has been found to inhibit polymerization (Peyrot et al., 1989). Therefore, we polymerized CeTb in the presence of MDL to look for inhibition. We found that MDL does inhibit the polymerization of CeTb (Figure 2; quantitation is in Table 1), indicating that it does interact with CeTb.

It is reasonable to suppose that MDL binds to the colchicine site on CeTb, as it does to BTb, resulting in inhibition of polymerization of both tubulins. To be useful in a competition binding assay, it must also exhibit a spectroscopic change upon binding. We therefore looked for a change in the fluorescence of MDL in the presence of tubulin, including CeTb. The emission spectrum showed a large enhancement of fluorescence upon binding to CeTb, as well as to BTb (Figure 3). Fluorescence intensity increases rapidly upon addition of MDL to BTb or CeTb, and is then displaced by addition of 10-fold excess of the colchicine site ligand podophyllotoxin (structure in Figure 1). Displacement kinetics differ between BTb and CeTb, indicating a slower off-rate for MDL from CeTb compared to BTb. Addition of the benzimidazole nocodazole also displaces MDL, confirming the usefulness of this assay for binding of diverse compounds.

This emission enhancement allowed us to determine the binding affinity of MDL for CeTb, compared to brain tubulin from chicken brain (CBTb), rat brain (RBTb), and bovine brain (BBTb) (Figures 4A,B). Tubulins were titrated with MDL, and the concentration dependence of the emission signal was used to determine the affinity. The resulting binding isotherms are shown in Figure 4C and the fitted K_d values for binding are presented in Figure 4C and Table 2. It is notable that all the brain tubulins, including CBTb, yield very similar Kd for MDL binding to the protein, while MDL binding to CeTb shows affinity ∼5–fold weaker than to brain tubulins.

As a test of utility of MDL in a competition assay for ligand binding to the colchicine site, the binding of the reversible colchicine site ligand podophyllotoxin was chosen. It was shown in Figure 2 that excess podophyllotoxin could displace MDL from both BTb and CeTb, so we used titration of tubulin- MDL with increasing concentrations of podophyllotoxin to measure displacement by loss of fluorescence intensity. The resulting data for BTb and for CeTb are shown in Figures 5A,B, respectively. Displacement as a function of podophyllotoxin concentration is shown in Figure 5C, and the derived K_d for binding of podophyllotoxin to BTb and to CeTb (calculated as described in Materials and Methods) are given in the inset to Figure 5C and in Table 3. The K_d for podophyllotoxin binding to BTb was found to be ∼0.5 μM, essentially the same as the literature value of 0.55 µM found in rat brain tubulin using tritiated podophyllotoxin (Cortese et al., 1977). The K_d for CeTb, ∼6 µM indicates about a 12-fold weaker binding of podophyllotoxin for CeTb compared to BTb.

A similar analysis was performed using the compound cis-Combretastatin A4 (structure in Figure 1), This two-ring analog of colchicine binds with considerable affinity as well as specificity (the trans-form is notably less potent) and has been the subject of much preclinical development (Cushman et al., 1991) [reviewed in (Sackett, 1993)]. Titration of CeTb- and BTb-bound MDL with combretastatin A4 yielded a K_d of 5.7 ± 0.32 µM for CeTb, while the K_d for BBTb was found to be 0.13 ± 0.011 µM (Supplementary Figure S5) The K_d for BBTb is similar to the literature value of 0.12 µM (Lin et al., 1989). Results are included in Table 3, and indicate a ∼44-fold weaker binding to CeTb compared to BTb.

Since fluorescence titration and competition assays with MDL and two standard compounds yielded binding values consistent with literature values with BTb, we accept the values obtained for CeTb as well and turned our attention to other compounds using this assay. The first group of compounds we examined were the benzimidazoles. These compounds have a history as antihelmintics that target parasite tubulin over mammalian host tubulin, but are causing renewed interest as potential repurposed agents in human oncology (Son et al., 2020). Five compounds were chosen for a full competition titration against CeTb- and BTb-bound MDL. The titration curves and best fits to the data are shown in Figure 6, and the obtained K_d values are presented in Table 3. The data show that nocodazole yields a Kd of about 1 µM to BTb, similar to the published value of 2.5 µM (Head et al., 1985). Mebendazole binding to BTb is similar, and both compounds have similar Kd (within a factor of two) with either tubulin. Other compounds (albendazole, oxibendazole, carbendazim) show moderate binding affinity with CeTb, but no detectable binding to BTb.

The dissociation constants obtained for selected compounds are combined in Table 4 for both CeTb and BTb. The last column shows the ratio of the drug’s K_d for binding to CeTb compared to that for BTb. A ratio >1 (i.e., K_d for CeTb > K_d for BTb) indicates weaker binding to CeTb than to BTb. It is clear that the relative affinity for the two tubulins shows considerable variation, with the greatest differential in the first three “standard” colchicine site compounds found with Combretastatin A4, the strongest binding ligand for BTb, showing 44-fold lower affinity for CeTb. In the benzimidazoles, however, the range of K_d ratios is even broader, since two of the compounds do not show binding to BTb at all under these assay conditions.

Next, we used this method to screen a single concentration of a series of molecules. Many colchicine-site ligands are known, so these were an excellent starting point to look for drugs that have activity in CeTb. A series of known colchicine site drugs was obtained and screened for binding to CeTb. Some compounds are structurally close to colchicine, others not so similar, including benzimidazoles. The molecules studied above, podophyllotoxin and combretastatin, were included in the screen. The results for these two drugs were consistent with the full binding curves obtained above, so we felt that valuable information about the interaction of the various drugs with CeTb (and BTb) could be obtained from a single concentration reading. For this purpose, we used a relatively high concentration of test compounds in order to probe weaker as well as stronger binding.

The results of the single-point screen are shown in Tables 5–8. The test compounds are divided into the following groupings: three-ring colchicine analogs, two-ring colchicine analogs, a group of diverse colchicine-site compounds which are structurally unrelated to the colchicine groups, and the benzimidazoles. Information about the compounds is given in Supplementary Table S1 and structures in Supplementary Figures S1–S4.

A comment about this assay is in order. The assay measures the inhibition of MDL binding to CeTb and BTb by a single concentration of test compound. Since the two tubulins have 5-fold difference in K_d for MDL, a reduction in MDL binding by an identical fraction will imply different binding strengths in the two tubulins. Since MDL binding to CeTb is weaker than to BTb, it will be more easily competed away. Thus, with fixed tubulin and MDL concentrations of 1 and 5 μM, respectively, a competitor/compound that caused a 50% reduction in MDL binding to BTb at 50 μM, would imply a Ki of ∼4.5 µM, while a 50% reduction in MDL binding to CeTb by 50 µM competitor would imply a K_i of ∼17 µM. Since all of the drugs analyzed here by titration (Figures 4–6) yielded curves that indicate simple single-site binding, we analyzed the single concentration data using a single-site binding isotherm by non-linear regression. As with the titrations, this curve was used to obtain the IC_50% which served to calculate an apparent K_d (K_d(app)) using Eq. 4 (for more details see Section 2 and Supplementary Figure S6).

The drugs in the colchicine series (Table 5) displayed weak binding to CeTb, and in general notably lower binding to CeTb than to BTb. Isocolchicine does not bind to either tubulin, demonstrating that the stereospecific requirements for binding to CeTb mimic those of BTb. Interestingly, steganacin inhibits MDL binding to CeTb to the same extent (%) as to BTb (though note that this still indicates a lower affinity/higher K_d(app) to CeTb than to BTb). The compounds that show the strongest binding to CeTb are the same ones that bind strongest to BTb: allocolchicine methyl ketone and allocolchicine ethyl ketone. Both compounds bind to both tubulins more strongly than does the parent compound, allocolchicine, or colchicine itself. While both compounds significantly inhibited MDL binding to both tubulins, the binding to CeTb is weaker than to BTb. By comparison to podophyllotoxin, the K_d(app) for binding of both compounds to both tubulins can be estimated to be lower by a factor of about two than that obtained by titration with podophyllotoxin. This shows that small modifications to the structure of a colchicine-site ligand can improve binding to both tubulins. Of interest here is a rational drug design study by Paré et al. (2020) who showed that some small changes to colchicine yielded a lead compound with increased affinity for βVI tubulin and increased bioactivity towards neutrophils.

Data from the two-ring colchicine analogs (Table 6) demonstrate the greater relative importance of the cis configuration of the two rings in CeTb compared to BTb. This can be seen by comparing combretastatin-A2 and -A4 (cis), both of which have an unsaturated bridge holding the two rings in cis, with combretastatin-A4 (trans) and dihydrocombretastatin-A4, in which the rings are locked in trans, or are freely rotating, respectively. The potency of the two cis compounds as competitors approaches that of podophyllotoxin, and is only slightly less than the two allocolchicine analogs in Table 5. The lack of activity of trimethoxy resveratrol (trans) is consistent with this observation.

The compounds of diverse structure in Table 7 yielded only two compounds with significant binding to CeTb. Indanocine shows similar inhibition of MDL binding to BTb and to CeTb, comparable to steganacin in Table 5, consistent with a ca. 4.5-fold difference in K_d(app). D-64131 shows the highest inhibitory potency to both tubulins of all the compounds in this table, while still showing an approximately 10-fold ratio in K_d(app). Binding potency is similar to podophyllotoxin (for BTb) and slightly less so for CeTb. The largest differential between the tubulins appears to be with lexibulin, which is somewhat less potent than podophyllotoxin with BTb but nearly inactive with CeTb.

Benzimidazole compounds (Table 8) yield several patterns of comparative inhibition. Nocodazole and mebendazole are fairly potent inhibitors of MDL binding to both tubulins, comparable to podophyllotoxin. Both compounds yielded Kd (app) values within a factor of two of the Kd values obtained from titrations, with both tubulins. Three compounds show moderate inhibition of MDL binding to both tubulins, but slightly higher % inhibition to MDL-CeTb than to MDL-BTb: fenbendazole, flubendazole, and benomyl [Kd (app) values are within a factor of two for the two tubulins]. A final group of five shows no or very low inhibition of MDL binding to BTb (<10%), but moderate to significant inhibition of MDL-CeTb: thiabendazole, carbendazim, albendazole, ricobendazole, and oxibendazole. We were somewhat surprised at the lack of inhibition of MDL binding by these compounds, especially albendazole, which has shown activity in mammalian cells that has prompted discussion of repurposing this compound for cancer therapy (Nath et al., 2020; Will Castro et al., 2021). We were unable to find a published study of direct binding of albendazole or these other benzimidazoles to BTb, and therefore we checked albendazole for direct inhibitory activity against polymerization of BTb. We observed little or no inhibition of polymerization of BTb in the albendazole concentration range that we have been studying (Supplementary Figure S7), consistent with the lack of inhibition of MDL binding that we observed in the titration and single-point assays.