Taccalonolide C-6 Analogues, Including Paclitaxel Hybrids, Demonstrate Improved Microtubule Polymerizing Activities

Abstract

The C-22,23-epoxy taccalonolides are microtubule stabilizers that bind covalently to β-tubulin with a high degree of specificity. We semi-synthesized and performed biochemical and cellular evaluations on twenty taccalonolide analogues designed to improve target engagement. Most notably, modification of C-6 on the taccalonolide backbone with the C-13 N-acyl-β-phenylisoserine side chain of paclitaxel provided compounds with 10-fold improved potency for biochemical tubulin polymerization as compared to the unmodified epoxy taccalonolide AJ. Covalent docking demonstrated the C-13 paclitaxel side chain occupied a binding pocket adjacent to the core taccalonolide pocket near the M-loop of β-tubulin. Although paclitaxel-taccalonolide hybrids demonstrated improved in vitro biochemical potency, they retained features of the taccalonolide chemotype, including a lag in tubulin polymerization and high degree of cellular persistence after drug washout associated with covalent binding. Together, these data demonstrate C-6 modifications can improve the target engagement of this covalent class of microtubule drugs without substantively changing their mechanism of action.

Graphical Abstract

Microtubule stabilizers are highly effective against many solid tumors, including breast and ovarian cancers, where they are often used as first-line therapy. However, there is a lack of diversity in the structure and mechanism of action of clinically approved agents of this class, which are all taxane-site binding agents that are susceptible to common drug resistance mechanisms.¹ The taccalonolide microtubule stabilizers are a class of natural compounds that can circumvent multiple mechanisms of taxane-associated resistance due to their irreversible target engagement.^2–4 A combination of methodologies have recently been employed to fully map the taccalonolide pharmacophore. The generation of a taccalonolide AJ-tubulin co-crystal structure,⁵ subsequent molecular modeling,⁶ and cellular-based binding studies⁷ have demonstrated that the C-22-C-23 epoxide of the potent taccalonolides AF (1) and AJ (2)⁸ forms a highly specific covalent interaction with the Asp226 on β-tubulin. Taccalonolide AF has demonstrated in vivo antitumor efficacy, albeit with a short serum half-life and narrow therapeutic window with systemic injection.^{3, 9} Other epoxy-taccalonolides, including 2, demonstrate potent and highly persistent antitumor efficacy when administered by intratumoral injection, which underscores the exciting potential of this compound class and the need to further explore its unique interaction with tubulin.

Of particular interest was our recent unanticipated finding that the introduction of a fluorescent probe at C-6 of the taccalonolide backbone could engage a binding pocket adjacent to the core taccalonolide binding pocket to enhance microtubule-stabilizing activity as compared to the unmodified drug.⁷ While this fluorescent taccalonolide probe provided a valuable tool compound to elucidate binding specificity, it did not offer improved cellular potency and is not amenable to further development as a potential therapeutic. Previous studies have demonstrated that incorporation of the paclitaxel side chain onto the backbone of other bioactive compounds, including the microtubule stabilizer discodermolide, can alter and even improve their cellular properties.^10–14 In the current study, semi-synthesis was performed on the natural product taccalonolide skeleton to generate a series of 20 novel taccalonolide analogues designed to enhance the biochemical and cellular potency of this compound class by improving target engagement. The resulting structure-activity relationships of this study have further characterized the optimal taccalonolide binding pocket on tubulin and refine the modeling of the functional pharmacophore for this unique class of microtubule stabilizing drugs that will be critical in the design of a therapeutic lead compound of this class.

RESULTS AND DISCUSSION

Previous efforts to generate semi-synthetic taccalonolide analogues by direct esterification of the C-7 and C-15 hydroxyl groups have led to unsatisfactory biological effects, partially due to the instability of the resultant esters.¹⁵ In order to overcome this challenge, an efficient reductive amination reaction was optimized to generate the C-6 amino analogue, 6-NH₂-tacca, from the commercially available taccalonolide B (2) (Scheme 1). Further amidation of the C-6 amino group of 6-NH₂-tacca has enabled the synthesis of stable taccalonolide-based fluorogenic probes that retain binding specificity and cellular potency.⁷ Thus, the C-6 amino group provides a useful “handle” for generating stable amide and sulfonamide derivative to further optimize the binding affinity of the C-22,23-epoxy taccalonolides.

Scheme 1. Synthesis of Amide-taccas (3-13) and SA-taccas (14-22).^a.

^a Reagents and conditions: (a) NH₄OAc, NaCNBH₃, MeOH, 35°C, overnight; (b) HATU, DIPEA, EtOAc, 35°C, overnight; (c) MgO, DIPEA, MeCN, rt, 2 d; (d) DMDO, acetone, DCM, −20°C, overnight

A panel of 17 epoxy-taccalonolide C-6 derivatives containing either an amide or sulfomide linker to a benzene moiety with substitutions varying in size, electronegativity, and position were generated as compounds 3 – 10 and 14 – 22 (Figure 1 and Supporting Information). Each of these compounds demonstrated antiproliferative potency against HeLa cells ranging from 1.2 – 109 nM (Table 1). The most potent compound was 14, which was significantly more potent than the unmodified taccalonolide AJ (2), with a GI₅₀ value of 1.2 nM. Other compounds with slightly improved cellular potency included 18 and 19 with GI₅₀ values of 3.7 and 3.1 nM, respectively. The majority of compounds produced fell within a three-fold range of potency as compared to the unmodified taccalonolide, with the exception of 5 that had a GI₅₀ value of 109 nM (Table 1). This demonstrated that modifications at the C-6 position of the taccalonolide backbone are relatively well-tolerated and that some modifications are able to slightly improve the antiproliferative potency of this compound class.

Figure 1.

Structures of compounds 1 – 22.

Table 1.

Cellular Antiproliferative Potency (GI₅₀) and Rate of Biochemical Tubulin Polymerizing Activity for 2 – 22. Significant differences compared to 2 as determined by one-way ANOVA with Dunnett’s post hoc test are shown. Log transformed values were utilized for statistical GI₅₀ comparisons, *p < 0.05.

Compound	GI50 (nM) ± SEM	tubulin polymerization (min) ± SEM
-	-	12.7 ± 1.8*
2	5.0 ± 0.2	9.7 ± 0.5
3	9.4 ± 0.6*	5.6 ± 1.0*
4	4.9 ± 0.4	8.3 ± 1.4
5	109 ± 6*	7.8 ± 1.3
6	15 ± 3*	7.5 ± 0.6
7	9.3 ± 1.4*	8.6 ± 2.0
8	11.5 ± 0.9*	9.0 ± 2.4
9	11 ± 2*	8.3 ± 0.8
10	7.4 ± 0.7	7.6 ± 0.3
11	259 ± 17*	9.2 ± 2.4
12	134 ± 15*	6.4 ± 1.0
13	149 ± 16*	2.5 ± 0.5*
14	1.2 ± 0.1*	5.6 ± 0.6*
15	6.9 ± 0.2*	5.9 ± 0.1*
16	5.3 ± 0.2	5.7 ± 0.5*
17	10.6 ± 0.3*	7.1 ± 0.6
18	3.7 ± 0.1	5.6 ± 0.1*
19	3.1 ± 0.1	4.8 ± 0.5*
20	8.1 ± 0.5	6.3 ± 2.1
21	7.2 ± 0.6	8.3 ± 1.1
22	4.9 ± 0.3	4.3 ± 0.1*
Paclitaxel	3.2 ± 0.5	1.1 ± 0.1*

Due to the tolerability of modifications at C-6 of the taccalonolide backbone and the proximity of taccalonolide binding to the canonical taxane binding site on β-tubulin, additional modifications at C-6 were made to introduce the C-13 N-acyl-β-phenylisoserine side chain of paclitaxel, which is a major determinant of activity of this drug¹⁶ (Figure 1 and Supporting Information; compounds 11 – 13). While each of these three compounds with differing linkers between the taccalonolide core and paclitaxel side chain retained antiproliferative potency, they were each significantly less potent than the simplified C-6 modifications described above, having GI₅₀ values ranging from 134 – 259 nM (Table 1).

While it was encouraging that each of the C-6-modified taccalonolides, including the taccalonolide/paclitaxel fusion compounds, retained cellular antiproliferative efficacy in the nanomolar range, additional studies were performed to probe the direct effects of these compounds on microtubule polymerization. Taccalonolide AJ (2) promotes a concentration-dependent enhancement of tubulin polymerization that can be quantified by interpolating the inflection point in the microtubule growth curve using a Gompertz-based model. It was found that 1 μM of taccalonolide AJ (2) increased the rate of purified tubulin polymerization from 12.7 min to 9.7 min (Table 1). Surprisingly, each of the 20 C-6-modified compounds at this concentration demonstrated an increased rate of tubulin polymerization as compared to 2, with values ranging from 2.5 min for 13 to 9.2 min for 11 (Table 1). In particular, it was striking that one of the taccalonolide-paclitaxel conjugates, 13, which was one of the least potent compounds in cells with a GI₅₀ of 149 nM, had the greatest impact on the rate of polymerization of purified tubulin. This phenotype was somewhat reminiscent of the impact of adding the C-13 side chain of paclitaxel to rigid borneol esters, which demonstrated improved microtubule stabilizing activity as compared to paclitaxel, albeit with reduced cytotoxic potency.¹⁷ Compound 13 was determined to be over ten-fold more potent than taccalonolide AJ in its ability to polymerize purified tubulin such that a concentration of 1 μM of this analogue demonstrated an increased rate and extent of tubulin polymerization as compared to 10 μM taccalonolide AJ with rates of 2.5 min and 4.5 min, respectively (Table 1, Figure 2a). Furthermore, concentrations as low as 0.1 μM 13 promoted polymerization to a similar extent as 1 μM taccalonolide AJ (Figure 2a).

Figure 2.

(a) Comparison of tubulin polymerization between taccalonolide AJ (2) and the most potent analogue in this assay, 13. (b) Comparison of tubulin polymerization between paclitaxel and taccalonolides 2, 12, and 13 each at 20 μM, which is equimolar with tubulin. (c,d) Comparison of a model of β-tubulin (PDB ID: 5EZY) docked to taccalonolide derivatives 13 (c) and 11 (d). The primary taccalonolide binding site (pocket A), additional site occupied by the C-6 modifications (pocket B), and M-loop are indicated. The selected H-bonds and pi-pi stacking are displayed as yellow and azure dashed lines, respectively.

To provide a rationale for how the taccalonolide/paclitaxel hybrid 13 could improve the biochemical microtubule stabilizing activity as compared to the taccalonolide backbone alone, its binding was interrogated through covalent modeling studies. Previous covalent docking studies on fluorescent taccalonolide probes using Covdock recognized an additional binding pocket adjacent to the taccalonolides that afforded additional interactions of the C-6 fluorescein moiety with the M-loop of β-tubulin.⁷ In the present study, the same covalent docking method was applied to compare the binding modes of 13 and 11 with β-tubulin. Optimized structures of 13 and 11 were docked into the optimized 5EZY structure. The representative lowest-energy pose obtained from each docking experiment was displayed (Figure 2c,d). The taccalonolide core structure of each compound was positioned into the taccalonolide binding pocket (based on the 5EZY crystal structure of 2). Interestingly, the paclitaxel side chain on 13 was immersed into the adjacent binding pocket close to the M-loop, affording strong interactions with β-tubulin residues via additional hydrophobic interactions, H-bonds, and salt bridges (Figure 2c). In contrast, the paclitaxel moiety of 11 was only partially (ring C) and loosely inserted into the same binding pocket with the major part of the moiety (C-1”˗C-3” and ring B) hanging on the surface of the binding pocket (Figure 2d). An additional H-bond was predicted between R278 and C=O-4” affording stabilization of the loose binding mode. The overall docking results suggested that the enhanced ability of 13 to promote microtubule stabilization in biochemical assays could be attributed to improved binding affinity to β-tubulin afforded by these additional contacts. Although more comprehensive computational and experimental analyses would be required to provide a more complete understanding of the binding modes for these taccalonolide derivatives, the current binding analysis suggests that 13 may provide elevated binding affinity with β-tubulin via strong engagement of the binding pocket adjacent to the M-loop, while the binding affinity of 11 with β-tubulin was only slightly enhanced via a loose and less-favored binding mode. Although improved biochemical potency and modeling support that the incorporation of the paclitaxel side chain to the taccalonolide backbone improves target engagement, the taccalonolide lag in tubulin polymerization strongly suggests that this additional interaction near the M-loop of β-tubulin is not sufficient to promote taxane-induced M-loop stabilization, which is associated with almost immediate polymerization (Figure 2b).^{3, 18}

While 12 was not as potent as 13 with regard to its ability to polymerize purified tubulin, it was still slightly more effective in this biochemical assay than taccalonolide AJ in spite of its reduced cellular potency, demonstrating a non-linear relationship between the biochemical and cellular activities of these taccalonolide derivatives. Each of the 20 taccalonolide C-6-modified compounds generated in this study was represented graphically with regard to its relative cellular antiproliferative potency and effects on the rate of tubulin polymerization to generate a comprehensive view of their respective biological effects as compared to taccalonolide AJ (2) (Figure 3a). Four compounds were identified that represented the extremes of this compound class: 2 was one of the most potent of the compounds in cells but had the lowest potency in the biochemical tubulin polymerization assay; 13 was the most potent biochemically but had relatively low cellular potency; 14 was significantly more potent in both measures; and 11 was the one of the least potent compound in both readouts.

Figure 3.

(a) Correlation between antiproliferative potency and effects on the rate of biochemical tubulin polymerization for all 20 analogues generated in this study and taccalonolide AJ (2). The four compounds (2, 11, 13, 14) that represent the extremes of these activities are highlighted. (b) Cellular persistence of taccalonolide analogues. HeLa cells were treated with the indicated concentration of each compound (2, 11, 13, 14) that completely inhibited proliferation (left bars) or caused 75% cytotoxicity (right bars) in the SRB assay as compared to vehicle control (−). After 4 h of drug treatment, medium was removed and replaced and colonies that formed over the subsequent 8 days were quantified. (c) Comparison of the cellular persistence of 2 to paclitaxel in HeLa cells treated with 10 nM of each compound or vehicle control for 8 h prior to drug washout. (d) Immunofluorescence imaging of cellular microtubules in HCC1937 cells after a 4 h treatment with taccalonolide analogs at concentrations that caused 75% cytotoxicity in the SRB assay. Images (left to right): vehicle, 2 (20 nM), 14 (10 nM), 11 (1400 nM) and 13 (2700 nM).

We interrogated whether these four compounds demonstrated any differences in their cellular irreversibility as measured by their effects on long-term colony formation after an acute, 4-h exposure, which is a hallmark of the taccalonolides as compared to other classes of microtubule stabilizers.¹⁹ To compare this mechanistic readout among compounds with differing cellular potency, these studies were performed at two concentrations based on their relative potency in the 48 h SRB assay; one that inhibited net growth over the treatment period and a higher concentration that promoted 75% cytotoxicity as compared to the time of drug addition. For all four compounds, a 4-h treatment with the concentration that caused total growth inhibition in the 48-h cytotoxicity assay was not sufficient to inhibit long-term colony formation (Figure 3b). In contrast, a 4-h treatment with concentrations of these each of these compounds that caused 75% cytotoxicity in the 48-h assay was sufficient to completely abrogate the formation of colonies even after the compounds were removed from the medium (Figure 3b). As an additional control, we also found that 2 had a higher degree of cellular persistence as compared to paclitaxel (Figure 3c), consistent with previous reports.¹⁹ Importantly, parallel experiments demonstrated that the viability of these cells was not compromised within the 4-h drug incubation period. This long-term cellular persistence after acute drug treatment is a hallmark of compounds that promote irreversible target engagement and demonstrates that each of the taccalonolide analogues evaluated retain this property.

It was further demonstrated that each of these four taccalonolides with diverse cellular and biochemical potencies promoted cellular microtubule bundling within 4 h at equivalent concentrations (Figure 3d). Additionally, we prepared sufficient 14 for an in vivo pilot study in animals bearing MDA-MB-231 tumors that demonstrated a total dose of 4 mg/kg caused over 10% weight loss with no evidence of antitumor efficacy similar to the published effects of 2.³ Together, these data demonstrate that although a diverse array of C-6 taccalonolide modifications can alter target engagement and cellular potency, these modifications do not alter the overall mechanism of action of these compounds as microtubule stabilizers that promote persistent cellular effects due to their irreversible target engagement. This is consistent with previous reports demonstrating that the addition of this paclitaxel side chain onto the discodermolide backbone resulted in compounds with activity that resembled discodermolide albeit with improved cellular potency.^{11, 14} Additionally, we found that the analogs in this study impact biochemical and cellular activities in a non-linear manner, suggesting distinct effects of C-6 modifications on direct target engagement versus binding of the tubulin target in cells due to compound permeability or stability. While these data suggest that the C-6 modifications produced in this study are not likely to have improved antitumor efficacy as compared to other taccalonolides, we propose that the diversity of structural modifications that are permitted as C-6 modifications offers unprecedented access to facilitate tumor targeting and prodrug strategies that may increase the antitumor efficacy of this compound class, which could provide long-term antitumor efficacy, particularly in taxane-resistant settings.

EXPERIMENTAL SECTION

General Experimental Procedures.

Optical rotations were measured on a Rudolph Research Autopol III automatic polarimeter. NMR data were obtained on a Varian VNMR spectrometer (500 MHz for ¹H, 125 MHz for ¹³C) with a broad band probe at 25 ± 0.5 °C. High-resolution electrospray ionization mass spectrometry (HRESIMS) was performed on an Agilent Technologies 6224 TOFLC/MS mass spectrometer. Prepative HPLC separations were performed on a Shimadzu system using a SCL-10A VP controller and a Gemini 5 μm C18 column (110 Å, 250 × 21.2 mm) or a Kinetex 5 μm F5 column (110 Å, 250 × 21.2 mm) with flow rate of 10 mL/min. All solvents were of ACS grade or better.

Synthesis of Amide and Sulfonamide Taccalonolide Analogues.

The detailed procedures and reaction schemes for the synthesis of taccalonolide analogues 3 – 22 are described in the Supporting Information.

Computational Modeling.

The computational modeling experiments were conducted using the Schrödinger Small-Molecule Drug Discovery Suite (2018-4) as previously described.⁷ Briefly, the crystal structure 5EZY was downloaded from PDB and optimized using the Protein Preparation Wizard. The optimized protein structure was simplified by only retaining chain B comprising the taccalonolide AJ-β-tubulin complex. All water molecules were removed except for that forming H-bonds between β-tubulin Thr223 and the C-26 carbonyl group of taccalonolide AJ. The ligand structures were optimized using the Ligand Preparation Wizard (LigPrep). Further covalent docking experiments were performed using CovDock with Asp226 selected as the reactive residue. The docking box was centered on the coordinates X 3.2 / Y −63.5 / Z 22.6 in the length of 20 Å. The top 10 low-energy poses were generated and retained for each docking experiment. The lowest-energy pose showing the correct spatial arrangement of the taccalonolide core structure was selected for analysis of the ligand-protein binding modes.

Evaluation of Antiproliferative and Cytotoxic Potencies.

The antiproliferative and cytotoxic effects of compounds in HeLa cells (ATCC) were evaluated using the sulforhodamine B (SRB) assay. Briefly, cells were plated in a 200 μL volume in 96-well plates and allowed to adhere overnight, after which compounds were added in 1 μL of EtOH vehicle. A companion, time 0, plate was developed at the time of compound addition as a measurement of the cellular density at the time of drug addition to allow for the distinction between antiproliferative and cytotoxic efficacy. Treated plates were developed 48 h after compound addition and the antiproliferative and cytotoxic efficacies calculated as compared to the time of compound addition (y = 0) as well as the cellular density of 48 h vehicle controls (y = 100). Concentrations that caused 50% growth inhibition (y = 50) were determined by non-linear regression analysis of at least three independent experiments and calculated as the average ± SEM.

Biochemical Tubulin Polymerization.

The effects of the test compounds on the polymerization of purified tubulin were determined using purified porcine brain tubulin (Cytoskeleton) at a concentration of 20 μM in G-PEM buffer containing 80 mM PIPES pH 6.9, 2 mM MgCl₂, 0.5 mM EGTA, 1 mM GTP, and 10% glycerol. Compounds (in 1 μL EtOH vehicle) were added to 100 μL of the tubulin in G-PEM at a final concentration of 0.1 – 20 μM on ice and tubulin polymerization was monitored over time turbidometrically as an increase in absorbance at 340 nm using a Spectramax plate reader (Molecular Devices) and compared to vehicle treated controls. The rate of tubulin polymerization for each sample was determined by non-linear regression analysis in GraphPad Prism using a Gompertz growth curve fit and the inflection point of the polymerization curve (1/K) reported as an average ± SEM from at least two independent experiments.

Persistence Assay.

HeLa cells were plated in 60 mm plates at a density of approximately 500 cells per plate and allowed to adhere overnight. These cells were then treated for 4 – 8 h with each compound at relative concentrations that either completely inhibited the growth of cells in the SRB assay over the 48 h drug treatment period (TGI) or concentrations that caused 75% cytotoxicity of cells (LC₇₅) in the 48 h SRB assay. After the acute treatment period, the drug-containing medium was removed and replaced with fresh medium and cells were allowed to grow into colonies over an additional 8 days after which they were fixed and stained with 0.5% crystal violet in 20% MeOH. Colonies were counted from acquired images using ImageJ and presented as the average ± SEM from two independent experiments.

Microscopy.

HCC1937 breast cancer cells (ATCC) were plated on coverslips and allowed to adhere overnight after which they were treated with compounds or vehicle for 4 h at concentrations that caused 75% cytotoxicity of cells (LC₇₅) in the 48 h SRB assay. Cells were then fixed with cold methanol and microtubules visualized by immunofluorescence using a β-tubulin primary antibody (Sigma T-4026, 1:500) and FITC-conjugated secondary antibody (Sigma F-3008, 1:500) along with DAPI staining. Images were acquired on a Nikon i80 fluorescence microscope by compressing multiple y-stacked images using NIS elements software.