A Simplified Synthesis of Novel Dictyostatin Analogues with In Vitro Activity against Epothilone B–Resistant Cells and Antiangiogenic Activity in Zebrafish Embryos Free

Microtubules are an important component in cell division and mitosis. Interference with microtubule dynamics causes a block in cell-cycle progression and, eventually, programmed cell death (apoptosis), desirable results for treating rapidly dividing cancer cells. Microtubule-perturbing agents such as taxanes, epothilones, or Vinca alkaloids, which stabilize or destabilize microtubules, are successfully used in the treatment of solid or hematologic malignancies (1). The clinical successes of these anticancer agents have made microtubules one of the most validated molecular cancer targets. Currently, the Food and Drug Administration (FDA)-approved microtubule-stabilizing agents are the taxanes paclitaxel (Taxol), docetaxel (Taxotere), cabazitaxel (Jevtana), an albumin-bound form of paclitaxel (Abraxane), and a semisynthetic analogue of epothilone B, ixabepilone (Ixempra). In spite of their success, the development of drug resistance reduces the effectiveness of these agents (2), resulting in a continued effort to develop novel microtubule-perturbing agents.

Several microtubule-stabilizing agents are currently under investigation as potential anticancer therapies (3). A particularly promising agent, (+)-discodermolide, a potent microtubule stabilizer with activity superior to paclitaxel, entered into phase I clinical trials but failed disappointingly due to pulmonary toxicity (4). Previously overshadowed by (+)-discodermolide, (−)-dictyostatin, a closely related compound, has recently gained attention as a potential anticancer agent. A decade after isolation, the complex structure was finally resolved (5), and 2 total syntheses (6, 7) provided enough sample for a detailed characterization (7, 8). Extensive structure–activity relationship (SAR) studies have provided important information for the development of several (−)-dictyostatin analogues (9–11). These studies culminated in the discovery of 6-epi-dictyostatin, which was shown to have antitumor activity superior to paclitaxel in mice bearing human breast cancer MDA-MB-231 xenografts (12). In spite of these promising preclinical results, the complex structure and difficult synthesis of (−)-dictyostatin and analogues present major obstacles to their further preclinical development.

We recently reported a streamlined synthesis that generated new 16-desmethyl-25,26-dihydrodictyostatins that were considerably easier to make and, in preliminary biological studies, retained much of the potency of (−)-dictyostatin (13). Based on the biological activity of the series, which suggested that reduction of the C25–C26 double bond is well tolerated but removal of the C16 methyl group results in loss of activity against paclitaxel-resistant cells (13), we applied the new streamlined synthesis to generate 25,26-dihydrodictyostatin (1a) and 6-epi-25,26-dihydrodictyostatin (1b; Fig. 1). High-content cellular analysis revealed that 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin induced mitotic arrest and stabilized cellular microtubules with potencies similar to that of the natural product. In vitro, both agents caused tubulin assembly with potency similar to paclitaxel and displaced [³H]paclitaxel and [¹⁴C]epothilone B from preformed microtubules. The new analogues inhibited the growth of human cancer cells at low nanomolar concentrations, retained antiproliferative activity in epothilone B- and paclitaxel-resistant cancer cell lines, were able to synergize with paclitaxel, and had antiangiogenic activity in a zebrafish model. These data validate 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin as bona fide microtubule-stabilizing agents and identify them as candidates for continued preclinical development.

The dictyostatin analogues 1a and 1b were prepared by full syntheses. The Supplementary Material contains complete characterization details and copies of nuclear magnetic resonance spectra. Full experimental details of the synthesis will be published elsewhere. [³H]Paclitaxel was obtained from the Drug Synthesis and Chemistry Branch, National Cancer Institute. [¹⁴C]Epothilone B was a gift from Novartis Pharma.

HeLa human cervical carcinoma cells [obtained from the American Type Culture Collection (ATCC)], A549 human lung cancer cells, and their epothilone B–resistant counterparts EpoB40/A549 (a gift from Susan Horwitz, Albert Einstein College of Medicine) were maintained in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) containing 10% FBS (Cellgro), 2 mmol/L l-glutamine (Invitrogen), and 1% penicillin–streptomycin (Invitrogen). Maintenance medium for EpoB40/A549 cells contained 40 nmol/L epothilone B, which was removed before experimental setup. The HeLa/DZR cell line was generated as previously described (14), using ethyl methane sulfonate mutagenesis followed by stepwise increased concentrations of the antimitotic, tubulin assembly-inhibiting, macrocyclic polyketide disorazole C1 (0.1–10.8 nmol/L), resulting in ∼30-fold resistance to disorazole C1. These cells were valuable in our studies because they are resistant to natural products, at least in part, due to the overexpression of the ATP-binding cassette ABCB1 transporter (14). Thus, HeLa/DZR cells are cross-resistant to the natural products vinblastine, doxorubicin, and paclitaxel but not to cisplatin (14). Cells were cultured as previously described (14).

MDA-MB-231 human breast cancer cells (ATCC), 1A9 human ovarian carcinoma cells, and their paclitaxel-resistant clones 1A9/PTX10 and 1A9/PTX22 (a gift from Drs. Tito Fojo and Paraskevi Giannakakou) were maintained in RPMI-1640 medium (Invitrogen) containing 10% FBS. The maintenance medium for 1A9/PTX10 and 1A9/PTX22 cells was further supplemented with 17 nmol/L paclitaxel and 10 μmol/L verapamil. Forty-eight hours before test agent analyses, paclitaxel and verapamil were removed and the cells were placed into phenol red–free RPMI-1640 medium supplemented with 10% FBS and antibiotics. All cells were maintained in a humidified atmosphere of 95% air–5% CO₂ at 37°C. The identities of the HeLa and MDA-MB-231 cell lines were confirmed by the Research Animal Diagnostic Laboratory at the University of Missouri, Columbia, MO (http://www.radil.missouri.edu), using a PCR-based method that detects 9 short tandem repeat loci, followed by comparison of results to the ATCC short tandem repeat database.

We used our previously reported cell-based immunofluorescence assay (11, 15) for high-content analysis of mitotic arrest and microtubule stabilization. In brief, 7,500 HeLa cells per well were seeded into the wells of two 384-well collagen-coated microplates (Becton Dickinson), allowed to adhere for 5 hours, and treated for an additional 21 hours with either vehicle control [dimethyl sulfoxide (DMSO)] or test agents. Cells were fixed with 4% formaldehyde containing 20 μg/mL Hoechst 33342, permeabilized with 0.2% Triton X-100, and immunostained with the following antibody combinations: anti-α-tubulin (Sigma Aldrich; T9026; mouse monoclonal; 1:3,000 dilution)/fluorescein isothiocyanate (FITC)-labeled donkey anti-mouse immunoglobulin G (IgG; Jackson ImmunoResearch; 715-095-150; 1:500 dilution) and anti-phospho-histone H3 (Millipore; 06-570; rabbit polyclonal; 1:500 dilution)/Cy3 (indocarbocyanine dye with 3-methine linker)-labeled donkey anti-rabbit IgG (Jackson ImmunoResearch; 711-165-152; 1:500 dilution) for mitotic arrest, or anti-acetylated tubulin (Sigma Aldrich; T7451; mouse monoclonal; 1:1,000 dilution)/Cy3-labeled donkey anti-mouse IgG (Jackson ImmunoResearch; 715-165-150; 1:500 dilution) for quantitation of stabilized cellular microtubules. Cells were imaged on the ArrayScan II HCS Reader (Thermo Fisher Cellomics) using a ×20 objective and an Omega filter set XF93 (Omega Optical) at excitation/emission wavelengths of 350/461 nm (Hoechst), 494/519 nm (FITC), and 556/573 nm (Cy3). For each condition, images of 1,000 cells were acquired and analyzed using a Target Activation Bioapplication Algorithm (Thermo Fisher Cellomics), essentially as described (16). An image mask was generated from the Hoechst-stained nuclei. Microtubule density and acetylation were defined as the average pixel intensity in an area defined by the nuclear mask. For determination of mitotic index and nuclear condensation, thresholds for Hoechst 33342 and phospho-histone H3 intensities were defined as 1 SD above the average Hoechst 33342 or Cy3 intensity obtained from 28 vehicle-treated wells located at the center of the microplate. Cells were classified as positive if their average Hoechst 33342 or Cy3 intensity exceeded this threshold. Minimal detectable effective concentrations (MDEC) were estimated from concentration–response curves as described (17).

Growth inhibition of A549 and EpoB40/A549 cells was assessed over 3 days, using a modified version of our previously described high-content cytotoxicity assay (18). Cells were plated in 384-well collagen-coated plates at 1,000 cells per well, allowed to adhere overnight, and treated in quadruplicate with 10-point 2-fold serial dilutions of individual test agents or vehicle control (DMSO) for an additional 72 hours. After the 72-hour treatment period, cells were fixed and nuclei were stained with 10 μg/mL Hoechst 33342. Four imaging fields were acquired on the ArrayScan II HCS Reader (Thermo Fisher Cellomics) at excitation/emission wavelengths of 350/461 nm, using a ×10 objective, and nuclei were enumerated as described (18). Cell densities were calculated as objects per imaging field and normalized to vehicle control density at the end of the study.

Growth inhibition of 1A9 human ovarian cancer cells and the paclitaxel-resistant clones 1A9/PTX10 and 1A9/PTX22 was assessed over 3 days, using a previously described colorimetric assay (8). Cells were seeded at a low density into 96-well plates. Following a 48-hour attachment and growth period, cells were treated with a concentration range of individual test agents in quadruplicate or vehicle control (DMSO; n = 8) for an additional 72 hours. Cell proliferation was assessed spectrophotometrically after exposure to MTS that was followed by an absorbance reading at 490 nm minus the absorbance reading at 630 nm. One full microplate was developed at the end of the attachment period to determine cell numbers at the time of treatment. The 50% growth inhibitory concentrations (GI₅₀) of test agents were calculated from the spectrophotometrically determined expansion of the control cells over the 72-hour period.

HeLa/DZR cells were transfected with 20 nmol/L ABCB1 short interfering RNA (siRNA) or scrambled siRNA (Stealth siRNA and Negative Control Hi GC, both from Invitrogen) as described previously (14). Treatment with this ABCB1 siRNA caused greater than 75% decrease in ABCB1 protein levels at 24 and 72 hours after transfection, as measured by Western blotting (14). Briefly, HeLa/DZR cells were plated at a density of 7.5 × 10⁴ cells/well into a 6-well tissue culture plate and transfected 24 hours thereafter with 20 nmol/L ABCB1 siRNA or scrambled siRNA using 5 μL/well Dharmafect 1 Reagent (Dharmacon) and 480 μL/well Opti-MEM I Reduced Serum Media (transfection medium; Invitrogen) in a total volume of 2 mL/well. After 5 hours, the transfection medium was replaced with fresh medium. Twenty-four hours later, cells were detached with 0.05% trypsin, seeded on 96-well plates at a density of 1,000 cells/well, and allowed to attach overnight. Cells were then treated with test agents or vehicle control for 72 hours. Growth inhibition was determined by measuring Hoechst 33342–stained nuclei as described above.

Combination cytotoxicity studies were carried out as described previously (19). MDA-MB-231 cells were treated in quadruplicate for 96 hours with 10-point 2-fold serial dilutions of paclitaxel, test agents, or a fixed ratio of test agent and paclitaxel based on the GI₅₀ values of the individual agents. Images were acquired on the ArrayScan II HCS Reader (Thermo Fisher Cellomics) and nuclei enumerated as described above. Affected fractions (F_a) were calculated as F_a = cell density of drug treated cells/cell density of vehicle-treated cells. Data were analyzed using the median-effect analysis of Chou and Talalay (20), assuming mutually exclusive drug effects. The degree of synergism, additivity, and antagonism was measured by calculating combination indices (CI) over a range of affected fractions exactly as described previously (19).

Experiments were conducted as previously described (11) using tubulin purified in our laboratory from bovine brains by the method of Hamel and Lin (21). Microtubules were preformed by incubating 2 μmol/L bovine tubulin with 40 μmol/L 2′,3′-dideoxyguanosine 5′-triphosphate (ddGTP) in 0.75 mol/L monosodium glutamate, with pH 6.6, at 37°C for 30 minutes. In separate tubes, a 50 μL solution of 8 μmol/L test agent and 4 μmol/L radiolabeled [³H]paclitaxel or [¹⁴C]epothilone B in 0.75 mol/L monosodium glutamate, at pH 6.6, with a final DMSO content of 1%, was incubated for 10 minutes at 37°C. An aliquot (50 μL) of the preformed microtubules was added to the radioligand/test agent mixture and incubated at 37°C for an additional 30 minutes. Final concentrations of tubulin, radioligand, and test agent were 1, 2, and 4 μmol/L, respectively. Reaction mixtures were then centrifuged at 17,000 × g for 30 minutes at room temperature, and the amount of unbound radioligand determined by analyzing 50 μL of the supernatant by scintillation spectrometry (Beckman-Coulter LS6500). To account for nonspecific radioligand binding, the amount of bound radioligand was calculated by subtracting the amount of radioligand in the supernatant in the presence of test agent from the amount of radioligand in the supernatant in the presence of a large molar excess of the agent with the highest binding affinity [20 μmol/L (−)-dictyostatin; refs. 8, 11]. The extent of displacement was then calculated as % inhibition = [1 − (radioligand bound with test agent/radioligand bound with DMSO)] ×100.

Tubulin assembly was monitored turbidimetrically at 350 nm in a temperature-controlled, multichannel Beckman-Coulter 7400 spectrophotometer as described previously (8, 22). Reaction mixtures without test compounds consisted of bovine brain tubulin (1 mg/mL) in 0.1 mol/L (4-morpholino)ethane sulfonate (Mes) and were cooled to 2.5°C to establish baselines. Compounds predissolved in DMSO were added to give the indicated final concentrations, and each reaction mixture (0.25 mL final volume) was subjected to a temperature gradient. From the precooled state, the temperature was rapidly raised to 30°C (in approximately 1 minute) and maintained there for 20 minutes. The temperature was then rapidly lowered back to 0.25°C to 2.5°C. Absorbance at 350 nm was monitored every 15 seconds.

The Tg(Fli1:EGFP)^y1 transgenic zebrafish line (obtained from Dr. Brant Weinstein) was maintained as described (23). Embryos were collected at 24 hours postfertilization (hpf) and staged according to the method described by Kimmel and colleagues (24). For each condition, 5 Tg(Fli1:EGFP)^y1 transgenic zebrafish embryos were placed in 500 μL E3 medium (5 mmol/L NaCl, 0.33 mmol/L CaCl₂, 0.17 mmol/L KCl, 0.33 mmol/L MgSO₄) and treated with vehicle (DMSO, 0.5%) or various concentrations of test agents (1–25 μmol/L) for an additional 24 hours. After manual removal of the chorions, single embryos were transferred to wells of a 96-well half-area plate (Greiner) containing 40 μg/mL MS222 (tricaine methanesulfonate; Sigma) in E3 for imaging.

Photomicrographs of fluorescent intersegmental vessels (ISV) were acquired with the ImageXpress ULTRA Confocal High-Content Screening (HCS) System (Molecular Devices) using a ×4 objective and 488-nm argon laser. Images were uploaded into the Definiens Developer software suite (Definiens AG) and analyzed with a custom-designed Cognition Network Technology ruleset as described previously (25). Thresholding modifications were made to the Cognition Network Technology ruleset to accommodate the higher resolution and pixel depth of the ImageXpress system than the previously used ArrayScan (25). Total embryo size and intensity measurements were used to identify dead embryos, plate-loading artifacts, and autofluorescent compounds. Wells that contained no embryos, or embryos in which no dorsal region could be detected, were eliminated. For the remaining wells, the ruleset provided numerical measurements of ISV development (area, length, and shape). The parameter that most robustly measured ISV development was the total ISV area (in pixels). Data were normalized to vehicle controls. Experiments were repeated at least 3 times.

We recently reported a streamlined synthesis of dictyostatin and used it to prepare 2 16-desmethyl-25,26-dihydrodictyostatins epimeric at C6 (13). Based on the biological activity of the series, we concluded that the reduction of the C25–C26 double bond is well tolerated but that removal of the C16 methyl group causes loss of activity against paclitaxel-resistant cells (13). Accordingly, we selected 25,26-dihydrodictyostatin 1a and 6-epi-25,26-dihydrodictyostatin 1b as target compounds.

The streamlined route, which features high convergence, modularity, a relative ease with which structurally complex new analogues of dictyostatin can be prepared without ambiguity in the C2 and C3 configuration, and reliability of the fragment couplings, was used to make the new analogues 1a and 1b. Fragment couplings and completion of the syntheses are summarized in Fig. 1. Briefly, a Horner–Wadsworth–Emmons (HWE) reaction (26) was used to couple the known top fragment 4 (13) with new middle fragment 3 to give 5. 1,4-Reduction of the enone, removal of the para-methoxybenzyl (PMB) group, stereoselective ketone reduction, and monosilylation then provided 6. Intermolecular esterification with epimeric acid chlorides 7a,b incorporated the bottom fragment (27) to give 8a,b. Selective removal of the primary tert-butyldimethylsilyl (TBS) group and oxidation provided aldehydes 9a,b that were substrates for an intramolecular Nozaki–Hiyama–Kishi reaction (13) to give macrolactone 10a,b. Selectivity in the formation of the new stereocenter at C9 depended on the configuration at C6 with the b isomer being more selective (10b, 10/1; 10b, 3/1). Desilylation and careful purification to remove the C9-epimers provided the target products 1a and 1b. This strategy enabled the total synthesis of both analogues in a total of 39 steps, with a longest linear sequence of 11 steps from commercially available starting material.

We first characterized novel agents for mitotic arrest and microtubule perturbation by using our multiparameter high-content analysis assay (11, 15), as described in Materials and Methods. Immunofluorescence images of HeLa cells treated with test agents for 21 hours show that the new analogues, like 6-epi-dictyostatin, caused microtubule bundling (shown in green), chromatin condensation (blue), and elevated levels of phospho-histone H3 (red) at nanomolar concentrations (Fig. 1B). All agents caused concentration-dependent changes (Fig. 1C). From the range of concentrations tested, an MDEC value was determined (28). Data indicate that the new agents were equipotent to 6-epi-dictyostatin and paclitaxel. A detailed summary of the mitotic arrest assay results can be found in Supplementary Table S1.

We next asked whether the new agents stabilized microtubules in cells and caused microtubule assembly of isolated tubulin in vitro. It was previously shown that acetylated tubulin is a marker for stabilized cellular microtubules (29). Cells were stained with antibodies against α-tubulin or acetylated tubulin, respectively, to visualize cellular microtubules and microtubule acetylation. Figure 2A shows distinct differences in the concentration–response curves of tubulin and acetylated tubulin staining obtained with (−)-dictyostatin, a known microtubule stabilizer, or vincristine, a known microtubule destabilizer. In cells treated with (−)-dictyostatin, we observed a steady increase in cellular microtubule density as well as acetylated microtubules that plateaued at high concentrations. In contrast, vincristine caused an initial increase in cellular microtubule density and microtubule acetylation at low concentrations that was lower in magnitude and that reversed at higher concentrations. This bimodal response is characteristic of microtubule-destabilizing agents: the initial increase results from morphologic changes (i.e., cell rounding); the subsequent decrease is due to extraction of monomeric tubulin into the permeabilization buffer during cell processing and staining (15). Both the shape and the magnitude of microtubule and acetylated microtubule density curves caused by the dictyostatin analogues (Fig. 2A) were identical to that elicited by (−)-dictyostatin, suggesting that 25,26-dihydrodictyostatin (1a) and 6-epi-25, 26-dihydrodictyostatin (1b) caused microtubule stabilization. Immunofluorescence micrographs of acetylated microtubules confirmed the results of the automated analysis (Fig. 2B).

To further confirm the microtubule-stabilizing activity of the new analogues, we carried out in vitro tubulin assembly studies by using a turbidity assay (22), with paclitaxel as a positive control. Isolated tubulin from bovine brain was incubated with vehicle (DMSO) or various concentrations of test agents and subjected to a temperature gradient as shown in Fig. 2C. The new agents induced rapid and vigorous tubulin assembly with potency similar to paclitaxel and (−)-dictyostatin (Fig. 2C). Assembly was concentration-dependent and the resulting polymer was cold-stable, similar to paclitaxel and consistent with what we had previously observed with 6-epi-dictyostatin (11).

We previously showed that (−)-dictyostatin competes with [³H]paclitaxel and [¹⁴C]epothilone B for binding to tubulin polymer formed in the presence of ddGTP (8). We, therefore, tested whether the new analogues retained this ability. Discodermolide, (−)-dictyostatin, and the new analogues were incubated with preformed microtubules labeled with [³H]paclitaxel or [¹⁴C]epothilone, and the amount of unbound tracer was measured by scintillation spectrometry. Table 1 shows that the new analogues displaced [³H]paclitaxel and [¹⁴C]epothilone B with similar potency to discodermolide or (−)-dictyostatin. These experiments provided conclusive evidence that the new dictyostatin analogues bind the taxoid site on tubulin polymer with affinities similar to that of (−)-dictyostatin.

(−)-Dictyostatin has antiproliferative activity in paclitaxel-resistant cells (11). To assess whether the analogues remained active in drug-resistant cancer cell lines, we tested 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin in paclitaxel-resistant 1A9 human ovarian cancer cells with β-tubulin mutations, Phe270 –> Val and Ala364 –> Thr (30), induced by long-term culture with paclitaxel, and in epothilone B–resistant A549 human lung cancer cells that harbor a point mutation in β-tubulin (Gln292 –> Glu) as a result of long-term exposure to epothilone B (31). Table 2 shows that cross-resistance to paclitaxel in the 1A9/PTX10 cells was reduced from 49-fold to 15-fold with (−)-dictyostatin and further reduced with the new analogues (7- and 8-fold for 1a and 1b, respectively). Similarly, cross-resistance to epothilone B was reduced with (−)-dictyostatin [from 94-fold for epothilone B and 18-fold for paclitaxel to 10-fold with (−)-dictyostatin], and further diminished with the new analogues (5-fold and 3-fold, respectively, for 1a and 1b). Furthermore, diminished cross-resistance was observed in a recently described disorazole C1–resistant human cervical carcinoma cell line that overexpresses the ABCB1 P-glycoprotein pump (14). Consistent with previously published data (14), these cells were 1,395- and 502-fold resistant to paclitaxel and vinblastine, respectively (Table 2). In contrast, the new dictyostatin analogues showed greatly reduced cross-resistance to disorazole C1 compared with paclitaxel and vinblastine, with a residual 12- and 18-fold resistance for 1a and 1b, respectively. To investigate further whether the new analogues were affected by multidrug-transport proteins, we carried out siRNA knockdown of ABCB1, which reversed the residual cross-resistance in the disorazole C1–resistant cells (Table 2).

Discodermolide and paclitaxel represent a synergistic drug combination in human cancer cells (32). Therefore, we examined the novel dictyostatin analogues in combination with paclitaxel to determine whether they also resulted in synergy. We used our previously described growth-inhibition assay (18), together with median-effect analysis (20), to quantify synergism, additivity, and antagonism. MDA-MB-231 cells were treated with comprehensive concentration gradients of paclitaxel, discodermolide, 6-epi-dictyostatin, 25,26-dihydrodictyostatin 1a, 6-epi-25,26-dihydrodictyostatin 1b, or equipotent, fixed mixtures thereof in combination with paclitaxel for 4 days, and cell densities quantified by counting Hoechst 33342–stained nuclei. Median effect (D_m), slopes (m), and correlation coefficients (r) for individual agents and their combinations can be found in Supplementary Table S2. CIs were then calculated for various effect levels by the method of Chou and Talalay (20, 33) as described previously (18).

As shown in Fig. 3, we reproduced the results of Martello and colleagues (32), who found the combination of paclitaxel and discodermolide to be synergistic at lower effect levels and antagonistic at high effect levels. The dictyostatins had CI profiles similar to that of discodermolide, although the degree of synergism was lower. The least potent combination was with 6-epi-25,26-dihydrodictyostatin 1b (Fig. 3D), which was additive over much of the effect range. These results were consistently repeated over the course of multiple independent experiments. The data suggest that (−)-dictyostatin and the new analogues share the ability of discodermolide to synergize with paclitaxel, a feature that is potentially favorable for clinical use.

Some microtubule-perturbing agents have antiangiogenic activity that contributes to in vivo anticancer activity (34). Solid tumors require an adequate supply of blood vessels to survive, grow, and metastasize (reviewed in ref. 35), and agents targeting tumor angiogenesis are now FDA-approved anticancer medicines (e.g., bevacizumab, Avastin). We, therefore, asked whether the dictyostatin analogues had antiangiogenic activity. We used the Tg(fli1:EGFP)^y1 zebrafish line that expresses enhanced green fluorescent protein (EGFP) under the control of the Fli1 promoter, thereby labeling all blood vessels and providing a live visual marker for vascular development (36). Zebrafish have a stereotypical vertebrate vasculature that develops in response to the same signals that guide mammalian blood vessel development (37, 38). Furthermore, zebrafish vasculature recruitment occurs in response to human glioma xenografts (39, 40), mimicking conditions found in mammals.

Tg(fli1:EGFP)^y1 zebrafish embryos at 24 hpf were treated for 24 hours with vehicle control or various concentrations of test agents and imaged. Figure 4A shows that, as expected, vehicle-treated embryos had well-established ISV that extended from the dorsal aorta (DA) and connected to the dorsal longitudinal anastomotic vessel (DLAV; Fig. 4A; ref. 41). Visually, all of the dictyostatin analogues stunted ISV outgrowth and prevented the establishment of the DLAV (Fig. 4A, top panels). Our previously described image analysis algorithm (25) quantified the antiangiogenic phenotype (Fig. 4A, bottom panels). All agents inhibited angiogenesis in a concentration-dependent manner (Fig. 4B), with concentrations required to reduce ISV area by 50% compared with control (IC₅₀) of 8.8, 6.1, and 6.7 μmol/L for 6-epi-dictyostatin, 25,26-dihydrodictyostatin 1a, and 6-epi-25,26-dihydrodictyostatin 1b, respectively. Importantly, at concentrations that were antiangiogenic, we observed no obvious signs of toxicity such as the appearance of necrotic opaque cells. At the highest concentration tested (25 μmol/L, data not shown), the test agents caused a bent-tail phenotype, suggesting that the compounds at this concentration would likely cause developmental defects in the embryo.

The complex chemical structure and difficult synthesis of the dictyostatins are major impediments to their development into novel antineoplastic agents. This study validates that our recently described synthetic route (13) can be used to rapidly generate new analogues. The streamlined route features a bimolecular esterification to create the C1–O21 bond in place of the usual macrolactonization. This bypasses a major problem of Z/E isomerization of the C2 and C3 alkene that has plagued the macrolactonization. In turn, the large ring is closed by a mild Nozaki–Hiyama–Kishi reaction to make the C9–C10 bond. It should be possible to access many more analogues, thanks to the modularity of this route and the reliability of the fragment couplings and end-game steps.

Consistent with earlier findings, removal of the C16 methyl moiety did not dramatically affect antiproliferative activity in human tumor cells expressing wild-type (WT) tubulin but diminished the ability of the compounds to inhibit the growth of paclitaxel-resistant clones harboring mutations within β-tubulin (10). We, therefore, reasoned that retaining the C16 methyl group would preserve the lack of cross-resistance to paclitaxel, and selected 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin as target compounds. Consistent with existing SAR, both new agents showed low nanomolar antiproliferative activity in HeLa, A-549, and MDA-MB-231 cells and reduced cross-resistance in paclitaxel- and epothilone B-resistant cells with mutant tubulin.

To confirm that the new analogues directly interact with their proposed target, we carried out radioligand binding studies. These experiments show that the new analogues have affinities for the taxane site similar to paclitaxel, epothilone B, or discodermolide. The precise location of the dictyostatin binding site has not been established, because the interaction of the dictyostatins or discodermolide with tubulin has not been solved by cryoelectron microscopy as it has been for paclitaxel and epothilone A (42, 43). Furthermore, 2 binding sites have been described for taxanes: an internal luminal binding site and an external transient binding site. The radioligand competition studies are unable to distinguish between the 2 sites; however, growth-inhibition studies of the natural product (8) and of the 16-desmethyl analogues using 1A9/PTX10 ovarian cancer cells with the Phe270 —> Val mutation that we carried out previously (13) are consistent with dictyostatin and analogues binding to the internal site.

The new analogues retained some, but not all, of the ability of discodermolide to synergize with paclitaxel in human breast cancer cells. Modeling studies based on nuclear magnetic resonance structures have suggested that the bound conformer of dictyostatin resembles that of discodermolide and provides similar contacts with tubulin (44). Because it is unusual for 2 drugs that bind to identical sites on the same target to show synergy, the combination cytotoxicity data do support the previously proposed model of overlapping binding sites for paclitaxel and the dictyostatins (44). The extent of synergy varied with the analogues; the least potent agent was 1b, although all of them showed a trend toward higher synergy at lower effect levels. Therefore, our results confirmed a synergistic relationship, particularly at the lower concentrations of the 2 drugs, as reported by Horwitz's group (32). The reasons for the differential activity of the analogues in this assay are unknown. The fact that the dictyostatins were essentially equivalent in all of our assays, including the in vitro radioligand binding studies, makes it seem unlikely that differences in binding affinity or cellular distribution would account for the observed differences. To formulate a valid hypothesis based on structural terms, however, physical evidence such as a high-resolution cryoelectron microscopy structure of the dictyostatins and discodermolide is needed. Alternatively, the different degree of synergy of the dictyostatins compared with discodermolide may be a result of off-target effects. As pointed out by Martello and colleagues (32), discodermolide induces apoptosis by mechanisms unrelated to microtubule binding, and it is currently not known whether dictyostatins share these activities. The data do suggest, however, that the combination of paclitaxel with either 6-epi-dictyostatin or 1a merits exploration in in vivo antitumor studies.

Drug resistance is a major problem with microtubule-perturbing agents in clinical use. One clinically important resistance mechanism is overexpression of P-glycoprotein efflux pumps (45). In cultured cells, additional resistance mechanisms have been observed that involve tubulin mutations induced by long-term culture of cell lines in the presence of microtubule-perturbing agents (31, 46), although such drug-induced mutations have not been found in clinical samples. In 3 such cellular models with mutant tubulin, the new analogues retained activity against both paclitaxel- and epothilone B–resistant cells and appeared less cross-resistant than the natural product. The 1A9/PTX10 cell line harbors a Phe270 —> Val mutation that is located within the taxane binding site (43) and confers 49-fold resistance to paclitaxel. Consistent with our previous studies with (−)-dictyostatin and 6-epi-dictyostatin (13), resistance was reduced to less than 10-fold with the new analogues (Table 2). As expected, no cross-resistance was found in the 1A9/PTX22 cell line, which has a Ala364 —> Thr mutation that is adjacent to the taxane-binding pocket. In epothilone B–resistant A-549 cells with a 292Gln —> Glu mutation, which is located at the periphery of the taxane pocket and makes contact with epothilone but not paclitaxel (42), the analogues showed only 12- to 18-fold resistance compared with epothilone B (94-fold resistance). These data indicate that reduction of the terminal double bond does not alter the mode of tubulin binding. They are consistent with a mode of binding to tubulin, as proposed by Canales and colleagues (44), that involves the taxane-binding pocket but not residues outside the pocket that make contact with the taxane side chain.

The analogues showed a unique behavior toward cells with acquired resistance against the natural product disorazole C1 (14), which owe their resistance phenotype, at least in part, to overexpression of the ABCB1 P-glycoprotein pump. All agents were subnanomolar inhibitors of WT HeLa cells. Paclitaxel and vinblastine were 1,395- and 502-fold less active, respectively, in the resistant cells (HeLa/DZR cell line; Table 2). Knockdown of the P-glycoprotein pump, ABCB1, restored most of their activity (HeLa/DZR/ABCB1 siRNA; Table 2). In contrast, the HeLa/DZR cells showed only minor cross-resistance to the dictyostatin analogues (12- to 18-fold; HeLa/DZR; Table 2), and it was fully reversed by ABCB1 knockdown. These data suggest that the dictyostatins may be only weak substrates for ABCB1. Moreover, because the HeLa/DZR cells were generated by a single exposure to the mutagen ethyl methanesulfonate followed by a stepwise increased disorazole C1 exposure, it is likely that resistance mechanisms other than elevated ABCB1 exist, but these do not appear to influence cellular sensitivity to the dictyostatin analogues.

We had previously shown that microtubule-perturbing agents inhibit angiogenesis in Tg(fli1:EGFP)^y1 transgenic fluorescent zebrafish embryos (15). In this study, we showed that the new analogues also have this property, which is thought to be beneficial for clinical activity (34, 46). In the Tg(fli1:EGFP)^y1 model, the agents appeared to have antiangiogenic rather than antivascular activity. During development, ISVs sprout from the DA at 24 hpf, and, at 48 hpf, are fully established and connected to the DLAV. To assess the effect of test agents on new vessel outgrowth (angiogenesis), embryos were treated at 24 hpf (when ISVs are just beginning to sprout and are barely visible; ref. 15) and analyzed for ISV formation 24 hours thereafter. Although the analogues caused a concentration-dependent inhibition of new vessel growth, they did not affect existing blood vessels, as the head and large trunk vessels were intact. Furthermore, heart beat, circulation, and twitch response (assessed visually) were all normal (data not shown). Furthermore, we did not observe tissue necrosis, which would appear as opaque cells in the fluorescence micrographs (see Fig. 4). In addition, test agent–treated embryos showed little difference in gross morphology when compared with control embryos (Fig. 4), although we did observe a bent-tail phenotype at the highest concentration tested (25 μmol/L). Although the model is currently not sufficiently well-characterized to suggest therapeutic safety in the context of angiogenesis inhibition, the data indicate the new dictyostatins have antiangiogenic activity in a zebrafish model of angiogenesis at nontoxic concentrations.

In summary, we have used our previously reported, highly convergent, streamlined synthesis (13) to generate 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin, 2 new analogues of the highly complex natural product (−)-dictyostatin. Consistent with existing SAR studies and a mode of action involving high-affinity binding to the taxane site on tubulin, the new analogues retained essentially all of the biological activities of (−)-dictyostatin and 6-epi-dictyostatin, the only analogue whose activity in adult mammals has been described to date (12). While the new analogues do not represent a significant simplification from a structural viewpoint, reduction of the exposed double bond eliminates chemical reactivity and a potential metabolic soft spot, as has been shown for discodermolide (47). Future experiments should focus on this issue. The results identify 25,26-dihydrodictyostatin and 6-epi-25,26-dihydrodictyostatin as candidates for scale-up synthesis by using the improved synthesis procedure and further preclinical development.

No potential conflicts of interest were disclosed.

We thank Dr. Brant Weinstein for the transgenic Tg(fli1:EGFP)^y1 line, Dr. Susan B. Horwitz for the epothilone B-resistant cells, Drs. Tito Fojo and Paraskevi Giannakakou for the paclitaxel-resistant clones, the National Cancer Institute for [³H]paclitaxel, and Novartis Pharma for [¹⁴C]epothilone B.

This work was supported by the NIH (grants CA78039 to A.Vogt and J.S. Lazo, HD053287 to N.A. Hukriede) and the Fiske Drug Discovery Fund.