In Vitro and In Vivo Anticancer Activities of Synthetic Macrocyclic Ketone Analogues of Halichondrin B

INTRODUCTION

Halichondrin B is a large polyether macrolide (Fig. 1) ⇓ found in a variety of marine sponges, including Halichondria okadai (1 , 2) , Axinella sp. (3) , Phakellia carteri (4) , and Lissondendryx sp. (5) . In 1986, halichondrin B was shown to have remarkable in vitro and in vivo anticancer activities against murine melanomas and leukemias (2) . Subsequent studies by several groups (6, 7, 8, 9) confirmed the sub-nm in vitro potency of halichondrin B and identified a tubulin-based antimitotic mechanism. Although halichondrin B is classified as a tubulin depolymerizer (similar to Vinca alkaloids, dolastatins, cryptophycin, and so on), the patterns of its interaction with tubulin are highly specific and place it in its own unique class within this group (6, 7, 8, 9) . Although speculative, unique interactions with tubulin may help explain the marked in vivo anticancer potency of halichondrin B, now well documented in several human tumor models including melanoma, osteogenic sarcoma, lung cancers, and colon cancers (10) . More importantly, unique tubulin interactions relative to existing tubulin-based drugs suggest the possibility of unique tumor specificities or other desirable clinical characteristics.

Although the extraordinary potency of halichondrin B generated considerable interest for development, limited availability of the natural product severely restricted such efforts. However, the existence of a synthetic route for halichondrin B (11) and knowledge that its activity resides in the macrocyclic lactone C1-C38 moiety 3 (see also Refs. 12, 13, 14 ) have permitted development of structurally simplified synthetic analogues that retain the exceptional activity of the parent. In this study, we describe ER-076349 and ER-086526, two macrocyclic ketone analogues of halichondrin B with highly potent activities in vitro and in vivo. We show that these agents exert their effects via a tubulin depolymerizing antimitotic mechanism similar or identical to that reported for parental halichondrin B.

MATERIALS AND METHODS

Chemical Synthesis.

Halichondrin B macrocyclic ketone analogues ER-076349 (NSC 707390) and ER-086526 (NSC 707389) were prepared as described previously (15) . The biotinylated analogue ER-040798 and the biotinylated negative control ER-040792 were synthesized based on appropriate modifications to the published synthetic route for halichondrin B (11) . The structures of halichondrin B and all of the synthetic analogues are presented in Fig. 1 ⇓ .

Cells and Cell Culture.

The following human tumor cell lines were obtained from the American Type Culture Collection (Rockville, Maryland) and grown under American Type Culture Collection-recommended conditions: COLO 205 and DLD-1 colon cancer, HL-60 promyelocytic leukemia, U937 histiocytic lymphoma, and LNCaP and DU 145 prostate cancer. LOX human melanoma cells were obtained from the Division of Cancer Treatment-NCI 4 Tumor Repository (Frederick, Maryland) and were grown in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum and 2 mm l-glutamine. MDA-MB-435 human breast cancer cells were generously provided by Dr. Mary J. C. Hendrix (University of Iowa College of Medicine, Iowa City, Iowa) and were grown in DMEM (high glucose) supplemented with 10% heat-inactivated fetal bovine serum, 20 mm N-2-hydroxyethyl-piperazine-N′-2-ethanesulfonic acid and 1 mm sodium pyruvate.

All of the cell lines were grown at 37°C in a humidified atmosphere containing 5% CO₂. Seeding densities for each cell line were empirically optimized for passaging twice weekly. Quantification of cell number was by hemocytometer counting, and viability determinations were made by standard trypan blue exclusion techniques. Routine harvesting of monolayer cultures was by standard trypsinization procedures.

Cell Growth Inhibition Assays.

Cells were plated in 96-well plates at 7,500 cells/well (except LNCaP cells, which were at 10,000 cells/well) and grown in the continuous presence of test compounds for 4 days. For monolayer cultures (DU 145, LNCaP, LOX, and MDA-MB-435), growth was assessed using modifications (16) of a methylene blue-based microculture assay (17) . For suspension cultures (HL-60, U937) or loosely adherent monolayers (COLO 205), growth was assessed using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide-based assay (18) modified as follows. After 4 days of incubation with test compounds, sterile-filtered 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (Sigma Chemical Co., St. Louis, MO) was added to each well (final concentration, 0.5 mg/ml), and plates were incubated at 37°C for 4 h. Acid-isopropanol (0.1 n HCl in isopropanol, 150μ l) was then added to each well, and the resultant formazan crystals were dissolved by gentle mixing. Absorbances at 540 nm were measured on a Titertek Multiscan MCC/340 plate reader.

Cell Cycle Analysis.

DNA content analysis of U937 cells was performed by single channel flow cytometry as follows. U937 cells were exposed for 0–14 h to 30 nm ER-076349 or 100 nm ER-086526; concentrations were roughly 10-fold higher than those that minimally induce complete G₂-M blocks based on dose/response studies. Typically, 2–4 × 10 6 cells were collected by centrifugation at 300 × g for 10 min and resuspended in 3 ml of saline, followed by the immediate addition of 7 ml of ice-cold 100% ethanol while vortexing. After fixing overnight at 4°C in the saline/ethanol solution, cells were recovered by centrifugation, washed in PBS, and resuspended in 0.5 ml PBS containing 200 μg/ml RNase A (Sigma; preboiled 5 min as 10 mg/ml stock in PBS; stored at −80°C). RNase digestion was at 37°C for 30 min, followed by the addition of 0.5 ml of 10 μg/ml propidium iodide (Sigma) in PBS at least 15 min before flow analysis. Single channel flow cytometry was performed on a Becton Dickinson FACScan with a Macintosh computer; the collection and analysis of data were performed using Becton Dickinson CELLQuest software. Doublet events were eliminated from analyses by proper gating on FL2-W/FL2-A primary plots before histogram analysis of DNA content (measured as FL2-A).

Fluorescence Microscopy.

For simultaneous anti-tubulin and DAPI staining of DU 145 cells, cells were plated in 4-well plastic Lab-Tek chamber slides (Nunc, Inc., Naperville, IL) at 7.5 × 10 4 cells/chamber in 0.9 ml of complete medium. After culturing overnight, 100 μl of 10 × concentrates of test compounds in complete media were added to a final concentration equal to three times the empirically determined IC₅₀ of each compound; control wells received 100 μl of complete media only. After 20 h, cell monolayers were fixed with ice-cold anhydrous methanol for 6 min at −20°C, followed by two washes with 0.1% (v/v) Tween 20 in PBS and two washes with PBS alone. Fixed cells were then incubated with 1:50 diluted (v/v in PBS) anti-β-tubulin TUB 2.1 monoclonal antibody (Sigma) for 1 h at room temperature with rocking, followed by three 5-min washes with PBS and incubation with 1:50 diluted (v/v in PBS) FITC-conjugated goat F(ab′)₂ antimouse immunoglobulin (BioSource International, Camarillo, CA) for 1 h at room temperature with rocking. Cells were washed 3× with PBS, incubated with 2 μg/ml DAPI in PBS for 10 min at room temperature, followed by one 5-min wash with PBS and two 5-min washes with 2 mm Tris, 0.1 mm EDTA (pH 7.5). Chamber walls and gaskets were then removed, and the slides were air-dried. Coverslips were mounted with Gel/Mount (Biomeda, Foster City, CA), and fluorescence microscopy was performed using a Nikon Labophot-2 microscope equipped with 100 × oil immersion objective and B-2A and UV-2A filter cubes for FITC and DAPI staining, respectively. Images were acquired with a Photometrics NU 200 CCD camera and digitized using IPLab Spectrum software (Scanalytics, Fairfax, VA).

Affinity Binding of Tubulin to Biotinylated Macrocyclic Analogue ER-040798.

UltraLink™ Immobilized NeutrAvidin Plus (Pierce Chemical Co., Rockford, IL) was charged with ER-040792, ER-040798 (135μ m final concentrations), or buffer alone at room temperature for 60 min in 20 mm Tris, 0.15 m NaCl, 1 mm Na₂EDTA (pH 7.5; wash-elution buffer). After five washes in wash-eluted buffer, matrices were incubated in the same buffer with 375 μg/ml bovine brain tubulin (Molecular Probes, Eugene, OR) at room temperature for 30 min, followed by three washes in wash-elution buffer and elution with 20% acetonitrile (v/v in water). Eluted proteins were dried in a Speed-Vac, resuspended in SDS sample buffer, and subjected to modified discontinuous SDS PAGE on 8% gels as described by Matthes (19) for separation of α- and β-tubulin subunits. Gels were silver-stained (20) or transferred to polyvinylidene difluoride membranes (Sigma) for Western immunoblotting (21) with subtype-specific anti-α-tubulin or anti-β-tubulin monoclonal antibodies (clones B-5-1-2 and TUB 2.1, respectively; Sigma) and horseradish peroxidase-conjugated sheep antimouse immunoglobulin secondary antibodies (Amersham Corp., Arlington Heights, IL) using a chemiluminescence protocol (ECL Western Blotting Detection System; Amersham).

In Vitro Tubulin Polymerization Studies.

ER-076349, ER-086526, and vinblastine in 10 mm anhydrous DMSO stocks were diluted into 10% DMSO, 90% PEM. Substocks were made to final concentrations of 1, 10, 100, and 1000 μm; these were diluted into 100-μl volumes of 3.0 mg/ml bovine brain tubulin (Cytoskeleton, Inc., Denver, CO) in PEM buffer plus 1 mm ATP, 3% (v/v) glycerol to achieve desired test compound concentrations. The amount of 10% DMSO, 90% PEM was made up to 11μ l in all of the samples to achieve 1% final DMSO concentrations. Microtubule polymerization was initiated by raising the temperature from 4°C to 37°C over a 3-min period. Absorbance (A_{340 nm}) was measured once/min for 60 min.

In Vivo Xenograft Studies.

Anticancer effects of ER-076349 and ER-086526 were evaluated in the MDA-MB-435 human breast cancer, COLO 205 human colon cancer, and LOX human melanoma xenograft models using 5–6-week-old female Swiss nude mice and in the NIH:OVCAR-3 human ovarian cancer model using 7-week-old female BALB/c nude mice. All of the studies using laboratory animals were approved either by the Eisai Research Institute (Andover, MA) or Tsukuba Research Laboratories (Tsukuba, Japan) Institutional Animal Care and Use Committees and adhered to all of the applicable institutional and governmental guidelines for the humane care and use of laboratory animals.

On day 0 of each experiment, mice received injections s.c. with 1 × 10 6 cells (MDA-MB-435, COLO 205, and LOX) or 6 × 10 5 cells (NIH:OVCAR-3; previously adapted for enhanced penetrance via tumor passage in vivo). For COLO 205 and MDA-MB-435 models, mice received injections with 200 μl of test compound in saline on Monday/Wednesday/Friday i.p. (COLO 205) or i.v. (MDA-MB-435) schedules, respectively, beginning on day 13 for four weekly cycles. Treatment of mice bearing LOX tumors was by daily i.p. injection (200μ l) in saline on days 3–7 and 10–14 inclusive. In the NIH:OVCAR-3 model, test compounds were injected i.v. (200μl/20g body weight) in 4% DMSO/saline beginning on day 40 and continuing Monday/Wednesday/Friday for three weekly cycles. Control groups in all of the studies received appropriate vehicle injections. Paclitaxel at its empirically determined MTD for each treatment regimen was included in all of the experiments and was administered using Diluent 12 (1:1 Cremophor EL/ethanol diluted into 5% glucose in water) as described by others (22 , 23) . Group sizes were 10 mice for all of the tumor models except NIH:OVCAR-3, which used 6 mice/group. Tumor volumes were determined by caliper measurements and were calculated according to V = 0.67π[(L + W)/4] 3 (i.e., volume[ V] of a semisphere with diameter equal to the mean of tumor length [L] and width [W]). Body weights and water consumption were measured in all of the experiments to assess toxicity.

RESULTS

Inhibition of Human Tumor Cell Growth in Vitro.

More than 180 analogues of the macrocyclic C1-C38 moiety of halichondrin B were synthesized and evaluated in search of suitably active and chemically stable drug development candidates. This culminated in the identification of two simplified macrocyclic analogues, ER-076349 and ER-086526, in which the C1 lactone ester of halichondrin B is replaced by ketone, C31 methyl is replaced by methoxy, and the tricyclic C29-C38 system is replaced by a single five-membered ring (Fig. 1) ⇓ . The two analogues themselves differ only at C35, with hydroxyl and primary amine substituents for ER-076349 and ER-086526, respectively.

On the basis of the known anti-tubulin mechanism of halichondrin B (6, 7, 8, 9) , in vitro growth inhibitory effects of ER-076349 and ER-086526 were compared with those of the microtubule destabilizer vinblastine and the microtubule stabilizer paclitaxel in MDA-MB-435 human breast cancer cells (Fig. 2) ⇓ . In this experiment, ER-076349 and ER-086526 inhibited cell growth at sub-nm levels (IC₅₀ values of 0.15 nm and 0.07 nm, respectively) with greater potency than vinblastine (0.54 nm) or paclitaxel (2.6 nm). Similar low- or sub-nm potencies were observed in several other human cancer cell lines, including COLO 205 and DLD-1 colon cancers, DU 145 and LNCaP prostate cancers, LOX melanoma, HL-60 monocytic leukemia, and U937 histiocytic lymphoma (Table 1) ⇓ . Interestingly, DLD-1 cells were about 10-fold less sensitive to ER-086526 compared with the other seven lines (overall mean without DLD-1 data, 0.7 ± 0.2 nm); this was not seen for ER-076349. The reasons for the relative insensitivity of DLD-1 cells to ER-086526 are unknown, although it is noted that these cells were also less sensitive to vinblastine and paclitaxel.

IC₅₀ values for growth inhibition by ER-076349 and ER-086526 agree well with values previously obtained using synthetic halichondrin B, where an overall mean of 0.4 ± 0.1 nm was obtained across 17 human cancer cell lines (24) . Thus, the structural simplifications of ER-076349 and ER-086526 relative to halichondrin B have not significantly altered growth inhibitory potency.

Despite sub-nm growth inhibition of several human cancer cell types, neither ER-076349 nor ER-086526 showed cytotoxic effects against quiescent IMR-90 human fibroblasts at concentrations up to 1μ m, as assessed using an ATP-dependent luciferase-based viability assay (data not shown). Thus, growth inhibition by low- or sub-nm levels of ER-076349 and ER-086526 is specific for proliferating cells and not secondary to nonspecific cytotoxicity.

Cell Cycle Effects: G₂-M Phase Block.

Cell cycle analysis of U937 human histiocytic lymphoma cells was performed after 0–14 h-incubation with 30 nm ER-076349 or 100 nm ER-086526. As shown in Fig. 3 ⇓ , treatment with either ER-076349 or ER-086526 for as little as 2 h led to increased numbers of G₂-M phase cells. In both groups, G₂-M cells accumulated at the expense of G₁, which was almost completely depleted by 8 h. Progressive depletion of S phase then followed beginning at 10 h because of cessation of new cells entering S from G₁. Depletion of S phase was essentially complete by 14 h, at which time almost all of the living cells were blocked in G₂-M. For both compounds, increased numbers of hypodiploid events were seen after 8–10 h, indicating apoptosis of cells after prolonged G₂-M blockage. The fact that both G₁ and S phases became progressively depleted under continuous ER-076349 or ER-086526 exposure confirms normal movement of cells through these phases even in the presence of a drug. Thus, ER-076349 and ER-086526 are rapid and effective G₂-M phase blockers that do not affect cell cycle progression through the G₁ or S phases or the G₁-S transition point. Similar G₂-M blockage of U937 cells was observed previously using synthetic halichondrin B (24) .

Disruption of Mitotic Spindles.

The known anti-tubulin effects of halichondrin B (6, 7, 8, 9) together with G₂-M blockage by ER-076349 and ER-086526 led us to assess effects on mitotic spindles in DU 145 human prostate cancer cells (Fig. 4) ⇓ . In this study, cells in log phase growth were treated for 20 h with vehicle, 2.1 nm ER-076349, 2.7 nm ER-086526, 11 nm vinblastine, or 28 nm paclitaxel, representing empirically determined 3 × IC₅₀ concentrations in this cell line (preliminary studies comparing 0.3, 1, 3, and 10 × IC₅₀ levels established 3 × IC₅₀ as optimal; data not shown). After drug exposure, cells were fixed and dual stained with anti-β-tubulin monoclonal antibodies and DAPI to label microtubules and DNA, respectively. As shown in Fig. 4 ⇓ , treatment with either ER-076349 (B) or ER-086526 (C) led to marked disruption of normal mitotic spindle architecture (A). As expected, mitotic spindle disruption was also seen with the anti-tubulin agents vinblastine (D) and paclitaxel (E). With all of the four drugs, cells with disrupted spindles showed the condensed but disorganized chromosomal material characteristic of prometaphase blockage. Occasional ER-076349- and ER-086526-treated cells were seen that had not yet reached the mitotic block point during the 20-h drug treatment period. These interphase (G₁, S, or G₂) cells, identified by intact DAPI-stained nuclei, appeared essentially normal with respect to filamentous cytoplasmic microtubules and intact, well-defined nuclear perimeters, consistent with the lack of effect of ER-076349 and ER-086526 on progression through G₁ and S seen by flow cytometry (Fig. 3) ⇓ . Similar disruption of DU 145 mitotic spindles was observed previously using synthetic halichondrin B (24) .

Subtle drug-to-drug differences in mitotic spindle disruption patterns were noted at the 3 × IC₅₀ concentrations used in this study. Although such differences might derive from different tubulin-based mechanisms (inhibiting versus promoting tubulin polymerization, differences in tubulin binding, and so on), it is known that spindle disruption patterns by Vinca alkaloids and paclitaxel at concentrations close to growth inhibitory IC₅₀ levels look strikingly similar regardless of mechanistic differences (25 , 26) . Indeed, we and others (data not shown and Refs. 25, 26, 27 ) have noted that only small changes in drug concentration close to growth inhibitory IC₅₀ levels can lead to marked differences in spindle appearance. Thus, we cannot conclude that the subtle differences in spindle disruption patterns noted in Fig. 4 ⇓ are related to mechanistic differences.

Binding of Biotinylated Analogue ER-040798 to Tubulin.

To directly address whether macrocyclic halichondrin B analogues were capable of tubulin binding, the ability of the biotinylated analogue ER-040798 (Fig. 1) ⇓ to bind tubulin under affinity chromatography conditions was tested. As shown in Fig. 5 ⇓ , immobilized ER-040798, but not the biotinylated negative control ER-040792, retained two M_r 50,000–55,000 protein bands from bovine brain tubulin (Lanes A–E). These bands were recognized by subtype-specific anti-α-tubulin (Lanes F–J) and anti-β-tubulin (Lanes K–O) monoclonal antibodies, confirming their identities as α- and β-tubulin, respectively. Thus, ER-040798 directly binds tubulin, suggesting that the biological activity of halichondrin B macrocyclic analogues derives from mechanisms similar or identical to parental halichondrin B. It should be noted that this study did not identify the tubulin subtype to which ER-040798 binds, because binding to either subtype in α/β-tubulin heterodimers would lead to retention of both. Finally, specific retention of α- and β-tubulin by immobilized ER-040798 was also seen using U937 whole cell lysates (data not shown), which demonstrated that the retention of α- and β-tubulin seen in Fig. 5 ⇓ was not simply because of the presence of only these two proteins in the purified bovine brain tubulin preparations used.

Inhibition of Tubulin Polymerization in Vitro.

To determine whether the tubulin-depolymerizing mechanism of halichondrin B (6) had been conserved during structural simplification to macrocyclic ketones, effects of ER-076349 and ER-086526 on polymerization of bovine brain tubulin in vitro were assessed and compared with the known tubulin depolymerizer vinblastine. As shown in Fig. 6 ⇓ , ER-076349, ER-086526, and vinblastine inhibited both rates and extents of tubulin polymerization in dose-dependent fashions. The inhibition of polymerization rates and extents was similar for each drug, with calculated IC₅₀s for ER-076349, ER-086526, and vinblastine of 5.7 μm, 6.9μ m, and 1.8 μm for rate inhibition, and 5.5 μm, 6.0μ m, and 2.1 μm for extent inhibition, respectively. By either criterion, ER-076349 and ER-086526 were about 3–4-fold less effective as inhibitors of in vitro tubulin polymerization compared with vinblastine.

Although concentrations of all of the three drugs required to inhibit tubulin polymerization in vitro are several log orders higher than those that inhibit cell growth, such large discrepancies with tubulin-active agents may result from inhibited microtubule dynamics at low drug concentrations rather than net changes in cellular microtubule content which occur only at higher drug levels (25 , 26) . In addition, Vinca alkaloids and paclitaxel are known to accumulate inside cells to high levels (25 , 26) ; this may also occur with ER-076349 and ER-086526.

Inhibition of Human Tumor Xenograft Growth in Vivo.

The abilities of ER-076349 and ER-086526 to inhibit tumor growth in vivo were tested in several human tumor xenograft models, including MDA-MB-435 breast cancer, COLO 205 colon cancer, LOX melanoma, and NIH:OVCAR-3 ovarian cancer (Fig. 7) ⇓ . In all of the cases, treatment with ER-076349 and ER-086526 in the 0.05–1 mg/kg range led to significant antitumor effects, with ER-086526 showing greater efficacy in all of the models. Thus, in the MDA-MB-435 model (Fig. 7, A and B) ⇓ , treatment with 0.25–1.0 mg/kg ER-076349 led to 60–70% inhibition at day 42, whereas ER-086526 at the same levels led to >95% inhibition, including actual regression of measurable tumors present on day 14. For both compounds, tumor regrowth rates after cessation of dosing allowed efficacy comparisons with paclitaxel at its empirically determined MTD of 25 mg/kg. Again, ER-086526 showed superiority over ER-076349; regrowth in all of the ER-076349 groups preceded the paclitaxel group by about 5 weeks, whereas all of the doses of ER-086526 were either equally efficacious (0.25, 0.5 mg/kg) or superior (1 mg/kg) to 25 mg/kg paclitaxel. Interestingly, the therapeutic window of ER-086526 seemed unusually large for a cytotoxic drug; >95% tumor suppression occurred over the 4-fold dosing range of 0.25–1.0 mg/kg with no evidence of toxicity based on body weight losses or decreased water consumption (data not shown). In contrast, complete tumor suppression by paclitaxel with this dosing regimen is only seen between 15–25 mg/kg, with 10 mg/kg inducing only partial inhibition (data not shown); the therapeutic window for paclitaxel in this model is thus just 1.7-fold.

The effects of 0.125 and 0.5 mg/kg ER-076349 and ER-086526 were also examined in the COLO 205 colon cancer xenograft model (Fig. 7, C and D) ⇓ . As in the MDA-MB-435 study, ER-086526 showed superior efficacy, with frank regression and long-term suppression of tumor regrowth at 0.5 mg/kg and small but measurable decreases in growth rates at 0.125 mg/kg. Suppression with 0.5 mg/kg ER-086526 was superior to that of paclitaxel at its empirically determined MTD of 20 mg/kg, both in rapidity of regression onset and in duration of regrowth suppression. In contrast, inhibition by ER-076349 was less pronounced at these dosing levels; no measurable effects were seen at 0.125 mg/kg, and inhibition at 0.5 mg/kg was less than that with 20 mg/kg paclitaxel. As in the MDA-MB-435 study, there was no evidence of toxicity based on body weights or water consumption (data not shown).

Evaluation of 0.1–1.0 mg/kg ER-076349 and 0.05–0.5 mg/kg ER-086526 in the LOX melanoma model also showed potent antitumor effects of both compounds (Fig. 7, E and F) ⇓ . With ER-076349, all of the doses led to >90% tumor suppression by day 17, three days after cessation of dosing. The three top doses, 0.25, 0.5, and 1 mg/kg, led to virtually complete tumor suppression up to day 17, with inhibition of regrowth persisting well beyond that seen with 12.5 mg/kg paclitaxel, its empirically determined MTD with this regimen. Again, complete tumor suppression by 0.25–1.0 mg/kg ER-076349 in the LOX model represents an unusually wide 4-fold therapeutic window. In contrast, doubling the paclitaxel dose to 25 mg/kg in this model is lethal, whereas halving it to 6.25 mg/kg leads to only minor inhibition (data not shown); the therapeutic window for paclitaxel with this dosing regimen is thus <2-fold.

Inhibition of LOX tumor growth by ER-086526 was similar to that with ER-076349 but with about 2-fold greater potency. Thus, 0.05 mg/kg ER-086526 inhibited tumor growth by 78% at day 17, with higher doses of 0.1, 0.25, and 0.5 mg/kg ER-086526 leading to complete tumor suppression. Again, this represents an unusually wide 5-fold therapeutic window. Tumor regrowth rates in the 0.5 mg/kg ER-086526 group were delayed significantly beyond the 12.5 mg/kg paclitaxel group; regrowth in the 0.25 mg/kg ER-086526 group was slightly delayed relative to paclitaxel. Significantly, one and three mice in the 0.25 and 0.5 mg/kg ER-086526 groups (n = 10), respectively, became tumor-free by day 17; all of the mice had measurable tumors on day 10. These cured mice remained completely tumor-free for an additional 7 months. Other than a 2% decrease in body weights in the 0.5 mg/kg ER-086526 group during the first five daily injections, there were no other indications of toxicity in any group with either compound, as assessed by body weights or water consumption (data not shown). The slight decrease in body weights in the latter group resolved itself before the beginning of the second cycle of five daily injections and was not observed again.

Finally, effects of 0.125–1.0 mg/kg ER-076349 and ER-086526 on NIH:OVCAR-3 ovarian cancer xenograft growth were examined (Fig. 7, G and H) ⇓ . During actual dosing periods, both compounds inhibited tumor growth to about the same degree, with 0.5 and 1 mg/kg leading to inhibition during dosing equivalent to paclitaxel at its empirically determined MTD of 20 mg/kg. Two of six mice each in the 1 mg/kg ER-076349 and ER-086526 groups and six of six mice in the paclitaxel group became tumor-free by the end of dosing; these mice remained tumor-free through the end of the experiment (day 89). After cessation of dosing, increased efficacy of ER-086526 over ER-076349 became evident in the 0.5 and 1.0 mg/kg groups; no dose of ER-076349 achieved the long-term complete regrowth suppression seen with paclitaxel, whereas 1 mg/kg ER-086526 achieved complete suppression out to day 89, similar to paclitaxel. Significant body weight losses during dosing were only seen in the 20 mg/kg paclitaxel and 1 mg/kg ER-086526 groups (12 and 18% decreases, respectively); body weights in these groups recovered within about 10 days after cessation of dosing (data not shown).

DISCUSSION

In the 15 years since its discovery, the extraordinary in vitro and in vivo anticancer potency of halichondrin B (2 , 10) has led to considerable interest in developing this agent as a new anticancer drug. Indeed, work on this compound by the NCI was ongoing as late as 1999 (28) . Nevertheless, the extremely limited supply of halichondrin B from marine sources has all but eliminated the possibility that this promising agent might become a viable drug candidate. Fortunately, the complete synthesis of halichondrin B (11) and the subsequent discovery that its biological activity resides in its macrocyclic lactone C1-C38 moiety 3 (see also Refs. 12, 13, 14 ) created the possibility of developing structurally simpler, fully synthetic analogues that retain the remarkable activity of halichondrin B. The resultant synthesis and evaluation of more than 180 macrocyclic analogues have culminated in the identification of ER-076349 and ER-086526.

ER-076349 and ER-086526 differ only in their C35 alcohol and amine substituents, respectively. Both showed sub-nm in vitro growth inhibition against a variety of human cancer cell lines (Table 1) ⇓ , with the alcohol ER-076349 being about 2-fold more potent than the amine ER-086526 in most lines. Both were roughly 5–10-fold more potent than vinblastine and paclitaxel, two tubulin-based antimitotic agents run as internal standards in each experiment.

Direct comparisons of ER-076349 and ER-086526 with halichondrin B were not performed in the current study because of the lack of availability of either synthetic or natural halichondrin B. However, the low- to sub-nm in vitro potencies measured for the two macrocyclic analogues were similar to historical halichondrin B data from our laboratory and others, indicating that the in vitro potency of halichondrin B had been retained during structural simplification. Thus, in an early study with synthetic halichondrin B, we measured a mean IC₅₀ of 0.4 nm against 17 human tumor cell lines representing eight different cancer types (24) , whereas other laboratories using natural halichondrin B have reported IC₅₀ values of 0.08 nm, 0.3 nm, and 5 nm against murine B16 melanoma, murine L1210 leukemia, and PtK1 normal kangaroo rat kidney cells, respectively (2 , 6 , 8) . Moreover, performance of ER-076349 and ER-086526 in the 60-cell line screen of NCI was essentially identical to that of natural halichondrin B 5 (data not shown). We conclude that virtually all of the growth inhibitory potency was retained during structural simplification, including removing the entire C36-C54 polyether “left half” of halichondrin B and replacing the macrocyclic lactone ester of the latter with a nonhydrolyzable ketone bridge.

Our results indicate that the growth inhibitory mechanisms of ER-076349 and ER-086526 are probably the same as those of halichondrin B. Thus, ER-076349 and ER-086526 induce G₂-M cell cycle arrest and disrupt mitotic spindles, similar to previous results with synthetic halichondrin B (24) and consistent with the tubulin-based antimitotic mechanism of the latter (6, 7, 8, 9) . Moreover, our studies with the biotinylated macrocyclic lactone analogue ER-040798 showed direct binding to tubulin under affinity chromatography conditions (Fig. 5) ⇓ . Although ER-040798 was prepared before discovery of the macrocyclic ketone series, direct tubulin binding by ER-040798 is probably representative of this entire class of drugs because simplification from halichondrin B to the macrocyclic ketones involved numerous incremental modifications made without significant changes in growth inhibitory potency, G₂-M arrest characteristics, or mitotic spindle disruption. Indeed, ER-076349 and ER-086526 directly inhibit tubulin polymerization in vitro with IC₅₀ values (5.7 μm and 6.9μ m, respectively) almost identical to the 7.2μ m value reported for halichondrin B (6) . We conclude that the anticancer activities of ER-076349 and ER-086526 derive from a tubulin-depolymerizing antimitotic mechanism similar or identical to that reported for halichondrin B.

ER-076349 and ER-086526 showed significant in vivo anticancer efficacy at doses well below 1 mg/kg in several human tumor xenograft models. Interestingly, despite somewhat lower in vitro potency in cell growth and tubulin polymerization studies, ER-086526 showed greater in vivo efficacy, particularly in the MDA-MB-435 breast and COLO 205 colon cancer models. One explanation for this might be that ER-076349 is metabolized in vivo more readily than ER-086526; pharmacokinetic studies are under way to investigate this possibility. Another explanation might relate to our finding that mitotic blocks induced by ER-086526 are much less reversible after drug washout than those induced by ER-076349 (data not shown). This could result in greater in vivo tumor cell killing under the intermittent dosing schedules used in our studies. This observation, which correlates with potential for in vivo efficacy within the halichondrin B macrocyclic analogue series, will be described in a separate manuscript. 6

In all of the four in vivo models, ER-076349 and ER-086526 were effective at much lower doses compared with paclitaxel run at empirically determined MTD levels. For instance, potency differentials for ER-086526 versus paclitaxel based on complete tumor suppression ranged from 20-fold in the NIH:OVCAR-3 model, to 40-, 50-, and 100-fold in the COLO 205, LOX, and MDA-MB-435 models, respectively. More importantly, ER-076349 and ER-086526 showed significantly wider in vivo therapeutic windows. For paclitaxel, empirically determined therapeutic windows were relatively small, 1.7 in the MDA-MB-435 model and <2.0 in the LOX model. In marked contrast, therapeutic windows for the macrocyclic ketone analogues were much wider. In the MDA-MB-435 model, ER-086526 showed a 4-fold therapeutic window, whereas in the LOX model, therapeutic windows of 4- and 5-fold were observed for ER-076349 and ER-086526, respectively. We speculate that the wide therapeutic windows seen with ER-076349 and ER-086526 contribute to their substantial in vivo efficacy, in that the ability to increase doses 4–5-fold above fully tumor-suppressive doses probably leads to more complete eradication of residual tumor cells. This might explain the superior in vivo efficacy of ER-086526 over paclitaxel MTD dosing in three of the four models tested and the existence of complete cures in the LOX model by ER-086526. The reasons for the unusually wide therapeutic windows of ER-076349 and ER-086526 are not currently known.

In summary, we have demonstrated highly potent in vitro and in vivo anticancer activities of two fully synthetic, macrocyclic ketone analogues of halichondrin B, ER-076349, and ER-086526. These agents exert their anticancer effects by mechanisms currently indistinguishable from the microtubule-destabilizing effects of halichondrin B. An unexpected but exciting finding was the existence of unusually wide in vivo therapeutic windows, which may contribute to the remarkable in vivo efficacy of these new agents. Our encouraging early results with ER-076349 and ER-086526 strongly support their continued development as novel anticancer agents for human use.