Cell-based Screen for Antimitotic Agents and Identification of Analogues of Rhizoxin, Eleutherobin, and Paclitaxel in Natural Extracts¹

Antimitotic agents are compounds that arrest cells in mitosis. Several are clinically important anticancer drugs, including the Vinca alkaloids vinblastine, vincristine, and vinorelbine(1) and the taxanes paclitaxel and docetaxel(2). They cause mitotic arrest by interfering with the assembly or disassembly of α- and β-tubulin into microtubules. At high concentrations, the Vinca alkaloids and most other antimitotics cause complete microtubule depolymerization, whereas the taxanes cause bundling of microtubules by stabilizing them against depolymerization. At low concentrations, neither depolymerization nor bundling is observed, but there is sufficient alteration in the dynamics of tubulin loss or addition at the ends of mitotic spindle microtubules to prevent the spindle from carrying out its function of attaching to and segregating the chromosomes, and cells arrest in mitosis (3, 4). Prolonged arrest eventually leads to cell death, either in mitosis or after an eventual escape from mitotic arrest (5, 6). Another class of antimitotic agents,represented by estramustine, does bind tubulin (7) but may also bind microtubule-associated proteins and prevent them from regulating interactions between tubulin polymers (8). Agents that are not known to interact with microtubules, such as inhibitors of protein phosphatases 1 and 2A and mitotic kinesin inhibitors, can also arrest cells in mitosis (9, 10, 11).

The Vinca alkaloids were isolated from the periwinkle plant,which originally attracted attention because of reported hypoglycemic properties. However, periwinkle extracts showed no antidiabetic action but were found to prolong the life of mice bearing a transplantable lymphocytic leukemia (1). This led to the identification of vincristine and vinblastine. Paclitaxel was isolated from the bark of the Pacific yew tree, an extract of which showed antineoplastic activity in the NCI³large-scale screen (2). Vinorelbine and docetaxel are semisynthetic analogues.

These drugs, although extremely valuable, are not ideal. They have numerous toxicities, principally myelosuppression and neurotoxicity. More importantly, many cancers are inherently resistant to these drugs or become so during prolonged treatment (1, 2). This is often the result of multidrug resistance caused by overexpression of P-glycoprotein, which functions as a drug efflux pump. Other sources of resistance include increased expression of tubulin isotypes to which a particular drug binds less effectively and alterations in α- andβ-tubulin structure, by mutation or posttranslational modification, that reduce binding.

Antimitotics with different chemical structures might show increased specificity to mitotic microtubules rather than neuronal microtubules and reduce unwanted side effects and might be effective against resistant cancers. Many other antimitotics have been discovered, some of which show promise in preclinical studies or have entered clinical trials (12). However, they were discovered either by serendipity or by cytotoxicity screening, or because they showed patterns of cytotoxic activity against panels of cancer cell lines similar to patterns shown by other antimitotic agents(13). The search for better antimitotics would be greatly aided by rational assays for use in drug screens.

We have developed a rapid and reliable cell-based screen for antimitotic agents. In this report, we describe the assay, its application to a screen of >24,000 natural extracts, and the purification and characterization of paclitaxel analogues and new rhizoxin and eleutherobin analogues.

Human breast carcinoma MCF-7 cells were cultured as monolayers(14). The cells were seeded at 10,000/well of 96-well polystyrene tissue culture plates (Falcon) in 100 μl of medium and were allowed to grow overnight. Crude extracts were then added at about 10 or 1 μg/ml from 1000-fold stocks in DMSO. Untreated samples received an equivalent amount of DMSO and served as negative controls. Cells treated with 100 ng/ml nocodazole (Sigma), from a 1000-fold stock in DMSO, served as positive controls. Cells were incubated for 16–20 h. The relative number of cells in mitosis was then determined by microscopy (14), by ELISA, or by ELICA (see below).

After incubation with extracts, the cell culture medium was withdrawn carefully using a pipettor. This did not result in any loss of the rounded-up mitotic cells, which remained attached to the plates. The cells were lysed by adding 100 μl of ice-cold lysis buffer (1 mm EGTA, pH 7.4, 0.5 mm phenylmethylsulfonyl fluoride) and by pipetting up-and-down 10 times. The cell lysates were transferred to 96-well PolySorp plates (Nunc) and dried completely in a stream of air at about 37°C from a hair dryer. Vacant protein binding sites were blocked by adding 200 μl/well of antibody buffer [10 mm Tris-HCl (pH 7.4), 150 mm NaCl, 0.1 mm phenylmethylsulfonyl fluoride, and 3% (w/v) dried nonfat milk (Carnation)] for 1 h at room temperature. This was removed and replaced with 100 μl of antibody buffer containing 0.1–0.15 μg/ml TG-3 monoclonal antibody (15, 16). This antibody recognizes a phosphoepitope on nucleolin that is present only at mitosis and was provided by Dr. Peter Davies (Albert Einstein College of Medicine, Bronx, NY). After 16–20 h incubation at 4°C,the antibody solution was removed, and the wells were rinsed twice with 200 μl of 10 mm Tris-HCl (pH 7.4), 0.02% Tween 20. HRP-conjugated goat antimouse IgM secondary antibody (Southern Biotechnology Associates) was added at a dilution of 1:500. After overnight incubation at 4°C, the antibody solution was removed, and the wells were rinsed three times with 200 μl of 10 mmTris-HCl (pH 7.4), 0.02% Tween 20. One hundred μl of 120 mm Na₂HPO₄, 100 mm citric acid (pH 4.0) containing 0.5 μg/ml 2,2′-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) and 0.01%hydrogen peroxide was added for 1 h at room temperature, and absorbance at 405 nm was determined using a Dynex MRX plate reader.

After incubation with extracts, the medium was withdrawn carefully using a pipettor, and 100 μl of 10 mm Tris-HCl (pH 7.4),150 mm NaCl, containing 3.7% formaldehyde, were added to fix the cells for 30 min at 4°C. The fixative was removed and replaced with 100 μl of cold (−20°C) methanol for 5 min to permeabilize the fixed cells. The methanol was removed, and the wells were rinsed briefly with 200 μl of antibody buffer. Then, 100 μl of antibody buffer containing 0.1–0.15 μg/ml TG-3 monoclonal antibody and HRP-conjugated goat antimouse IgM secondary antibody at a dilution of 1:500 was added for 16–20 h at 4°C. The plates were washed twice with 200 μl of 10 mm Tris-HCl (pH 7.4), 0.02% Tween 20. One hundred μl of 120 mmNa₂HPO₄, 100 mmcitric acid (pH 4.0) containing 0.5 μg/ml 2,2′-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) and 0.01%hydrogen peroxide were added for 1 h at room temperature, and absorbance at 405 nm was measured. Additional information about this assay is provided in “Results.”

Approximately 250 g each of marine invertebrates were collected by hand, using scuba, from the cold temperate waters of the Pacific Ocean along the coast of British Columbia, from tropical Pacific Ocean reefs off Motupore and Madang in Papua New Guinea, and from tropical waters off the island of Dominica in the Caribbean. Samples were deep frozen on site and transported to Vancouver over dry ice. Voucher samples of each invertebrate are stored in methanol at −20°C at the University of British Columbia for taxonomic identification. Marine microorganisms were isolated from the invertebrates on site using various marine culture media, and pure cultures were grown as lawns on solid agar marine media in 10-cm Petri plates for several days and then freeze-dried.

Extracts of invertebrates were prepared by homogenizing in methanol∼200 g of each sample. The homogenates were filtered and concentrated to dryness in vacuo to give a gummy residue. Extracts of microorganisms were prepared by extracting the freeze-dried culture(cells and agar) multiple times with dry methanol:acetone, followed by lyophilization. A small amount of each extract was dissolved in DMSO for the antimitotic screen. Extracts of terrestrial plants were obtained from the Open Repository Program of the Natural Products Repository of the NCI Developmental Therapeutics Program as 500-μg samples that were dissolved in 100 μl of DMSO. All diluted extracts were stored at −20°C.

The TG-3 monoclonal antibody, originally described as a marker of Alzheimer’s disease (15), is highly specific for mitotic cells. Flow cytometry shows that TG-3 immunofluorescence is >50-fold more intense in mitotic cells than in interphase cells(16). In Western blots, the antibody reacts with a M_r 105,000 protein that is present in abundance in extracts of cells treated for 20 h with the antimitotic agent nocodazole but present at only low levels in extracts from cycling MCF-7 cells (Fig. 1). This protein has been identified as a mitotically phosphorylated form of nucleolin (17). Densitometric scanning of the bands showed a 27-fold difference in intensity between nocodazole-treated and untreated cells, corresponding well to the difference in the number of mitotic cells in the two samples: 80% for the nocodazole-treated sample and 3% for the untreated sample, as measured by microscopy.

TG-3 also recognizes mitotic cells in ELISA using microtiter plates(18). In this standard assay (19), cells grown in 96-well plates are lysed, and the lysates are transferred to protein-binding ELISA plates for adsorption to the plastic surface. The antigen is detected by incubating with TG-3 antibody, followed by an HRP-conjugated secondary antibody and colorimetric determination of HRP activity.

We first tested the suitability of the ELISA for quantifying the activity of antimitotic agents. MCF-7 cells were incubated for 20 h with different concentrations of the antimitotic drug paclitaxel, and the proportion of cells arrested in mitosis was measured by counting mitotic figures in the microscope and by ELISA. Paclitaxel induced mitotic arrest in a concentration-dependent manner with half-maximal activity at 10 nm measured by microscopy (Fig. 2,A) and at 4 nm measured by ELISA (Fig. 2 B).

ELISA is a lengthy and labor-intensive procedure requiring the preparation of cell lysates, their transfer to protein-binding plates,and many solution changes. We subsequently simplified it, reducing the time of the procedure and the number of steps by half and avoiding transfer of samples to ELISA plates. In this procedure, the cells are fixed with formaldehyde in their microtiter culture plate and permeabilized with methanol and detergents, and the TG-3 primary antibody and HRP-conjugated secondary antibody are added simultaneously. Colorimetric detection of HRP activity remains unchanged. Because cell fixation and permeabilization in situ are steps commonly used in immunocytochemistry, we termed the assay ELICA.

The ELICA was tested as above. Dose-dependent arrest of cells in mitosis by paclitaxel was detected by ELICA with half-maximal activity at 1.5 nm (Fig. 2). The ELICA showed a higher signal at low paclitaxel concentrations and a lower signal at high concentrations than did the ELISA (Fig. 2 B). These differences probably resulted from higher nonspecific staining of interphase cells because of reduced washing and from lower specific staining of mitotic cells because of fixation and reduced antibody incubation times. Nevertheless, the ELICA consistently showed sufficient difference in absorbance between cells treated or not with antimitotic agents to allow unambiguous detection of mitotic cells. Measurements obtained by ELICA consistently showed smaller SDs than obtained by ELISA, probably because the reduced number of manipulations reduced experimental variation.

We first tested the suitability of the ELISA for drug screening using a small selection of crude extracts from marine microorganisms (Table 1). Of the 264 extracts tested, 261 showed no activity, giving absorbance readings not statistically different from those of untreated cells. Three extracts clearly showed activity, with absorbance readings of 1.135, 1.437, and 1.245, close to the values obtained with nocodazole as a positive control.

We then screened over 2000 crude extracts of marine sponges, tunicates,soft corals, starfish, and nudibranchs. This screen identified 16 additional extracts with antimitotic activity. The positive extracts were retested by counting mitotic cells in the microscope, and all were confirmed to arrest cells in mitosis.

Finally, we screened crude extracts of terrestrial plants from the NCI Natural Products Repository by ELICA. The suitability of the ELICA for drug screening is illustrated in Table 2, which displays a screen of 264 plant extracts from three randomly selected 96-well plates. Five extracts showed activity, with absorbance readings close to or higher than those obtained with nocodazole. These positive readings were well above those obtained with negative controls or extracts showing no activity. Of 21,600 plant extracts tested in this manner, 100 showed activity, all of which were confirmed to be positive by microscopy.

All positive extracts from marine organisms and most positive extracts from plants were then rescreened using tubulin immunofluorescence microscopy (14) to examine their effects on microtubule structure. We then purified and identified the active agents in the three microbial extracts, in the single marine invertebrate extract that produced paclitaxel-like bundling of microtubules, and in the terrestrial plant extract that showed clearest evidence of microtubule bundling. The other extracts remain to be studied.

Marine bacterial isolate MK7020 collected off the coast of British Columbia was identified as a Pseudomonas sp. by gas chromatographic analysis of cellular fatty acids. The active compounds 1 and 2 (Fig. 3) were purified by chromatographic procedures using the ELISA to guide fractionation. The two other microbial extracts were found to be independent isolates of the same Pseudomonas species and contained the same active compounds as MK7020.

Compound 1 is identical to WF-1360C (20, 21), a previously reported analogue of the antimitotic agent rhizoxin (Fig. 3). Compound 1 showed half-maximal antimitotic activity(IC₅₀) at 52 nm, as determined by ELISA (data not shown). Compound 2 is a new δ-lactone seco-hydroxy acid and had an IC₅₀ of 8 nm (data not shown).

An extract of octocoral Erythropodium cf. caribaeorumcollected from shallow reefs near Dominica showed antimitotic activity and bundling of microtubules. The active compounds 3–10(Fig. 4) were isolated, and their chemical structure was elucidated as described in detail elsewhere (22).

Compound 3 was identified as eleutherobin, a recently discovered antimitotic agent that acts like paclitaxel by stabilizing microtubules (23, 24). Compound 4 was identified as sarcodictyin A (25) and differs from eleutherobin by replacement of the C-15 β-linked 2′-O-acetyl-d-arabinopyranose side chain of 3 with a methyl ester and replacement of the C-4 methoxyl with a hydroxyl group. Compounds 5–10 have not been reported previously. Desacetyleleutherobin (5) retains the arabinose, but not the 2′ acetyl substituent. Isoeleutherobin A(6) has an acetyl group at the 3′ position instead of the 2′position. Z-Eleutherobin (7) is a geometric isomer of eleutherobin at the C-2′ to C-3′ double bond of the C-8 N-(6′)-methylurocanic acid ester side chain. Desmethyleleutherobin (8) differs from eleutherobin by the presence of a hydroxyl instead of a methoxyl at C-4. Caribaeoside(9) differs from eleutherobin by the presence of a hydroxyl at C-11 of the tricyclic core, and a double bond at C-12 to C-13 instead of C-11 to C-12, significantly altering the cyclohexene ring. Caribaeolin (10) differs from caribaeoside only by the presence of a –CH₂OCO-CH₃substituent in the C-3 side chain.

The antimitotic activity profile of these compounds determined by ELICA is shown in Fig. 5. Eleutherobin (3) had an IC₅₀ of 100 nm. The activity of Z-eleutherobin(7) was close, with an IC₅₀ of 250 nm. Desmethyleleutherobin (8) and isoeleutherobin A (6) were slightly more potent than eleutherobin, with IC₅₀ of 20 and 50 nm, respectively. Desacetyleleutherobin(5) was slightly less potent, with an IC₅₀ of 400 nm. Sarcodictyin A (4) showed lower activity, with an IC₅₀ of 2 μm. Caribaeoside (9) and caribaeolin(10 ) were considerably less potent, with an IC₅₀ of 20 μm for both compounds.

NCI Natural Products Repository extract N29701 was obtained from the stem bark of the tree Ilex macrophylla in Kalimantan,Indonesia. It showed antimitotic activity and caused bundling of microtubules. The active compounds were isolated and analyzed using ELICA and identified as the known paclitaxel analogues 10-deacetylaxuyunnanine A (11) and 7-(β-xylosyl)-10-deacetyltaxol C (12) (Fig. 6) by analysis of their nuclear magnetic resonance data and comparison with published values (26, 27). Compounds 11(IC₅₀, 0.3 μm) and 12 (IC₅₀, 10μ m) were much less potent than paclitaxel(IC₅₀, 1.5 nm).

We have described a cell-based assay for antimitotic compounds. When searching for therapeutic agents, cell-based assays are particularly valuable compared with cell-free assays because they select not only for activity against a particular target but also for other desirable properties, such as the ability to permeate cells and to retain activity in tissue culture medium and in cells. In one study, >90% of compounds found on the basis of in vitro target-based assays showed no cytotoxic activity because they did not cross the plasma membrane or were degraded rapidly (28). In addition,assays based on measuring arrest of cells in mitosis have the potential to identify not only agents that interact with microtubules but also agents that cause mitotic arrest by other mechanisms, such as protein phosphatase inhibitors and mitotic kinesin inhibitors(9, 10, 11).

The ELISA and the ELICA procedures both allow unambiguous detection of antimitotic activity in crude natural extracts. The ELICA was used for most of the screening described here because it is faster, less labor-intensive, and less costly than the ELISA.

Our screen of over 24,000 crude extracts from different natural sources identified unambiguously 119 with antimitotic activity. The absence of false-positive results was confirmed by microscopy, and all five positive crude extracts that were subjected to further study yielded known or novel antimitotic agents; three extracts from the pilot screen contained members of the rhizoxin family, one marine invertebrate extract contained compounds related to eleutherobin, and a tree extract contained paclitaxel analogues.

The assay is useful not only for identifying and purifying antimitotics but also for providing a quantitative measure of their antimitotic activity. This is a helpful indicator of a compound’s pharmacological potential because it measures not simply the interaction of the compound with its target, as an in vitro assay would do, but its ability to interact with its target within a cell. We used it to compare the antimitotic activity of different analogues of rhizoxin,eleutherobin, and paclitaxel.

Rhizoxin is a 16-membered macrolide isolated in 1984 (29)and later found to cause the accumulation of cells in mitosis(30, 31) and to inhibit microtubule assembly (31, 32). Rhizoxin is very cytotoxic to cancer cells in vitro or in mice (20, 30), including cell lines resistant to the Vinca alkaloids (30). It has been the subject of several Phase I and II clinical trials, but results have been disappointing (reviewed in Ref. 33). To the best of our knowledge, the seco-hydroxy acid 2 was not known previously as a natural product, having been reported in the patent literature only as a semisynthetic derivative of the correspondingδ-lactone. WF-1360C (1) was 15-fold less toxic to P388 cells than rhizoxin (20). It differs from rhizoxin by the presence of a hydroxyl group instead of a methoxyl at C-17 and the absence of the two epoxides at C-2 to C-3 and C-11 to C-12. Compound 2 retains the methoxyl and one epoxide but has an open lactone ring. Comparison of the antimitotic activity of WF-1360C(IC₅₀, 52 nm) to that of compound 2 (IC₅₀, 8 nm) and to published cytotoxicity data for other analogues (20, 32) indicates that a closed lactone ring is not required for antimitotic activity and that the presence of a methoxyl substituent at C-17 contributes to the high potency of rhizoxin.

Eleutherobin was identified as a compound with paclitaxel-like properties in 1997 (23), but sarcodictyins A-D were the first members of the eleutherobin class of compounds to be identified(25, 34), their paclitaxel-like properties were recognized only later (35). Sarcodictyin A (4) was 20-fold less active than eleutherobin (3), indicating that the C-15β-linked 2′-O-acetyl-d-arabinopyranose side chain or the C-4 methoxyl group is important for antimitotic activity. Desmethyleleutherobin (8) was active, showing that it is the C-15 side chain and not the C-4 methoxyl that is required. Desacetyleleutherobin (5) and isoeleutherobin A(6) showed activity similar to eleutherobin, indicating that the acetyl group does not contribute importantly to activity. Therefore, although the sugar moiety is not absolutely required for antimitotic activity, it contributes to the high potency of eleutherobin.

Isomerization of the C-2′ to C-3′ double bond of the C-8 side chain of Z-eleutherobin (7) had little effect on the antimitotic activity of the compound, showing that the Econfiguration in eleutherobin is not required for antimitotic activity. Desmethyleleutherobin (8) was the most active of the compounds tested, suggesting that the C-4 hydroxyl might enhance activity through additional hydrogen bonding, or that the C-4 methoxyl somehow hinders the activity of eleutherobin. Caribaeoside(9) was 200-fold less active than eleutherobin, revealing the importance of the C-11 to C-13 segment for antimitotic activity. Caribaeolin (10) differs from caribaeoside (9)only in the C-3 side chain, and the activities of these compounds are similar. Likewise, sarcodictyin A differs from desmethyleleutherobin only in the C-3 side chain, but its activity is lower than that of desmethyleleutherobin. These data indicate that the C-15 acetyl-d-arabinopyranose can be replaced with an acetoxy functionality without significant loss of activity, confirming earlier data with synthetic analogues (36, 37), but not with a methyl ester.

Thirteen synthetic eleutherobin analogues have recently been described and tested in tubulin polymerization and cytotoxicity assays(36, 37, 38). Overall, these studies underlined the importance of the C-8 and C-3 side chains for activity, the C-8 side chain being essential and the sugar or another bulky substituent being needed at C-3 for optimal activity. All of the synthetic analogues retained the original eleutherobin core and therefore provided no information about the importance of segments of the tricyclic core.

Eleutherobin represents one of five chemical structural types known to arrest cells in mitosis by stabilizing microtubules. The other four are paclitaxel, discodermolide, the epothilones, and the laulimalides(39, 40, 41). Several pharmacophores have been proposed for members of this group (42, 43, 44). The latter(44) included eleutherobin and proposed three regions of common overlap between the chemotypes, shown as boxes A, B and C in Fig. 4. Region A of eleutherobin consists of the C-8 side chain, region B encompasses the C-11 to C-13 segment of the tricyclic skeleton, and region C consists of the C-15 substituent. The importance of regions A and C is supported by the published structure-activity data for eleutherobin analogues mentioned above (36, 37, 38). Our demonstration that caribaeoside (9), which differs from eleutherobin only in region B, shows a 200-fold lower activity demonstrates an important role for this region in antimitotic activity. Further studies will be required to determine whether the reduced antimitotic activity of caribaeoside is attributable to reduced affinity for tubulin and microtubules or to factors such as drug uptake, extrusion, or metabolism.

Paclitaxel is an approved drug for the treatment of advanced ovarian cancer and metastatic breast cancer. It was originally isolated from Taxus brevifolia in 1971 (45). Since then, over 350 related diterpenoids have been isolated from different species of the genus Taxus (46), including compounds 11 and 12 described here. Compound 11differs from paclitaxel in the nature of the N-acyl substituent on the C-13 phenylisoserine side chain and in the absence of the acetyl substituent at C-10. It was less active than paclitaxel,showing that the C-13 and C-10 substituents, although not essential for activity, contribute to the high potency of paclitaxel. Compound 12 further differs from paclitaxel by the presence of aβ-xylosyl substituent at C-7. Compound 12 was less active than compound 11, indicating that the C-7 substituent also contributes to the potency of paclitaxel.

An unexpected outcome of this study is that although the active compounds we isolated belong to known antimitotic chemotypes, they were found in organisms not known or suspected to produce them. To our knowledge, rhizoxin compounds have previously only been isolated from the rice seedling blight fungus Rhizopus chinensis and unidentified species of the same genus (29). We have now identified rhizoxin analogues in marine bacterial isolates of the genus Pseudomonas, which is common in Pacific Northwest waters. Eleutherobin was originally isolated from the soft coral Eleutherobia sp. (possibly E. albiflora)collected in Western Australia (23). We now identify eleutherobin in the Caribbean octocoral Erythropodium caribaeorum. This is of practical significance because it has not been possible to obtain sufficient amounts of natural or synthetic eleutherobin for preclinical development (47). The taxonomic classification of this source was confirmed by the identification of large quantities of the erythrolide diterpenoids characteristic of this species (48). In contrast to Eleutherobia, E. caribaeorum is widespread in the Caribbean and Florida (49, 50, 51), abundant in certain areas,and has been grown in aquaria. It may thus constitute a suitable source of eleutherobin for preclinical and early phase clinical trials. Paclitaxel and analogues have all been isolated from the bark of yew trees (46), from endophytic fungi isolated from the Taxus species or Taxodium distichum (52, 53), and recently from an epiphytic fungus on the rubiaceous plant Maguireothamnus speciosus (54). It was surprising to find paclitaxel analogues in the bark of a non-Taxus tree. The taxonomic classification of our extract was confirmed by the presence of the triterpenoid glycosides characteristic of the genus Ilex (55). It is possible that an endophytic fungus is responsible for their production in Ilex macrophylla.

Perhaps the most important outcome of this study is that the assay permitted us to detect antimitotic agents in extracts that were not found to contain them using other methods. E. caribaeorumhas been subjected to extensive chemical characterization(48), but eleutherobin compounds were not detected because they are very minor components. The COMPARE algorithm, which is able to detect similar differential patterns of growth inhibition for the 60 human cell lines in the NCI anticancer drug screen, has been used successfully to identify new antimitotic agents within the NCI chemical repository of pure compounds (13). Extracts in the NCI Natural Products Repository have also been tested against the NCI cell line panel. A COMPARE analysis using paclitaxel as the probe compound identified 47 plant extracts with a Pearson correlation coefficient above 0.6 (not shown). Three of these extracts were positive in our antimitotic screen, and all three were from Catharanthus roseus, the plant from which the Vinca alkaloids were originally isolated (1). The analysis did not identify the extract from Ilex macrophylla, the growth inhibition pattern of which does not resemble that of paclitaxel and other antimitotic compounds. This illustrates the usefulness of the cell-based assay for the identification of active compounds present in very low abundance in crude natural extracts. The assay should greatly facilitate the discovery and development of novel antimitotic agents and their characterization in the context of living cells.

We thank Peter Davies for providing TG-3 antibody; Hans Behrisch and Ross University for logistic support in the Commonwealth of Dominica; and Michael Leblanc, David Williams, and Robert Britton for collecting E. caribaeorum.