Chemists and phytochemists have always been impressed by the fact that compounds found in nature display an almost unbelievable range of diversity in terms of their structures and physical and biological properties. Most of these compounds are secondary metabolites whose functions in plants, fungi and marine organisms are still not widely understood. Currently, it is believed that many of these compounds act in defence of the harmful effects of toxins, carcinogens or mutagens found in the plant1,2 or attack by external predators.3
Phytochemical screening for cortisone precursors
During the period 1950–59, my colleague, the late Dr ME Wall, was the director of a large programme at the Eastern Regional Research Laboratory (ERRL), US Department of Agriculture (USDA) (Wyndmoor, PA, USA), which involved a screening study of thousands of plants, searching for steroids that would be cortisone precursors. The plant collections were conducted by botanists under the auspices of the Plant Introduction Division of the USDA. Thousands of plants were collected and meticulously identified before shipping to the ERRL. The survey included not only quantitative data for steroidal sapogenins, but also qualitative analysis for sterols, alkaloids, tannins and flavonoids.4–7 ERRL USDA saved thousands (but not all) of the plant alcoholic extracts, particularly of the more unusual plants.
One of the extracts so saved and stored was prepared from the leaves of Camptotheca acuminata, Nyssaceae, a tree which is native to China, growing in relatively warm areas of south-eastern provinces of China such as Szechwan, Yunnan and Kwangsi.8 This plant was introduced several times into the USA and eventually a few trees were growing at a USDA Plant Introduction Garden in Chico, CA, USA, from which this sample was procured.
Discovery of antitumour activity in extracts of C. acuminata
In 1957, after a visit by the late Dr Jonathan Hartwell of the Cancer Chemotherapy National Service Centre (CCNSC) (Bethesda, MD, USA), ERRL agreed to send CCNSC 1000 ethanolic plant extracts for testing for antitumour activity. Almost a year later the astonishing result came back that the C. acuminata extracts were the only ones to have high activity in the CA-755 assay then used as one of the standard test systems.
Discovery of camptothecin
Dr Wall became intensely interested in discovering the nature of the compound(s) responsible for the antitumour activity of C. acuminata extracts. For administrative reasons it was not possible to do this in the USDA and hence, in July 1960, Dr Wall left the USDA and established a natural products group at the Research Triangle Institute (RTI) with support from the National Cancer Institute (NCI). By 1963, a sizeable sample of approximately 20 kg of the wood and bark of the tree became available to the RTI, and in 1964 I joined Dr Wall’s group commencing a fruitful 40-year collaboration.
Use of L1210 assay for fractionation
We commenced a typical biodirected isolation in which the fractionation of the sample would be directed entirely by the antitumour assay. By this time it was known that the crude extracts of C. acuminata were very active in the L1210 mouse leukaemia life prolongation assay. This was, and still is, highly unusual. Most of the hundreds of plant extracts with which we and others subsequently worked were never sufficiently active in L1210 to conduct fractionations in this system. However, this was the case for C. acuminata and aroused very great interest on the part of NCI. In general, after conducting a particular phase of the isolation, it would often take us 3 months or more before the results were received. This was primarily because the L1210 assay is based on life prolongation. In the case of highly active extracts the study would have to be carried out for 30 days or more.
Fractionation of C. acuminata
Figure 1 shows the results obtained from the extraction of almost 20 kg of dry plant material, which consisted of the wood plus wood bark of the tree. Our procedure involved a continuous hot extraction with heptane. The residual plant material was extracted with hot 95% ethanol and, after concentration, the aqueous ethanolic residue was extracted with chloroform as shown in Figure 1. The fractions shown in Figure 1 were given L1210 assay. Only the chloroform fractions were active.
Craig countercurrent partition
Several procedures involving chromatography, particularly on alumina, were tested on a small scale. They were all unsuccessful for reasons unknown to us at the time, although we now know that camptothecin (CPT) is adsorbed very tightly on alumina, and that this is not an appropriate chromatographic agent for CPT. We then tested the Craig countercurrent partition procedure. This methodology was designed and brought to a high pitch of perfection by the late LC Craig, who received the Nobel Prize for his work.9,10 Various types of countercurrent equipment were available in the early 1960s. Most of this equipment has now become museum pieces, as more convenient and space-saving equipment is now available.
Most of the chloroform phase (see Figure 1) after concentration was subjected to an 11-stage preparative Craig countercurrent distribution as shown in Table 1. In this very simple use of the Craig methodology, the partition was carried out in large separatory funnels using a chloroform–carbon tetrachloride–methanol–water partition system with the ratios shown in Table 1. All of the fractions were analysed by both the in vivo L1210 mouse life prolongation assay and by a 9KB in vitro cytotoxicity. There was reasonable correlation between the two assays. Tubes 2–6 were judged to contain the most active material. It should be kept in mind that, in the in vivo life prolongation assay, it is the combination of the lowest dose with the largest treated over control (T/C) activity that is the basis for selecting active fractions. In the cytotoxicity assay, it is simply the lowest ED50 dose that inhibits the growth of the 9KB cells. We had thus removed about 80% of the total weight while concentrating the most active fractions in 20% of the original weight of the chloroform extracts. By comparison, with a pure sample obtained later but again given the standard L1210 assay (see Table 1, K-128), it can be estimated that the most active fractions contained between 1% and 2% of the active compound. It should be noted that 9KB activity was reasonably correlated with the L1210 assay.
|NSC number||Tube number||Weight of fraction (g)||Dose (mg/kg)a||T/C||9KB|
When fractions 2–6 were combined and the solvent partially concentrated, a yellow precipitate was formed and collected by filtration. This material was subsequently further purified by chromatography on a silica gel column and crystallization. As stated above, the pure compound is very active in L1210. Doses as low as 0.5 mg/kg gave an appreciable life prolongation. A dose of 4 mg/kg was the maximum range prior to occurrence of toxicity.
Structure of camptothecin
It was found that CPT can readily be converted to an acetate and a chloroacetate. The chloroacetate was converted to the corresponding iodoacetate by treatment with sodium iodide–acetone. The iodoacetate crystallized in orthorhombic crystals suitable for X-ray analysis. This was carried out by AT McPhail and GA Sim, then in the Department of Chemistry, University of Illinois (Urbana, IL, USA).11 The structure of CPT and the corresponding sodium salt is shown in Figure 2. This structure was determined completely in accord with ultraviolet and infrared spectra, nuclear magnetic resonance (NMR) spectra and mass spectral data. The structure is unique. CPT has been shown to be related to the indole alkaloids, and the six-membered ring (ring B in Figure 2) and a five-membered ring (ring C in Figure 2) are formed by a ring expansion/ring contraction sequence of reactions. The pentacyclic ring structure is highly unsaturated. Some of the unique structural features involve the presence in ring E (see Figure 2) of an α-hydroxy lactone system, and in ring D (see Figure 2) an unsaturated conjugated pyridone moiety. On treatment with alkali, the compound readily opens, forming an open lactone sodium salt (see Figure 2). On acidification, the extremely water-soluble sodium salt is relactonized readily. The parent compound, however, is extremely water insoluble and is insoluble in virtually all organic solvents except dimethylsulfoxide, in which it exhibits moderate solubility.
Biological activity of camptothecin and analogues
The isolation and structure proof for CPT and some of the analogues obtained at an early stage, 10-hydroxycamptothecin (3) and 10-methoxycamptothecin (4)12 enabled extensive studies to be conducted on CPT and a few of its analogues. As stated previously, CPT (1) was remarkably active in the life prolongation of mice treated with L1210 leukaemia cells, showing activity in doses between 0.5 and 4.0 mg/kg in this L1210 mouse life prolongation assay. It showed activity of a similar order in the life prolongation assay for P388 leukaemia. The compound was also very active in the inhibition of solid tumours that were being studied at this early stage, including the Walker muscular carcinosarcoma (WM) tumour, which was completely inhibited by CPT. 10-Methoxycamptothecin (4) was found to be active, but somewhat less active than CPT, whereas 10-hydroxycamptothecin (3) was the most active compound in the series and was more active than CPT in both L1210 and P388 leukaemia life prolongation assays.12–14 Unfortunately, 10-hydroxycamptothecin (3) is found in nature in only very small quantities, probably amounting to about 10% of the CPT content. All of these previous compounds were extremely insoluble in water, but it will be recalled that the lactone could be opened under mild conditions with sodium hydroxide or sodium methoxide and it was found that this sodium salt of CPT was very soluble in water. It was not until much later that it was shown by definitive studies that this compound was only one-tenth as active as CPT in the P388 assay.13
Early clinical trial
Encouraged by the broad scope of animal antitumour activity of CPT, a decision was made by NCI to go to clinical trials with the CPT sodium salt (2). In contrast to CPT, the sodium salt was water soluble and hence was easily formulated for intravenous administration. In a phase I trial, Gottlieb and Luce15 studied 18 patients and found that it exhibited a partial response. These responses were primarily in gastrointestinal tumours and of short duration. Toxicity that involved mainly dose-limiting haematological depression was noted, along with some vomiting and diarrhoea. In another phase I trial16 only two partial responses were found in 10 evaluable patients. Because of the somewhat encouraging results obtained in the phase I study,15 a phase II study was undertaken in 61 patients with adenocarcinomas of the gastrointestinal tract, but only two patients showed objective partial responses.17 The drug was also studied as a sodium salt in China. Up to 1000 patients were included in the trial and effective results were reported in gastric cancer, intestinal cancer, head and neck tumours and bladder carcinoma.18 These results were more promising than those in the US trial, but this may be because in the USA only patients who had been treated with many other drugs previously and had become resistant were put under test.
Although our work on CPT discovery may be considered to have nominally ceased by 1970, our interest in CPT and/or its analogues did not wane, particularly with regard to structure–activity relationship (Figure 3). In 1969, we reported the isolation of 10-hydroxycamptothecin (3) and 10-methoxycamptothecin (4) as minor components of C. acuminata.12 Later we found that 10-hydroxycamptothecin (3) was much more active than CPT in a number of assays, and this stimulated synthetic efforts by both the SmithKline Beecham group (now of GlaxoSmithKline) and a Japanese pharmaceutical company, Daiichi (Tokyo, Japan), to prepare water-soluble 10-hydroxycamptothecin analogues. CPT reacted readily with ethylamine to give the amide (5) with lower activity. Oxidation of the ring B nitrogen (see Figure 3) with chloroperbenzoic acid to give the N-oxide (6) led to considerable loss of activity. Of great interest was the fact that hydroxylation in ring A (see Figure 3), at least at the 10 position, was compatible with activity, and indeed this compound, 10-hydroxycamptothecin (3), has greater activity than CPT. Major reduction of antineoplastic activity was noted as a result of reactions involving the hydroxyl or lactone moiety in ring E (see Figure 2). After acetylation of CPT, the resulting acetate (7) is virtually inactive. Other reactions also point to the absolute requirement of the α-hydroxy group, as shown by the fact that the chloro compound (8), and the corresponding reduction product, desoxycamptothecin (9), are inactive in L1210 leukaemia.12 Reduction of the lactone under mild conditions to give the lactol (10) also results in complete loss of activity.12
Total synthesis of camptothecin and analogues
Although interest in clinical use of CPT lagged after the poor results obtained with CPT sodium, we became interested in the utilization of CPT analogues based on the knowledge that 10-hydroxycamptothecin was very active in L1210 mouse leukaemia. We developed a procedure with considerable versatility for the total synthesis of CPT and analogues.19–21 Although this procedure afforded only racemic compounds with the 20(RS) configuration, we subsequently developed a method to separate the 20(S) and 20(R) tricyclic enantiomers.22 Only the former was active.
Inhibition of the enzyme topoisomerase I
Interest in CPT and analogues remained at a low ebb until 1985, when it was discovered that CPT, by a unique mechanism, inhibited the enzyme topoisomerase I (TI).23 The enzyme affects DNA in various transactions such as replication, transcription and recombination. CPT and analogues bind to a complex formed by DNA with the TI enzyme. With the mechanism now understood, there opened the possibility for clinical use of CPT and analogues by virtue of the inhibition of TI, which might be found in cancer cells and hence would inhibit tumour growth.
Many new CPT analogues synthesized at RTI were tested for their TI inhibition. The inhibition of TI activity closely paralleled in vivo mouse leukaemia assays.24,25 Analogues of 20(S) and 20(R) have been tested, and we have shown that the 20(R) form is inactive, both in the topoisomerase inhibition and in the in vivo assays.24
Another study conducted cooperatively involved the Stehlin Institute (Houston, TX, USA), New York University Medical School (New York, NY, USA), Johns Hopkins University (Baltimore, MD, USA) and the RTI.26 In these studies, utilizing nude mice, both 9-aminocamptothecin (11) and 10,11-methylenedioxycamptothecin (12) analogues had great potency in inhibiting human colon cancer xenografts in nude mice.26 In these studies, a large number of compounds commonly used in cancer chemotherapy were totally ineffective.
The unique mode of action of CPT rekindled interest in the molecule in the mid-1980s and led to renewed efforts to develop soluble analogues of CPT.27 First-generation analogues of CPT, irinotecan (Campto®, Pfizer Ltd, New York, NY, USA) (13) and topotecan (Hycamtin®, GlaxoSmithKline, Brentford, UK) (14), which are water-soluble derivatives of CPT with an intact lactone ring, were approved for use by the US Food and Drug Administration (FDA) in 1996. These are marketed in the USA for the treatment of metastatic colorectal, primary colon and metastatic ovarian cancers.
Discovery of taxol
Since 1988 the remarkable clinical efficacy of taxol, resulting in numerous observations of partial and complete remission of advanced ovarian cancer in women, and more recently reports of the drug’s efficacy in breast, lung and prostate cancer, has aroused great interest, discovered at RTI many years ago.28 Like so many other investigations of this type, a combination of serendipity followed by much hard work led to the discovery of this very active antitumour agent.
Initial procurement and isolation
As mentioned before, a screening programme for antitumour agents in plants was initiated in 1960 under the leadership of Dr Jonathan L Hartwell at NCI. In this programme, plant samples collected at random were supplied by the USDA under an interagency agreement with NCI. In August 1962, USDA botanist Arthur S Barclay and three college student field assistants collected 650 plant samples in California, Washington and Oregon, including bark, twigs, leaves and fruit of Taxus brevifolia in Washington state.29
Taxus brevifolia is a slow-growing tree which is found only primarily localized in the coastal areas of the above-mentioned western coast states. It had never received any chemical investigation until it was assigned to our laboratory by Dr Hartwell. The assignment of the plant was not entirely serendipitous. Some of these samples had been shown to have 9KB cytotoxicity. At that time there were only three groups working under contract to NCI, in the laboratories of Dr Jack Cole, University of Arizona (Tucson, AZ, USA), the late Dr S Morris Kupchan at the University of Wisconsin (Madison, WI, USA) and our laboratory at RTI. At that time, the other groups were not particularly interested in plants with 9KB activity. We had noted an excellent correlation in our CPT studies between L1210 in vivo activity and the 9KB cytotoxicity. Accordingly, we had requested Dr Hartwell to assign as many 9KB active compounds as possible to us, and from this arose the assignment of T. brevifolia to RTI.
The method finally adopted after several unsuccessful trials is shown in Figures 4 and 5. It involved our standard ethanol extraction, partition of the ethanolic residue between water and chloroform, followed by a large number of Craig countercurrent distribution treatments, the last of which involved a 400-tube Craig countercurrent distribution (see Figure 5). In this manner, approximately 0.5 g of taxol was isolated starting with 12 kg of air-dried stem and bark. The yield was about 0.004%. All the various steps were monitored by an in vivo bioassay, which, at that time, involved the inhibition of the Walker WM solid tumour. As is shown in Figure 4, increased purification is accompanied by lower T/C and dose values. The isolation was carried out laboriously, but in a manner in which no losses as a result of the treatment and no changes in the unknown chemical constitution of the eventual product occurred because of the mild countercurrent distribution methodology. Much simpler procedures have been subsequently developed both at RTI and elsewhere.
Biological activity of crude and purified taxol
The name ‘taxol’ was assigned to this compound before we really knew its complete structure, but it was evident that it did contain some hydroxyl groups and the name had a nice ring to it. Apart from the actual isolation of the pure material, the crude extracts were subjected to a large number of tests. In early work we found that the crude extracts were active not only in the Walker tumour inhibition assay, but also had shown modest activity in L1210 leukaemia and particularly high activity in the 1534 (P4) leukaemia assay. The latter assay is a life prolongation assay in mice, and it had been used some years before by scientists at Eli Lilly (Indianapolis, IN, USA) to isolate the vinca alkaloids, showing very great activity in life prolongation in this system. The same was noted by us for taxol with T/C values in the P4 system in excess of 300, even with crude extracts. The activity of pure taxol in a number of in vivo rodent assays is shown in Table 2. Particularly high activity was shown in the B-16 melanoma assay. Years later this was one criterion responsible for placing taxol in clinical trial.30
i.p., intraperitoneal; s.c., subcutaneous; SRC, subrenal capsule.
The structure determination of taxol
As soon as we had isolated taxol in pure form, a great deal of work on the structure of the compound was carried out by available spectroscopic methods. Although methods for ultraviolet, infrared and mass spectrometry were at a reasonably advanced stage in the late 1960s, NMR was relatively primitive compared with the very sophisticated instrumentation now available.
It was evident by this time that taxol probably contained a taxane skeleton. A number of taxane derivatives had been previously reported and it was evident that taxol was much more complex as its molecular formula from high-resolution mass spectrometry is C47H51NO14, corresponding to a molecular weight of 853. The evidence then indicated that taxol comprised a taxane nucleus to which an ester was attached, as preliminary experiments indicated that an ester moiety was easily cleaved from the rest of the molecule. Attempts were made to make crystalline halogenated derivatives of taxol. However, none had properties suitable for X-ray analysis.
Taxol was therefore subjected to a mild base-catalysed methanolysis at 0°C, which yielded a nitrogen-containing α-hydroxy methyl ester C17H17NO4, a tetraol C29H36O10, and methyl acetate. Full details are presented in our article dealing with the structure of taxol.28 The methyl ester thus obtained by the mild methanolysis procedure was converted to a parabromobenzoate ester and characterized by the X-ray analysis as C24H20BrNO5. The tetraol produced from the methanolysis of taxol was converted to a bis(iodo)acetate, C33H38O12I2, which again received X-ray analysis.28
Because the ester could have come originally from hydroxyl groups either at C7 or C13 (Figure 6), it was necessary to establish by a few chemical steps on which of these hydroxyl moieties the ester had originally been located. As reported in detail in our article,28 the esterified position was found to be the allylic C13 hydroxyl moiety. Consequently, taxol has the structure as shown in Figure 6.
Prior to the isolation of taxol, no natural taxane derivative was reported to have antitumour activity. There are many interesting features of the molecule, particularly the four-membered oxide ring, which is not found in any of the other natural taxanes. The ester moiety itself is of interest, containing two phenyl groups, one of which is attached as part of an amide function. The large number of asymmetrical carbon atoms renders total synthesis of taxol most difficult.
For antitumour activity, it is essential that the entire taxane molecule be present. We have shown that the ester and the tetraol formed by low-temperature cleavage of taxol are each essentially inactive.
Our initial discovery of taxol essentially ended with the publication of our article on the structure.28 We made repeated efforts to interest the NCI administrators who were then involved in this programme to obtain larger amounts of the bark and wood of T. brevifolia so that advanced animal studies, toxicology and eventually clinical trials could ensue. The response was that the compound was present in too limited a quantity, that the extraction and isolation was difficult and that the supply of the tree was limited.
Fortunately, two major developments occurred. It was noted years later that taxol had very high activity in the B-16 melanoma assay.30 Then during the next several years, studies by Dr Susan Horwitz at the Albert Einstein Medical Center (Bronx, NY, USA) showed that the mode of activity of taxol was due to its binding with tubulin in a novel manner.31
For some time there had been some interest in the mechanism of action of taxol. Was the unique structure accompanied by an unusual mechanism of action? Initial studies by Fuchs and Johnson32 indicated that taxol inhibited proliferation at the G2–M phase in the cell cycle and blocked mitosis. Thus, it appeared to be one in a series of naturally occurring spindle poisons such as vincristine and vinblastine, colchicine, podophyllotoxin, maytansine and others. The finding that taxol was a spindle poison was not necessarily encouraging. Shortly thereafter a more detailed investigation by the Horwitz group31 established that, while taxol was a mitotic inhibitor, the mechanism was unique in that it stabilized microtubules and inhibited depolymerization back to tubulin; this was the opposite effect of the other antimitotic agents cited above, which all bind to soluble tubulin and inhibit the polymerization of tubulin to form microtubules.30,33,34 This information was important in making the argument that, by virtue of uniqueness of both structure and mechanism, taxol was a worthy candidate for development.
Our early observations that taxol extracts and pure taxol were highly active in leukaemia P1534 (P-4) assay now became explicable. Tubulin is a protein involved in the process of mitosis. The vinca alkaloids were noteworthy for their activity against P-4 leukaemia. Taxol is also highly active against P-4 leukaemia. Both compounds bind to tubulin but, as described above, by completely different mechanisms.
Era of rapid progress 1982–94
Rapid progress both in chemical synthesis of taxol and in clinical developments occurred during 1982–94. The events in this decade have been covered in detail by a number of reviews30,35,36 so this section will present only a brief outline of the tremendous progress and extensive research conducted during this period.
Animal toxicology and formulation were in place by 1982–84 and clinical phase I and II trials were conducted over the period 1983–86. Great interest was generated in taxol, both in scientific circles and among the general public, by the announcement of the remarkable efficacy of taxol against ovarian cancer.37,38 Subsequently, taxol (15) and its analogue docetaxel (Taxotere®, Sanofi-Aventis Ltd, Guilford, UK) (16) were studied for efficacy against many solid tumours (e.g. breast and lung).30,35
In an effort to obtain adequate supplies of taxol, the NCI issued a Cooperative Research and Development Agreement (CRADA), which was open to competition. The CRADA was then issued to Bristol-Myers Squibb in 1991. The company moved rapidly to obtain FDA approval for the marketing and a New Drug Application (NDA) was filed in July 1992 and approved by the FDA in a record time of 6 months in December 1992.
Analogues of taxol
Taxol (15) continues to be modified for improved activity and its discovery has led to the development of two second-generation analogues, docetaxel (16) and cabazitaxel (Jevtana®, Sanofi-Aventis Ltd, Guilford, UK) (17), which are FDA approved for cancer treatment. Docetaxel has been a very important drug for the breast cancer and cabazitaxel (17) is FDA approved for the treatment of hormone-refractory prostate cancer. In another therapeutic development, a solvent-free, nanoparticle-based delivery system for taxol (paclitaxel; Abraxane®, Celgene Ltd, Uxbridge, UK) bypasses the need for premedication to prevent hypersensitivity reactions to the vehicle, cremophor EL, used in the first-generation taxol formulations. Last year, the FDA approved the supplemental NDA of paclitaxel as first-line treatment for patients with metastatic adenocarcinoma of the pancreas, in combination with gemcitabine. Adenocarcinoma, a subtype of exocrine tumours, accounts for approximately 95% of cancers of the pancreas.
The discoveries described in this article represent several contributions for the treatment of cancer. Most significantly, taxol and CPT analogues have been responsible for saving the lives of hundreds of thousands of people afflicted with cancer. Since the inclusion of taxol in the treatment of ovarian cancer, the survival rate has more than doubled. Furthermore, both taxol and CPT were found to inhibit cancer cell growth via novel mechanisms of action. Prior to their discovery, neither the stabilization of microtubule assembly nor the trapping of T1-DNA intermediate was known to be an effective way to circumvent the uncontrolled growth of cancer cells. Thus, these compounds have led to the identification of new cancer drug targets.
To me, it is a matter of tremendous personal satisfaction to be associated with discovery of two of the most important anticancer drugs that are benefiting humanity.