Scaled-up Synthesis of Discodermolide

Some 60 g of (+)-discodermolide, a potent inhibitor of tumor cell growth, has been prepared in a 39-step synthesis by the Novartis Chemical & Analytical Development Group in Basel, Switzerland. The synthetic material is now undergoing Phase I clinical trials for pancreatic cancer at the Cancer Therapy & Research Center in San Antonio, Texas.

"The large-scale total synthesis of such a complex natural product in such quantities was a first for Novartis and probably the entire pharmaceutical industry," says Novartis principal scientist Stuart J. Mickel, who played a significant part in the large-scale synthesis effort.

The synthesis is spectacular, according to Steven V. Ley, chemistry professor at the University of Cambridge, England. "It's probably the best piece of synthetic work to come out from an industrial company," he comments. "The ability to make something at this level of complexity as opposed to extracting it from natural product sources illustrates the power of modern synthetic chemistry." 

(+)-Discodermolide is a novel polyketide lactone natural product that was first isolated in 1990 from extracts of the rare Caribbean marine sponge Discodermia dissoluta by chemistry group leader Sarath P. Gunasekera and coworkers at Harbor Branch Oceanographic Institution, Fort Pierce, Fla. [J. Org. Chem., 55, 4912 (1990)].

"The sponge that yielded the first sample of (+)-discodermolide was collected by hand using scuba at Lucay, Grand Bahamas Island, at a depth of 33 meters," Gunasekera says. "The gross structure of (+)-discodermolide was determined by extensive NMR studies, and the relative stereochemistry was assigned by single-crystal X-ray crystallography."

The molecule's 24-member polyketide carbon skeleton is made up of eight polypropionate and four acetate units. The structure has 13 stereocenters, lactone and carbamate moieties, and three Z-configured alkenes, one of which is part of a terminal diene unit. It also features a stereo triad--methyl, hydroxyl, methyl--that is repeated three times.

THE COMPOUND has been shown to inhibit the proliferation of human and mouse cells by arresting the gap 2 (G2) and mitosis (M) stages of the cell cycle. It does so by binding to and stabilizing microtubules--hollow filaments consisting of tubulin protein subunits. Microtubules play an important role in cell division.

(+)-Discodermolide is one of a small, but structurally diverse, collection of naturally occurring microtubule-stabilizing agents discovered over the past decade. They include epothilones, eleutherobin, and laulimalide.

"The compound belongs to the class of antimitotic agents known to act by microtubule stabilization whose clinically used member is Taxol," Gunasekera says. "(+)-Discodermolide is more potent in stabilizing microtubules and more water soluble than Taxol. In addition, it is active in Taxol-resistant human cancer cell lines that overexpress P-glycoprotein, the multidrug-resistant transporter."

(+)-Discodermolide has been shown to be a promising candidate for clinical development as a drug for colon, ovarian, and breast cancers. It also, unusually, exhibits synergy with Taxol.

"Due to its strong microtubule-stabilizing properties and its strong activity against multiple drug-resistant tumors, Harbor Branch Oceanographic Institution licensed (+)-discodermolide to Novartis Pharmaceutical Corp. in early 1998 for development as an anticancer drug," Gunasekera notes.

Mickel points out that the marine sponge cannot provide the quantities of discodermolide needed for drug development. "The sponge has to be harvested using manned submersibles, and the compound accounts for only 0.002% by weight of the dried material," he says. "Some 3,000 kg of the sponge--a quantity that probably does not exist--would have been needed to deliver 60 g of (+)-discodermolide.

"Attempts to reproducibly isolate a discodermolide-producing microorganism for fermentation have not been successful to date," he continues. "Therefore, all discodermolide used for preclinical R&D activities as well as for the ongoing clinical trial has been supplied by total synthesis."

Chemistry professor Stuart L. Schreiber and coworkers at Harvard University reported the first total syntheses of the nonnatural (–)-antipode of discodermolide in 1993 [J. Am. Chem. Soc., 115, 12621 (1993)] and the natural (+)-antipode the following year [Chem. Biol., 1, 67 (1994)].

"THESE SYNTHESES were reported together with the discovery of a cellular (+)-discodermolide binding activity, which was shown in 1996 [Chem. Biol., 3, 2878, (1996)] to be the Taxol-binding site of microtubules," Schreiber notes.

Since the publication of the syntheses by Schreiber's group, several other total syntheses and preparations of various discodermolide fragments have been reported. According to Gunasekera, there are 24 U.S. patents and more than 300 scientific papers on discodermolide and its analogs in the literature.

The Novartis team evaluated all the available literature syntheses of the molecule with respect to yield, ease of reactions with regard to scale-up, availability of reagents, safety, and other requirements.

The synthetic route employed by the group was a hybrid of two literature approaches. The early stages used the 1-g synthesis of (+)-discodermolide achieved by chemistry professor Amos B. Smith III and coworkers at the University of Pennsylvania, Philadelphia [J. Am. Chem. Soc., 122, 8654 (2000)]. The latter stages, leading to the natural product itself, employed chemistry reported by a group at the University of Cambridge led by chemistry professor Ian Paterson [Angew. Chem. Int. Ed., 39, 377 (2000)].

"Novartis licensed our patented synthesis from the University of Pennsylvania," Smith notes. "They had access to all of our information as well as the patented material."

The Novartis-Smith-Paterson synthetic route, which took about 20 months to complete, was reported in a series of five papers earlier this year [Org. Process Res. Dev., 8, 92, 101, 107, 113, and 122 (2004)].

The first paper describes a multistep process, based on Smith's procedure, to convert the commercially available starting material (S)-3-hydroxy-2-methylpropionic acid methyl ester, known as Roche ester, to a Weinreb amide (an N-methoxy-N-methylamide). Weinreb amides are versatile compounds that are commonly used to prepare aldehydes and ketones. The one prepared by the Smith and Novartis groups contains the methyl-hydroxyl-methyl stereo triad that is repeated in three discodermolide structural segments--the C1­6, C9­14, and C15­21 fragments.

The Novartis team modified Smith's procedure to facilitate large-scale production of the amide in a pilot plant. Modifications included switching a reducing agent from lithium aluminum hydride to lithium borohydride to avoid the accumulation of large quantities of aluminum salts that do not allow for efficient filtration.

The large-scale preparations of the C1­6, C9­14, and C15­21 discodermolide fragments from the common Weinreb amide precursor using Smith's approach are reported in the second and third papers.

The fourth paper describes the preparation of the C7­24 intermediate--the "Paterson aldehyde"--that involves coupling the C9­14 and C15­21 fragments. In the course of this synthetic sequence, the transition is made from the Smith to the Paterson chemistry.

"Although the original Smith synthesis was attractive, it contained a step requiring a high-pressure reaction for the formation of a phosphonium salt from an alcohol intermediate for a late-stage Wittig reaction," Mickel says. "This step was not an option for us on any sort of scale. We then rationalized that this intermediate could be fed into the attractive end-game approach of Paterson."

The final paper describes the linkage of the C1­6 fragment (a methyl ketone) and the C7­24 aldehyde fragment using a boron-mediated aldol coupling procedure developed by Paterson's group.

The total synthesis required 17 large-scale chromatographic purifications, and many complex scale-up issues had to be solved along the way, according to Mickel.

"IMAGINE, after 30-odd steps and 18 months' work, you come to a crucial fragment coupling," Mickel says. "It has to work first time on a large scale. You can't quickly go back to the beginning and bring more material through. You have little material for a proper laboratory investigation, in contrast to a normal development project, so things have to work, and you have to take a calculated risk."

As an example of the challenges faced in scaling up the synthesis, Mickel cites the final step, which involved removing discodermolide's protecting groups to produce the final product.

"Easy chemistry, one might think," Mickel remarks. "We chromatographed the material and isolated it by evaporation from ethyl acetate. Now the HPLC [high-performance liquid chromatography] of the ethyl acetate solution revealed a material with a purity of more than 99%. After evaporation, we obtained a nice crystalline material. HPLC showed the presence of around 8% impurity. Panic broke out! After we calmed down, we realized that the lactone ring had probably opened and we had an equilibrium mixture. We produced the pure material again by adding a trace of acid."

The scale-up process progressed in three stages: proof of synthesis, preparation of a 6-g batch, and finally the production of 60 g of (+)-discodermolide. More than 43 chemists were involved in the synthesis concept, experimental design, and execution. They included chemists in the Process Research & Development Group in the Chemical & Analytical Development Department at Novartis Pharma, Basel; the Novartis Institutes for Biomedical Research, East Hanover, N.J; and the University of Cambridge.

"The option of optimizing the present synthesis further or replacing it with a better one is a topic of our ongoing studies, and we are confident of climbing this mountain as the situation demands," the authors note in the fifth paper of the series.

Paterson, who is one of the authors of parts four and five of the five-paper series, believes that the Novartis large-scale synthesis is a significant achievement. He is delighted that the chemistry developed by his group over a number of years has reached maturity.

"The work indicates that the challenging total synthesis of complex natural products, like discodermolide, can be performed successfully within the strict timelines set by the pharmaceutical industry, delivering sufficient material for human clinical trials that may ultimately generate new medicines," he comments. "It leads the way for the large-scale synthesis and clinical development of many other intriguing biologically active natural products, with even more complex structures, that are only available in very low natural abundance."

Smith's group has since overcome the high-pressure problem in his synthesis by replacing a large protecting group with a sterically less demanding one in the preparation of the Wittig salt during the final stages of the synthesis [Org. Lett., 5, 4405 (2003)].

Smith refers to the syntheses developed by his group in terms of generations. "Our first-generation synthesis was done shortly after the Schreiber first synthesis wherein we also synthesized the nonnatural enantiomer," he explains. "Our second-generation synthesis was the 1-g synthesis, and the third-generation synthesis was the work removing the high-pressure requirement in conjunction with other as yet unpublished improvements to our 1-g synthesis."

The longest linear sequence in the second-generation synthesis is 24 steps in a total of 34 steps, he points out, compared with 26 and 39 steps, respectively, in the Novartis large-scale synthesis. The overall yield of the Smith 1-g synthesis is 6.0%, whereas the Novartis yield is 0.65%.

"We anticipate that our new third-generation approach will further simplify the synthesis and reduce the cost of the clinical material," Smith remarks. "Clearly, the Novartis synthesis is a wonderful accomplishment, demonstrating that if a new drug candidate is sufficiently valuable, synthetic chemists will rise to the challenge of developing a viable synthetic approach no matter how complex the structure."