|Home | About | Journals | Submit | Contact Us | Français|
Natural products have a well-proven track record as cancer chemotherapeutic agents.i Beginning in the late 1960’s with the approval of the Vinca alkaloids vinblastine and vincristine for the treatment of a number of cancers, natural products have emerged both as treatments and as leads for further drug development. Without a doubt, the most important cancer chemotherapeutic agent to arise from a natural source is paclitaxel (Taxol,® Figure 1),ii a complex diterpene originally isolated from the bark of the Pacific yew Taxus brevifolia. Paclitaxel has met with considerable clinical success for ovarian, breast, and lung carcinomas, but the relative paucity of paclitaxel available from the Pacific yew (4000 trees were sacrificed to furnish 360 grams of the drug) nearly prevented the drug from being developed. The supply problem was ameliorated with the discovery that a structurally similar compound, 10-deacetylbaccatin III, could be extracted readily in much larger quantities from the renewable foliage of the European yew Taxus baccata, that, in turn, could be converted chemically to paclitaxel, or to the more potent synthetic congener Taxotere (Figure 1).
Paclitaxel comprises the first member of an ever-growing class of natural products that induce the polymerization of the ubiquitous cellular protein tubulin and suppress microtubule dynamics,iia thus preventing effective cell division. The second entry to this class, the epothilones A and B (Figure 1),iii also hold considerable promise as lead anticancer agents in part because, unlike paclitaxel, the epothilones showed efficacy against several multi-drug resistant (MDR) cell lines.iv However, the most interesting compounds to result from the epothilone studies are the structurally related synthetic analogues. For example, Danishefsky and coworkers discovered that 12,13-desoxyepothilone B (Figure 1), a late stage intermediate in the total synthesis of epothilone B, displayed in vivo tumor growth suppression superior to that of the natural product in mice bearing human xenograft tumors.ivc More importantly, 12,13-desoxyepothilone B (now termed epothilone D), exhibited little to no general cytotoxicity, whereas epothilone B often proved fatal to the treated mice. Further optimization ultimately led the Danishefsky team, in collaboration with Kosan Biosciences, to 26-trifluoro-E-9,10-dehydro-12,13-deoxyepothilone B (Fludelone, Figure 1), an orally available agent that led to complete tumor remission in two tumor xenograft models, with no relapse evident even after 200 days!v This synergy between total synthesis and drug optimization is not solely academic; in 2005, a joint Kosan/Roche venture entered epothilone D into Phase II clinical trialsvi for the treatment of breast, lung, and prostate cancer. Earlier this year that trial was halted in order to focus resources on a second, more promising epothilone B congener (9,10-didehydro-12,13-desoxyepothilone B, Figure 1)vi that was recently introduced into Phase II. Elsewhere, Bristol-Myers Squibb Company has completed Phase III clinical trials on Ixabepilone (another epothilone B analogue),vii and in June of 2007 applied for a New Drug Application for the monotherapy treatment of metastatic or locally advanced breast cancer.
A third structurally distinct microtubule-stabilizing natural product, (+)-discodermolide (1, Figure 1), was isolated in 1990 by Gunasekera and coworkers at the Harbor Branch Oceanographic Institute from the deep-sea marine sponge Discodermia dissoluta.viii Employing a battery of NMR experiments, including 1H, 13C, COSY, long-range COSY, and several 2D correlation experiments, the Gunasekera team determined that (+)-discodermolide is comprised of a linear polypropionate backbone, punctuated by Z-olefinic linkages at C(8,9) and C(13,14), a terminal Z-diene substituent at C(21–24), thirteen stereogenic centers (including four secondary hydroxyls and seven methyl substituents), a carbamate, and a fully substituted δ-lactone. While the relative stereochemistry was determined by X-ray crystallography, the absolute stereochemistry of (+)-discodermolide remained unknown until 1993, when Schreiber and coworkers reported the first total synthesis in the discodermolide area,ixa which unfortunately proved to be the unnatural antipode. The Schreiber group then prepared the natural congener.ixc
(+)-Discodermolide, like the epothilones, retains tumor cell growth inhibitory activity against several MDR cancer cell lines. However, (+)-discodermolide exhibits a number of characteristics unique among the micro-tubule-stabilizing agents, including a linear (not macrocyclic) framework, immunosuppressive properties both in vitroxa and in vivo,xb potent induction of an accelerated senescence phenotype,xi and synergistic antiproliferative activity in combination with paclitaxel.xii Importantly, despite the discovery of several additional microtubule-stabilizing natural products — including eleutherobin,xiii sarcodictyins A and B,xiv laulimalide and isolaulimalide,xv dictyostatin,xvi peloruside A,xvii FR182876,xviii WS9885B,xix taccalonolides,xx and the coumarinsxxi — discodermolide remains the most potent natural promoter of tubulin assembly yet discovered.xxii The intriguing biological activity profile has prompted a number of efforts directed towards the total synthesis of (+)-discodermolide, as well as towards the production and evaluation of synthetic analogues. The sections to follow present a detailed account of each of these aspects of discodermolide research.
Initial biological evaluation, carried out by researchers at Harbor Branch, demonstrated that (+)-discodermolide (1) is a highly potent immunosuppressive agent, with activity comparable to that of the clinically proven immunomodulator cyclosporin A, both in vitroxa and in vivo.xb Specifically, (+)-discodermolide was found to suppress the two-way mixed lymphocyte reaction (MLR) in both human peripheral blood leukocytes (PBL) and murine splenocytes. Mitogenic response of PBL in the presence of the immunostimulators concavalin A (ConA) or phytohemagglutinin (PHA) was similarly suppressed by (+)-discodermolide, with IC50 values in the low micro-molar range. Importantly, the immunosuppression response is observed at concentrations of (+)-discodermolide that are essentially non-toxic in vitro.
Follow-up experiments probed the efficacy of (+)-discodermolide as an in vivo immunosuppressant, utilizing the parental–F1 graft-versus-host reaction model for splenomegaly. At daily dosage levels of 1.25 and 5.0 mg/kg, discodermolide suppressed the splenomegaly response by 93% and 219%, respectively, which is 100–1000 fold more potent than cyclosporin A, and on par with the powerful immunosuppressant FK506.xb Mechanistic investigations, by researchers at Harbor Branch in collaboration with Roche and Merck, demonstrated that the immunosuppressive activity of discodermolide was largely due to a general antiproliferative effect on lymphoid cells, including PHA- and ConA-induced T cells, at low nanomolar concentrations.xxiii This antiproliferative activity was not limited to lymphoid cells, as (+)-discodermolide was demonstrated to maintain potent antiproliferative effects in several non-lymphoid cell lines.xxiii
Cell cycle studies designed to probe the general antiproliferative activity of (+)-discodermolide (1), first reported in 1993 by a collaboration between Harbor Branch, Roche, and Merck, demonstrated that discodermolide prevented murine DO11.10 T hybridoma cells from cycling normally through the phases of the cell cy-cle.xxiii In untreated controls, 68% of cells were found to comprise the G1 phase of the cell cycle, 31% were found at the S phase, and less than 1% could be seen at the G2/M phase. Three hours after treatment with discoder-molide, the characteristic percentages shifted to 52% at the G1 phase, 40% S phase and 8% G2/M phase. After 24 hours, an even more marked difference was noted, with only 25% of cells now comprising the G1 phase, 16% at the S phase, and 58% at the G2/M phase, indicating that discodermolide blocks the cell cycle at the G2/M phase. The cell cycle effect is reversible, as cells will resume normal cycling within 48 hours after the removal of discodermolide from the medium.xxiii
Subsequently (1996), ter Haar and coworkers, in collaboration with the Harbor Branch team, disclosed that discodermolide arrests mitosis via binding and stabilization of the microtubule network,xxiv findings which were confirmed by the Schreiber laboratory shortly thereafter.xxv
ter Haar and coworkersxxiv demonstrated that, under every reaction condition studied, tubulin polymerization was substantially more rapid upon treatment with discodermolide than with equimolar concentrations of paclitaxel. Indeed, discodermolide at a concentration of 10 µM proved able to promote assembly of microtubules at 37°C even in the absence of microtubule-associated proteins (MAPs) and/or GTP, conditions under which paclitaxel is inactive. With assistance from MAPs or GTP, 10 µM discodermolide is able to initiate tubulin polymerization at temperatures as low as 0°C; again, Taxol is inactive under these conditions. Stability to depoly-merization was also markedly superior in discodermolide-induced tubulin polymers. Specifically, addition of up to 5 mM CaCl2 to polymers at 0°C, in the presence or absence of MAPs and/or GTP, had no effect on the extent of assembly for the discodermolide-treated microtubule system. Taxol-induced polymers, on the other hand, began to disassociate in the presence of only 0.6 mM CaCl2, even in the presence of MAPs and GTP. A final point of comparison examined by ter Haar and coworkers was the average number and length of the tubulin bundles induced upon treatment with either of the two antitubulin agents, at temperatures ranging from 0°C to 37°C. Discodermolide-induced systems consistently displayed large numbers of very short polymers, with an average length between 0.5 and 0.6 µm; this result proved to be independent of temperature. On the other hand, Taxol-induced microtubules were fewer in number and grew to much longer lengths on average. At 10°C, polymers initiated by Taxol averaged 0.7 µm in length, with those observed at 20°C or 37°C averaging over 1.60 µm. These results highlight the extraordinary ability of discodermolide to hypernucleate tubulin assembly.
In a parallel series of experiments, the Schreiber group explored the competitive binding of paclitaxel and discodermolide, as well as the stoichiometry of the tubulin polymerization event;xxv in the process the Harvard group confirmed many of the ter Haar results. For example, pure tubulin dimers (10 µM) undergo polymerization in the presence of 10 µM Taxol, ensuring that each of the dimers is occupied with Taxol. Radiolabeled [3H]-discodermolide was then added in various concentrations, and the fraction of tubulin bound with discodermolide measured. Under these conditions, discodermolide binding proceeded with a 1:1 stoichiometry, reaching a saturation point at 10 µM. This result clearly demonstrated that the presence of bound Taxol does not interfere with the binding of discodermolide; however, whether discodermolide competes with Taxol or instead binds at a distinct site remained to be determined. Initially, this question was addressed via a series of competitive binding experiments, utilizing tritiated Taxol and discodermolide as a means to quantify their respective binding. These experiments demonstrated that discodermolide does indeed competitively displace [3H]Taxol from microtubules. Conversely, tubulin polymers bound with discodermolide remained largely unchanged even upon treatment with superstoichiometric concentrations of Taxol. Taken together, these results indicate that Taxol and discodermolide bind in a mutually exclusive manner to microtubules, implying the same or overlapping binding sites.xxv
In early 2006, in collaboration with Horwitz and coworkers, we disclosed the results of a series of radio-labeling/digestion experiments directed toward the elucidation of the precise discodermolide binding domain of tubulin.xxvi The study employed a radiolabeled discodermolide analogue bearing a benzophenone photoaffinity probe.xxvii Photoincorporation of the probe to tubulin was followed by formic acid hydrolysis, immunoprecipitaion experiments, and subtilisin digestion, as well as Asp-N and Arg-C digestion. These experiments implicated amino acid residues 355–359 as the site of photoincorporation, residues which lie very near the known taxol binding domain on the β-tubulin monomer. More specifically, residues 355–359 were found to comprise a sub-pocket within the taxoid binding site, an observation which guided the development of a model for discoder-molide binding wherein the discodermolide skeleton occupies the taxol binding site and the benzophenone appendage extends into the adjacent subpocket. It should be noted, however, that the precise mode and orientation of discodermolide binding at the molecular level remains speculative, in spite of extensive NMR and modeling studies.xxviii,xxix
In 1997, the Harbor Branch group assayed (+)-discodermolide against two multi-drug resistant cell lines (including Taxol-resistant lines) which overexpress the P-glycoprotein transporter.xxx The relative resistance factor was evaluated as the ratio between the IC50 in the drug-resistant cell lines and the IC50 in the parent drug-sensitive cell lines. The results indicate that while the SW620AD-300 colon carcinoma line displays a 930-fold increased resistance to paclitaxel relative to the drug sensitive variant SW620, discodermolide is only 25-fold less potent in the resistant cell line. Even more strikingly, paclitaxel diminishes in potency by a factor of 2800 in the drug-resistant A2780AD ovarian carcinoma cell line versus the parent 1A9 cell line, while the resistance factor of discodermolide is only 89.
Discodermolide was also evaluated against two 1A9-derived cell lines bearing β-tubulin mutations that render the cells resistant to treatment with Taxol. In both cases, the mutant cell line was over 20-fold less sensitive to Taxol than the wild-type 1A9. Conversely, complete sensitivity to discodermolide was retained in both mutant cell lines, a phenomenon which is thought to be the result of the superior hypernucleation activity of discodermolide relative to paclitaxel.xxx
Although paclitaxel has proven to be a highly successful chemotherapeutic, both alone and in combination with cisplatin, the growing incidence of clinical resistance to paclitaxel has prompted a search for new drugs and/or drug combinations. Traditionally, combination therapies utilize drugs whose mechanisms of action are distinct from one another, with the goal of targeting two independent disease-related pathways. However, therapies combining drugs with similar molecular targets (for example, paclitaxel and vinorelbine)vii have advanced into Phase I/II clinical trials, indicating that this tactic holds considerable promise. With this in mind, Horwitz and coworkers at the Albert Einstein College of Medicine, in collaboration with the Smith and Danishefsky groups, began to explore the effects of alternate microtubule-stabilizing agents against several carcinoma cell lines.xiia
These studies began with evaluation of several known microtubule-stabilizing agents, including paclitaxel, discodermolide, eleutherobin, and epothilones A and B, against a variety of P-glycoprotein expressing and non-P-glycoprotein expressing cell lines. By and large, cell lines that were sensitive to paclitaxel were also sensitive to the other four agents. However, paclitaxel-resistant cell lines which overexpress P-glycoprotein tend to display cross-resistance to eleutherobin, but remain sensitive to the epothilones and discodermolide. Horwitz and coworkers also discovered a mutant cell line, termed A549-T12, that does not overexpress P-glycoprotein, displays 9-fold resistance to paclitaxel, and in fact requires low levels of paclitaxel in order to cycle smoothly through the G2/M phase of the cell cycle. With the required low concentration of paclitaxel, this cell line exhibits cross-resistance to epothilone A, epothilone B, and eleutherobin, but not discodermolide, making discodermolide the only agent tested that retains potent cytotoxicity. Low concentrations of epothilone A, epothilone B, or eleutherobin were also able to maintain the viability of the A549-T12 cell line in the absence of paclitaxel, whereas discodermolide proved unable to prevent the G2/M-phase blockage at any concentration. Perhaps most interestingly, A549-T12 cells proved to be 20-fold more sensitive to discodermolide in the presence of low concentrations of paclitaxel (~ 2 nM), than in the absence, a result that is counterintuitive given the paclitaxel-dependence of the cell line.
The synergistic cytotoxicity of the discodermolide/paclitaxel combination in the A549-T12 cell line prompted Horwitz and Smith to investigate the phenomenon, with a paclitaxel/epothilone B combination serving as the control.xii Four carcinoma cell lines (A549, SKOV3, MCF-7, and MDA-MB-231) were treated with each of the two drug combinations; a Combination Index (CI) value was calculated, according to the method of Chou and Talalay,xxxi to determine whether the combination acts synergistically (CI < 1), additively (CI = 1), or antagonistically (CI >1). The outcome clearly revealed a high degree of synergy between paclitaxel and discodermolide; the calculated CI values were significantly less than 1 in all four cell lines, across a 3–4 fold log concentration of the drugs. In contrast, only additive interactions were observed for the combination of paclitaxel and epothilone B.
Follow-up studiesxiib demonstrated that most of the microtubule dynamic instability parameters (growth rate, shortening rate, growth duration, shortening duration, pause duration, growth length, shortening length, % time spent growing, % time spent shortening, % pause time, and overall dynamicity) were synergistically altered by the paclitaxel/discodermolide combination, with CI values ranging from 0.20 to 0.41. Additionally, the cell cycle progression was synergistically blocked at the G2/M phase with a CI of 0.59. Apoptosis was also enhanced cooperatively, although to a lesser degree (CI = 0.85). The overall degree of synergy is quite surprising, given that paclitaxel and discodermolide appear to share both a common binding site and mechanism of action. xiib
Synergism of paclitaxel and discodermolide was also observed in vivo using SKOV3 human tumor xenograft models.xxxii In fact, combination treatment suppressed angiogenesis and induced tumor regression at drug concentrations that independently led to only minimal tumor growth suppression. A closer look revealed that at low concentrations, treatment with paclitaxel or discodermolide resulted in drug-induced aneuploidy, while at higher concentrations, single-agent treatment resulted instead in mitotic arrest. The synergistic effect of combination treatment was most profound when administering relatively low concentrations of paclitaxel and discodermolide (i.e. a drug regimen which induced aneuploidy, but not mitotic arrest). Importantly, the combination of discodermolide and paclitaxel was remarkably well tolerated in the treated mice, with no toxicity noted at efficacious levels. While the precise mechanism of the observed synergism remains unknown, Horwitz and Smith speculate that the two agents might preferentially target different tubulin isotypes, despite exhibiting mutually exclusive binding to a likely common tubulin binding site. Additionally, discodermolide, but not paclitaxel, has been shown to induce an accelerated senescence phenotype (which will be the subject of the following section), providing a second potential mechanism for synergy. Efforts to elucidate the exact nature of this effect continue.
In 2005, Horwitz and coworkers, in collaboration with the Smith group, isolated a A549-derived cell line that proliferated in 8 nM discodermolide, a concentration approximately one-third the IC50 value for parental A549 cells.xi Further examination of this cell line, which was termed A549.Disco8, revealed that short-term exposure to the IC50 concentration (24 nM) of discodermolide results in both an alteration of cellular morphology and the expression of senescence-associated β-galactosidase (SA-β-gal) activity, both characteristic of an accelerated senescence phenotype. Hallmarks of senescence, which commonly occurs in cells that have undergone a series of cell divisions, include the following: a cessation of proliferation, an overall flattening of the cellular structure, an increased cytoplasmic area, and the onset of significant SA-β-gal activity.xxxiii Although many antitumor treatment regimens are known to induce the senescence phenotype, this phenomenon had not been observed previously with microtubule-stabilizing agents.
To explore the generality of this effect, Horwitz and coworkers separately treated HeLa, MDA-MB-231, HCT-116, and parental A549 cells with discodermolide, paclitaxel, and the known senescence-inducing agent doxorubucin (25 nM each). As expected, doxorubicin produced consistently high levels of SA-β-gal activity in all four cell lines. Discodermolide-treated cells also exhibited moderate to high levels of SA-β-gal activity across the panel, while paclitaxel induced far less, if any SA-β-gal activity.xi
Investigation of the expression and activity of several proteins known to be associated with senescence found several differences between discodermolide-treated cells and those treated with paclitaxel, providing a basis for the disparity in the effects of the two antimicrotubule agents. Most notably, discodermolide induced substantial up-regulation and/or activation in three proteins (p66Shc, Erk1, and Erk2) involved in the mitogenic activation signaling pathway; in contrast, little or no increase in activity was noted upon paclitaxel treatment. Several other cellular pathways were found to require a much higher concentration of paclitaxel than discodermolide (relative to IC50) to promote the senescence-related response. The longevity and/or intensity of the response induced by the two agents also tended to differ. xi
Traditionally, the search for the treatment of Alzheimers and related neurodegenerative diseases has focused on the prevention of the amyloid plaques and neurofibrillary tangles that are the hallmarks of these disorders.xxxiv Recently, however, a new therapeutic approach to neurodegenerative diseases was validated by Lee and Trojanowski at the University of Pennsylvania, which entailed the in vivo use of microtubule-stabilizing drugs to restore function to neurons that have been disrupted by the β-amyloid-induced sequestration of the microtubule-associated protein tau.xxxv
In healthy neurons, the function of tau protein is to stabilize the microtubule network, which serves as the “railroad track” upon which actin, tubulin, mitochondria, neurotransmitter-related enzymes, and vesicles carrying messenger proteins are delivered between the nucleus and the extremities of the elongated cell.xxxvi The presence of amyloid-β, however, has been demonstrated to lead to hyperphosphorylation and aggregation of tau protein into neurofibrillary tangles,xxxvii rendering tau protein ineffective at stabilizing the microtubule network.xxxviii These “tauopathies” result in a functional loss of axonal transport, leading to reduced microtubule numbers, and in turn, neurodegeneration and motor weakness.xxxix Utilizing a transgenic mouse model for human tauopathy, the Penn team demonstrated that exogenous paclitaxel stabilizes microtubules, and thereby restores fast axonal transport and microtubule numbers to physiologically viable levels, thus offsetting the loss of function caused by the hyperphosphorylation, misfolding, and aggregation of tau protein.xxxv Most importantly, motor-impaired mice exhibited a complete recovery of normal movement upon treatment with paclitaxel!
In a concurrent study, Georg et al. examined the use of microtubule-stabilizing compounds as neuropro-tective agents in vitro, using cortical neurons that had been exposed to amyloid-β.xl They found that the neuroprotective effects of paclitaxel, as measured by the neuron survival rate in treated cells versus untreated cells, were general for a variety of microtubule-stabilizing agents, including epothilone A, discodermolide, and the paclitaxel analogue docetaxel. The effective concentration (EC50) of the neuroprotective agents evaluated appears to mirror closely the tubulin assembly activity (see Section 1.3.1), with discodermolide proving to be almost an order of magnitude more potent than paclitaxel (EC50 1.7 nM and 13 nM, respectively); epothilone A also proved superior to paclitaxel, with an EC50 of 3.3 nM. Taken together, the promising in vivo efficacy of paclitaxel,xxxv coupled with the superior in vitro potency of discodermolide and epothilone A,xlb indicate that further evaluation of these natural products and related analogues as neuroprotective agents is warranted.
The impressive biological profile, coupled with both the growing interest from the medical and pharmaceutical communities and the lack of a ready natural source, has rendered (+)-discodermolide (1, Figure 1) an alluring target for total synthesis. To date, twelve unique syntheses have been disclosed by eight different research groups, both in academia and in industry.xli In addition, a variety of tactics directed at the construction of discodermolide substructures have also appeared.xlii As the body of literature expands, attention has turned towards shorter, more convergent, and generally more practical discodermolide syntheses (i.e., routes that are designed specifically to generate quantities of material sufficient for clinical evaluation).xlih,i,j,l Extensive efforts to cultivate the organism responsible for the natural production of discodermolide have thus far not been successful, leaving total synthesis as the sole source of this medicinally relevant natural product.
From the retrosynthetic perspective, all of the total syntheses reported to date disconnect (+)-discodermolide into three major fragments of roughly equivalent complexity, each of which comprise the methyl-hydroxyl-methyl triad of contiguous centers of stereogenicity in the target structure (Figure 2). It should be noted that several of the early accounts (specifically those of Schreiber, the first-generation Smith approach, and that of Myles) reported the synthesis of the unnatural antipode (−)-discodermolide. The summaries detailed herein, however, will all be depicted in the natural optical series for clarity. Signs of optical rotation for the various intermediates will only be provided for syntheses in the natural series.
In 1993, Schreiber and coworkers reported the total synthesis of the unnatural antipode (−)-discodermolide, and in turn established the absolute stereochemistry of the natural product.xlia A full report detailing the synthesis of the naturally occurring enantiomer appeared in 1996. Described here is the synthesis of the natural congener. From the outset, the Schreiber team designed a doubly convergent strategy that exploited the occurrence of the common triad of stereogenic centers (Scheme 1), beginning from known homoallylic alcohols 2 and 3,xliii both of which were derived from 3-hydroxy-2-methylpropionate. To begin, alcohol 2 was converted in four straightforward steps to diol 4. Oxidation to the corresponding keto-aldehyde, followed by vinyl iodide installation under Stork-Zhao conditions,xliv and then diene formation under the conditions of Negishi and coworkers,xlv furnished diene 5 in moderate yield.
For construction of alkynyl iodide fragment 7, Still-Gennari olefinationxlvi installed the requisite C(13)-C(14) Z-trisubstituted olefin; a 6 step sequence entailing homologation to the acetylenexlvii and iodination then furnished the desired iodide.
Synthesis of the C(1)-C(6) subunit 10 (Scheme 1) began with ozonolytic cleavage of alkene 3. Wittig olefination, followed by hemi-acetal formation and in situ intramolecular 1,4-addition of the resulting alkoxide.xlviii Two-step formation of the methyl acetal then afforded methyl ester 9, which was converted in five steps to the final key intermediate, mixed thioacetal 10.
The Schreiber endgame comprised a Nozaki-Hiyama-Kishi unionxlix of alkynyl iodide 7 with aldehyde 10, followed by partial hydrogenation of the alkyne to establish the Z-configuration at C(8)-C(9); three additional transformations afforded allylic bromide 11 (Scheme 2). Displacement of the allylic bromide with the lithium enolate derived from methyl ketone 5 completed assembly of the discodermolide backbone; subsequent methylation furnished ketone 12. A six-step sequence comprising conversion of the thioacetal to the corresponding ketone, carbamate installation,l directed reduction of the C(16) lactone, and global deprotection completed the synthetic sequence, providing discodermolide with a longest linear sequence of 24 steps (36 total steps) in an overall yield of 4.3%.
The second total synthesis of discodermolide, in this case the (−)-unnatural isomer, was reported in 1995 by Smith and colleagues.xlib The cornerstone of the triply-convergent Smith strategy entailed common precursor 14 (Scheme 3), from which each of the three advanced intermediates 16, 18, and 22 would ultimately derive. The common precursor was the product of a highly efficient five-step sequence, beginning from 3-hydroxy-2-methylpropionate, which utilized an Evans syn-aldol reactionli to set the requisite stereochemical arrangement.
Alkyl iodide 16 was generated in six transformations from the common precursor 14 (Scheme 3), including a second Evans aldolli to install the requisite stereogenicity at C(16) and C(17). Installation of the critical Z-trisubstituted vinyl halide under the conditions of Zhao and coworkers,lii albeit with modest yield and selectivity, was the central feature of the synthesis of vinyl iodide 18. A detailed investigation of the mechanism of this troublesome transformation was later reported by Smith and coworkers.liii Preparation of the third intermediate, aldehyde 22, began with a highly efficient six-step sequence from 14 to afford dithiane 19 (Scheme 3). Central to the construction of aldehyde 22 was addition of the anion derived from dithiane 19 to benzyl glycidyl ether, employing a tactic well-precedented in the Smith group.liv Seven steps, including installation of the C(7) stereo-center and thioacetal formation, then provided aldehyde 22, completing a sequence which ultimately required 14 synthetic operations beginning from the common precursor (14), placing this fragment squarely on the longest linear route of the Smith first-generation synthesis.
With the three intermediates 16, 18, and 22 in hand, attention turned to their union and subsequent elaboration to discodermolide (Scheme 4). Palladium(0)-mediated cross-couplingxlv of vinyl iodide 18 with the organozinc reagent derived from alkyl iodide 16, followed by oxidative removal of the PMB ether furnished primary alcohol 23, which was next subjected to what at the time proved to be a difficult two-step conversion to the corresponding phosphonium salt 24. In particular, a side reaction involving olefin-assisted cyclization of the C(13)-C(14) alkene onto the pendant iodide, led to a significant amount of cyclopentane-containing byproducts; at best 37% of the desired salt was achieved. Future iterations of the Smith approach would focus in large part on the optimization of this transformation (See Section 3.6). Wittig unionlv of phosphonium salt 24 with aldehyde 22 proceeded in good yield to furnish acetal 25. Three step installation of the diene employing the titanium—mediated conditions reported by Yamamoto and coworkers,lvi first used in the discodermolide area by Heath-cock,xliia completed the construction of the carbon skeleton. What remained was hydrolysis of the thioether, oxidation, carbamate installation,l and global deprotection, producing (−)-discodermolide with a longest linear sequence of 29 steps (42 total steps) in an overall yield of 2.0%.
Myles and coworkers reported their total synthesis of (−)-discodermolide in 1997,xlic marking the third approach to appear in the literature.A follow-up full account in 2003 recorded the synthesis of the (+)-antipode.lvii The Myles approach to the three major fragments 30, 32, and 34 is described in Scheme 5. Most notably, the Myles team employed a titanium-mediated hetero-Diels Alderlviii reaction of aldehyde 27 with the Danishefsky dienelix to construct the challenging Z-trisubstituted C(13)-C(14) olefin present in allylic iodide 32. With respect to ethyl ketone 30, after initially exploring a substrate-controlled approach (27 to 28)lx to induce the stereogenicity found in the C(16)-C(21) subunit, a longer but more selective strategy based on Evans aldolli chemistry was employed. The syn aldol was subsequently used in tandem with a tin-promoted asymmetric allylationlxi to set the stereogenicity of the final key intermediate, aldehyde 34. Importantly, fragment 34 bears the correct oxidation state at C(1), which in turn would minimize late stage manipulations.
Strategically, the Myles team envisioned discodermolide to arise via a Nozaki-Hiyama-Kishi nickel/chromium-mediated unionxlix of a fully elaborated C(8)-C(24) vinyl iodide (37) with the C(1)–C(7) aldehyde 34 (Scheme 6). Vinyl iodide 37, in turn, was constructed beginning with chelate-controlled addition of the enolate of the C(16)-C(21) ketone 30 to an appropriately functionalized allylic iodide (32). Schreiber had examined a similar transformation, with C(16) already bearing the methyl substituent, before employing an alternative two-step addition/methylation process. In the Schreiber system,ixc enolate coupling did not proceed via the expected chelation-controlled transition state, but instead via a transition state conformation governed by allylic (A1,3) strain, which led to the undesired C(16)-R configuration.
Contrary to the Schreiber results, alkylation of the enolate derived from 30 with allylic iodide 32 in a solvent mixture consisting of hexanes/THF (45/55) proceeded under chelation control to afford the desired C(16) methyl stereoisomer in good yield and with reasonable (6:1) diastereoselectivity (Scheme 6). Further manipulation, including installation of the requisite cis-disubstituted vinyl iodide,xliv afforded aldehyde 36, which would serve as substrate for a Peterson-type olefination,lxii employing the conditions reported by Roush and coworkerslxiii to introduce the diene. Acid-mediated deprotection of the C(19) MOM ether followed by carbamate installationl completed the synthesis of vinyl iodide 37. Unfortunately, the subsequent nickel/chromium mediated union of vinyl iodide 37 with aldehyde 34 proceeded in only modest yield and selectivity, but this route supplied sufficient material to enable global deprotection and completion of (−)-discodermolide. Overall, the Myles total synthesis of discodermolide entailed a longest linear sequence of 25 steps (44 total steps), with an overall yield of 1.5%.
Marshall and coworkers viewed the synthesis of (+)-discodermolidexlid as an exercise to showcase the utility of their asymmetric allenylmetal-homoaldol tacticlxiv to construct polypropionate frameworks. From the retrosynthetic perspective, the strategy incorporates two early disconnections, one at the C(14)-C(15) bond, setting the stage for a palladium(0)-catalyzed cross coupling à la Smith and coworkers (Section 3.2), and the second at the C(7)-C(8) bond, similar to that of both Schreiber (Section 3.1) and Myles (Section 3.3). The novelty of the Marshall approach lies in the utilization of non-racemic allenylmetal reagents for installation of various stereotriad subunits, either in a syn/syn sense, as in the alkyl iodide fragment (+)-44, or in a syn/anti sense, as in (+)-48 and (−)-51 (Scheme 7).
To begin, aldehyde (+)-39, produced in 3 steps from Roche’s ester, was subjected to allenyltributylstan-nane (+)-40 in the presence of BF3-etherate to furnish the syn/syn isomer (+)-41 both in excellent yield and with high selectivity. Protecting group manipulation, followed by partial reduction of the alkyne to the corresponding trans-olefin, permitted installation of the two remaining stereocenters via an asymmetric epoxidation-lxv methylation sequence. What remained to complete alkyl iodide (+)-44 was formation of the pendant Z-diene via the Paterson two-step one-pot protocol;lxvi the latter entails Nozaki-Hiyama-Kishi additionxlix of α-TMS-allyl bromide to the aldehyde derived from (−)-43, followed by Peterson syn-eliminationlxii of the resultant β-hydroxysilane to furnish (+)-44. For the C(8)-C(14) fragment (+)-48, aldehyde (+)-45 was treated with the homochiral allenylzinc reagent derived in situ from propargylic mesylate (+)-46 to furnish alkynyl alcohol (−)-47, possessing a syn/anti methyl-hydroxyl-methyl stereotriad. Protection as a methoxymethyl ether completed the two step sequence.
The synthesis of C(1)-C(7) aldehyde (−)-51 also proceeds from alkynyl alcohol (−)-47 (Scheme 7). Desilylation and acetal formation was followed by hydroxylmethylation of the alkyne to provide alcohol (−)-49. Partial alkyne reduction and epoxidation then furnished (−)-50, which upon reductive ring opening of the epoxide, followed by a 3-step protection of the secondary alcohol and oxidation, completed the synthesis of aldehyde (−)-51.
The Marshall endgame (Scheme 8) began with the addition of the alkynyl lithium derived from (+)-48 to aldehyde (−)-51, a transformation which proceeded in 92% yield, with 6:1 diastereoselectivity. Semihydrogenation of the resultant alkyne led to the desired cis C(8)-C(9) olefin, at which point 4 steps, including vinyl iodide installation,lii set the stage for final coupling. In the event, a Suzuki unionlxvii of vinyl iodide (−)-53 with the boronate derived from alkyl iodide (+)-44 in the presence of catalytic palladium(0) proceeded in good yield to provide the expected tetraene 54.
Completion of (+)-discodermolide entailed an 8-step sequence involving formation of the C(1) methyl ester, carbamate installation, and global deprotection with concomitant lactonization, to afford the natural congener with a longest linear sequence of 30 steps (48 total steps) with an overall yield of 1.3%.
In a thesis published in 1999, David Halstead, working under the direction of Evans at Harvard, described a fifth total synthesis of discodermolide.xlif The Evans strategy, not surprisingly, relies heavily on asymmetric aldol methodology for the elaboration of the polypropionate backbone, in conjunction with a novel Claisen condensation to construct the C(13)-C(14) trisubstituted olefin. The C(7)-C(8) disconnection is similar to that employed by Schreiber (Section 3.1) and Myles (Section 3.3).
Halstead and Evans designed their strategy around three major advanced intermediates, lactone (−)-58, alkyne (+)-62, and aldehyde (+)-65 (Scheme 9). Synthesis of (−)-58 began with the tin(II)-mediated syn aldollxviii reaction of β-ketoimide (−)-56 with aldehyde (−)-55, which set the stereogenicity of the C(15) and C(16) centers with high diastereoselectivity. Directed reduction of the ketone installed the remaining stereogenic center, which was followed by functional group manipulation and para-methoxyphenyl acetal formation to furnish (−)-58.
Aldehyde (+)-62 was produced by a 10 step sequence (Scheme 9) that began with an aldol reaction between the boron enolate of propionyl oxazolidinone (−)-59 and methacrolein; diastereoselective hydroboration of the resulting olefin with 9-borabicyclo[3,3,1]-nonane then afforded primary alcohol (−)-61. Oxidation, Corey-Fuchs reaction,lxix detritylation, and a second oxidation completed the synthesis of the C(8)-C(13) fragment (+)-62.
Construction of aldehyde (+)-65 began with the aldol reaction of the same β-ketoimide (−)-56 employed for the production of lactone 58. Here, treatment of aldehyde 63 with the dicyclohexylboron enolate of (−)-56 furnished the desired anti configuration of the nascent stereocenters,lxviii in contrast with the tin(II)-mediated conditions used earlier to induce the syn aldol product. Protection of the derived hydroxyl as the TES ether and zinc borohydride-promoted syn-reduction of the ketone then led to (+)-64, comprising the complete tetrad of stereogenicity of the C(1) to C(6) region of discodermolide. What remained was protecting group adjustment and oxidation to complete the synthesis of aldehyde (+)-65.
With (−)-58 and (+)-62 in hand, a two-step formal aldol condensation was utilized to effect their union, as well as to induce the requisite C(13)-C(14) Z-olefin geometry (Scheme 10). A three-step ring opening/deoxygenation protocol was then followed by protecting group adjustment, alkynylation, and iodination to furnish iodide (+)-68. Nozaki-Hiyama-Kishi unionxlix of (+)-68 with aldehyde (+)-65 proceeded in good yield with reasonable diastereoselectivity (ca. 6:1) at the C(7) hydroxyl. This result compared favorably with the 3:1 ratio obtained by Schreiber in an analogous transformation (see Section 3.1). Partial reduction of the C(8)-C(9) alkyne, TES protection of the C(7) hydroxyl, hydrolysis of the C(21) silyl ether, and oxidation then led to aldehyde 70. Wittig installation of the terminal diene exploiting a γ-selenophosphonium ylidelxx followed by oxidative elimination proceeded smoothly with both high selectivity and excellent yield. What remained was a 4-step sequence involving carbamate installation and lactonization to complete the synthesis of (+)-discodermolide, with a longest linear sequence of 31 steps (49 total steps) and an impressive overall yield of 6.4%.
In order for (+)-discodermolide to become a viable candidate for drug development, quantities of the natural product sufficient to permit pre-clinical evaluation were required. The natural sponge source, which provided only 7 mg of (+)-discodermolide from 434 g of sponge, was clearly not a viable option. Total synthesis therefore provided the only feasible means of access. The challenge to produce meaningful quanitites of (+)-discodermolide was addressed by Smith, et al. with the production of 1.043 g of totally synthetic (+)-discodermolide, based on a route which incorporated several practical improvements vis-à-vis the earlier triply-convergent, first-generation approach (see Section 3.2).xlie The one-gram synthesis marked a turning point in the development of (+)-discodermolide as a potential chemotherapeutic agent, as the material produced and the synthetic route employed for the production, were subsequently licensed to Novartis Pharmaceuticals. Phase I clinical trials began just three years later.
While the Smith first-generation synthesis (Section 3.2) of the unnatural antipode (−)-discodermolide could be readily modified to produce the natural congener, the route was far from optimal. A critical analysis of the weak points of the earlier approach was undertaken prior to initiating the preparative-scale synthesis. Two key areas required readdress. First, the linear sequence was simply too long and had to be reduced, ideally with a corresponding increase in overall convergence. With this idea in mind, an improved second-generation plan was developed which retained the efficient, triply-convergent nature of the first-generation approach (Scheme 11). Second, formation of an advanced intermediate phosphonium salt [such as 24, Scheme 4 or (+)-71, Scheme 11] from the corresponding primary alcohol required optimization.
Facile production of large quantities of common precursor (−)-14 was central to this endeavor. A notable improvement to the 5-step route to (−)-14 (Scheme 12) derived from the discovery that aldol adduct (+)-78,which contains the norephedrine-based oxazolidinone moiety, was a crystalline solid,lxxi while the corresponding first-generation intermediate, phenylalaninol-derived adduct (−)-76, proved to be an oil. Recognition of this fact allowed (+)-78 to be isolated directly from the reaction mixture by crystallization. As a result, chromatographic purification was not required for the first four steps of the sequence.
With an efficient and practical route to (−)-14 in hand, the early stages of the first-generation approach, comprising synthesis of alkyl iodide (+)-16 and vinyl iodide (+)-18 from (−)-14, proved readily scalable, and thus required little modification. The first major point of departure from the first-generation strategy entailed synthesis of the C(1) – C(8) lactone fragment (−)-72. Here, a strategic opportunity to carry the intact δ-lactone moiety unprotected into the endgame sequence was envisioned, which in turn would lead to fewer late stage manipulations. In the end, a 7-step route from (−)-14, comprising a chelation-controlled Mukaiyama aldollxxii reaction to set the C(5) stereogenicity, followed by a diastereoselective K-Selectride reduction of the resultant C(7) ketone (Scheme 12), were ultimately employed to furnish the advanced lactone aldehyde (−)-72.
With the three fragments in hand, union of vinyl iodide (+)-18 with the mixed organozinc reagent derived from alkyl iodide (+)-16 under modified Negishixlv conditions proceeded smoothly to afford trisubstituted olefin (+)-80 (Scheme 13). Further manipulation then furnished aldehyde (+)-81, which would serve as the substrate for the installation of the C(21)-C(24) terminal diene.lvi Deprotection of the C(9) trityl ether set the stage for exploration of the problematic two-step iodination / Wittig salt construction. The initial solution to this problem was the use of ultra-high pressure, a tactic which was pioneered by Dauben and coworkers to accelerate the alkylation of phosphines.lxxiii In the event, treatment of a highly concentrated benzene solution of the C(9) primary iodide with 10 equivalents PPh3 at 12.8 KBar for 7–10 days afforded the desired phosphonium salt (+)-71 in 79% yield, with only 20% of the undesired cyclopentane byproducts. Continuing, unionlv of the Wittig ylide derived from (+)-71 with the fully functionalized C(1) – C(8) lactone aldehyde (−)-72 proceeded in good yield, thus completing construction of the discodermolide carbon skeleton. What remained was a three-step sequence involving deprotection of the C(19) PMB ether, carbamate installation, and global deprotection.
While the ultra-high pressure tactic provided sufficient phosphonium salt [(+)-71] for the production of over one gram of discodermolide, a more practical method for the generation of phosphonium salt (+)-71 would remain the subject of intense investigation in the Smith laboratory. Two years following the one gram synthesis of (+)-discodermolide, a third-generation effort that simply replaced the bulky TBS ether protecting group at C(11) of the precursor iodide with a less sterically encumbering group, as in MOM ether (+)-84 (Scheme 14), was disclosed by Smith and coworkers. This tactic proved to bias the system to produce the desired phosphonium salt, at ambient pressure, at the expense of the undesired cyclopentanes.xlih Smith and coworkers later found that alkyl iodide (+)-82, which possesses the terminal diene, and vinyl iodide (+)-83, in which the requisite protecting group exchange had already been effected, served as viable coupling partners in the context of the sequence described above (Scheme 13).xlil This modification served to increase the overall convergence of the gram-scale route to (+)-discodermolide.
Taken together, the improvements described above ultimately afforded (+)-discodermolide with a longest linear sequence of 21 steps (35 total steps) and a much-improved overall yield of 6.0%.
The final iteration of the Smith synthesis of (+)-discodermolide was reported in 2005.xlil The highlight of the fourth-generation approach was the construction and sequential, bidirectional union of the vinyl iodide/phosphonium salt (−)-88 (Scheme 15). Due to the relatively deactivated nature of the trisubstituted vinyl iodide, phosphonium salt (−)-88 could be readily generated in two steps from alcohol (−)-86, with no evidence of concomitant cyclopentane formation that plagued several earlier syntheses. Wittig union with (−)-89 followed by Suzuki coupling with alkyl iodide (+)-82 efficiently furnished tetraene (+)-91. To complete the endeavor, debenzylation, carbamate installation, and global deprotection then afforded totally synthetic (+)-discodermolide (1) with an overall yield of 9.0% and with a longest linear sequence of 17 steps (36 total steps). Importantly, this synthesis entails the shortest linear sequence reported to date.
In 2000, Ian Paterson and coworkers at the University of Cambridge (UK) reported their initial efforts regarding a preparative approach toward (+)-discodermolide.xlig Retrosynthetically, the Paterson first-generation strategy (Figure 2) was based on two key aldol-type disconnections, employing aldehyde (+)-95, ester (+)-100, and methyl ketone (−)-103 (Scheme 16). Production of each subunit began with boron-mediated anti-aldol reactions as the stereoselective transformation, a tactic well-precedented in the Paterson laboratory.lxxiv
Employing a novel chelation-controlled aldol reaction for the construction of the C(16)-C(17) bond, the enolate of aryl ester (+)-100 was treated with aldehyde (+)-95 to furnish the desired anti-aldol adduct (+)-104, both with complete Felkin-Anhlxxv selectivity and in good overall yield (Scheme 17). Reductive removal of the aryl ester moiety was followed by an 8 step sequence including carbamate installation, and then a Still-Gennari Horner-Wadsworth-Emmons reactionxlvi to induce the requisite cis geometry of the C(8)-C(9) olefin. Reagent-controlled aldol union of the (+)-diisopinocampheylborane enolate derived from the C(1)–C(6) methyl ketone (−)-103 with aldehyde (+)-106 then led to desired carbon skeleton, simultaneously setting the stereogenicity at C(7). Directed reduction of ketone (+)-107, followed by global deprotection with concomitant lactonization, completed the Paterson first-generation synthesis of (+)-discodermolide, which required 23 steps as the longest linear sequence (42 total steps) and proceeded with a remarkable overall yield of 10.3%.
Three years later, in 2003, Paterson and colleagues communicated a revised approach to the synthesis of (+)-discodermolide based solely on substrate-derived stereocontrol.xlii A full reportlxxvi of this second generation route was published shortly thereafter. While the retrosynthetic strategy was largely analogous to their first-generation approach (Section 3.7), a few key changes were made. Specifically, the stereoselective incorporation of the C(1)-C(5) subunit 103 was accomplished via a dicyclohexylboron-mediated anti-aldol, which replaced the reagent-controlled transformation employed in the first-generation synthesis. Additionally, construction of the C(13)-C(14) trisubstituted olefin was achieved via a Still-Gennari modified Horner-Wadsworth-Emmons reaction.xlvi
The syntheses of each of the three major second-generation Paterson intermediates (+)-95, (+)-112, and (−)-114 are detailed in Scheme 18. Note that the key stereoselective transformation in each case is the same dicyclohexylboron-mediated anti-aldol reaction that was exploited to great success in the first-generation Paterson synthesis.
The choice of an aromatic ester proved to be critical for the construction and utility of the C(9)-C(16) fragment (+)-112. In the earlier Paterson approach, when the aromatic substituent was 2,6-dimethylphenyl, the aldol reaction of the enolate derived from aryl ester (+)-100 with aldehyde (+)-95 proceeded with 97% diastereoselectivity and in good chemical yield (cf. Scheme 17). The first-generation 2,6-dimethylphenyl ester (+)-100, in turn, was the product of esterification of the corresponding carboxylic acid (Scheme 16). Access to the requisite 2,6-dimethylphenyl ester (+)-100 via the second-generation approach, however, entailed alkylation of the aryl ester enolate with an allylic iodide, which at best proceeded in 38% yield (results not shown). Accordingly, a series of aryl substituents were evaluated; ultimately, the 2,6-dimethyl-4-methoxyphenyl substituent, as in 111, afforded the best results both in the alkylation reaction [Scheme 18, (−)-110 --> (+)-112] and in the subsequent aldol union (Scheme 19, 95 + 112). Continuing with the synthesis, a 3-step reductive removal of the aryl ester moiety, followed by silylation of the secondary alcohol furnished bis-PMB ether (+)-105 (Scheme 19). Debenzylation and TEMPO oxidation then provided an intermediate aldehyde, which upon Still-Gennari reactionxlvi installed the Z-disubstituted C(8)-C(9) olefin. After introduction of the C(19) carbamate, the stage was set for the final fragment union.
In the event (Scheme 19), treatment of aldehyde (−)-114 with the dicyclohexylboron enolate of methyl ketone (+)-116 proceeded in high yield and with good diastereoselectivity. Paterson proposes that the geometry of the nascent C(5) stereocenter is controlled by 1,6-stereoinduction, based on a transition state conformation governed by minimization of A(1,3) strain at C(10). To complete the synthesis, lactonization, stereoselective reduction of the C(7) ketone, and global desilylation then furnished (+)-discodermolide, with a longest linear sequence of 24 steps (35 total steps) and an overall yield of 7.8%.
Subsequently (2004), Paterson and coworkers reported a third-generation endgame,xlik in which a latestage Still-Gennari-type olefinationxlvi replaced the stepwise method used previously to incorporate of the C(1)-C(8) subunit, thus increasing the overall convergence. Accordingly, C(1)-C(8) phosponate ester (−)-118 (Scheme 20) was designed and constructed. A selective anti-aldol reaction of the dicyclhexylboron enolate derived from ethyl ketone (+)-96 with 2-benzyloxypropanal and in situ reduction of the intermediate aldolate installed the stereotetrad in high yield and with excellent diastereoselectivity (97:3 dr). Further manipulation furnished phosphonate ester (−)-118.
The endgame of the Paterson third-generation synthesis begins with bis-PMB ether (+)-105 (Scheme 20), an intermediate found in both previous Paterson routes. Horner-Wadsworth-Emmons reaction under the conditions of Still and Gennarixlvi effected the desired olefination to furnish tetraene (+)-119 with a cis:trans selectivity of 10:1 in 74% isolated yield after separation of the minor isomer. Carbamate installation, K-Selectride-mediated reduction of the C(7) ketone, and acid-promoted desilylation then furnished (+)-discodermolide, this time with a longest linear sequence of 21 steps (37 total steps) and with an impressive overall yield of 11.1%, the highest reported to date.
In early 2004, the work of Smith and Paterson directed at a practical, large-scale synthesis of (+)-discodermolide was validated by a disclosure from Novartis Pharmaceuticals, which detailed the production of sixty grams of the natural productxlij utilizing a hybrid of the Smith gram-scale approachxlie (Section 3.6) and the Paterson first-generation endgamexlig (Section 3.7). The material thus produced proved to be sufficient to permit Phase I clinical trials to evaluate the efficacy of (+)-discodermolide as an in vivo chemotherapeutic for adult patients presenting with advanced solid malignancies.
During the planning stages, the Novartis team carefully analyzed each of the existing synthetic routes (vide supra), evaluating parameters including the length of the sequence, overall chemical yield, potential for scalability, and cost of goods. From an economic perspective, the Smith gram-scale, triply-convergent strategy was particularly attractive, given the ready availability of the requisite common precursor (−)-14. At the time Novartis initiated their synthetic efforts, however, the Smith sequence had one major drawback, the ultra-high pressure reaction required for efficient formation of phosphonium salt (+)-71 (Scheme 13). To circumvent this transformation, the Novartis strategy ultimately constructed the C(9)-C(24) subunit exactly á la Smithxlie (Scheme 13) and then segued into the Paterson endgame,xlig albeit requiring the addition of two steps to the linear sequence (Scheme 21). Pleasingly, the Smith routes to amide (−)-14, alkyl iodide (+)-16, and vinyl iodide (+)-18 (Scheme 3, Scheme 12) proved to be readily amenable for production scale, with a few exceptions. The Paterson-type C(1)-C(6) subunit 120 was derived from the Smith common precursor (−)-14, via a route developed at Novartis.xlij Due to workup-related issues and safety concerns, pyrophoric aluminum reagents were replaced with more benign substitutes, a change which necessitated a two-step transamidation protocol for the formation of Weinreb amide (−)-14. Additionally, the use of the Swern oxidation was precluded due to the stench produced by methyl sulfide; TEMPO and bleach based oxidants were employed instead.
To effect the first key coupling, that of alkyl iodide (+)-16 with vinyl iodide (+)-18 (Scheme 21), the No-vartis process group initially examined zinc-mediated palladium(0)-catalyzed cross-coupling under the modified Negishi conditions used successfully by our group.xlie However, while this process did generate the desired product in 62% yield (not shown), the reaction product proved to be quite difficult to purify on scale. Attention therefore turned to a boron-mediated Suzuki-typelxvii cross-coupling, similar to that used by Marshall (Section 3.4) for the construction of the C(14)-C(15) bond. This resulted in a much cleaner reaction mixture, with the desired trisubstituted olefin (+)-80 isolated in a reproducible yield of 73%.
At this point, transition to the Paterson endgame began with construction of (+)-105 (cf. Scheme 17). To this end, diene installation was achieved in three steps.lxvi Still-Gennari modified Horner-Wadsworth-Emmons reaction,xlvi and introduction of the C(19) carbamate then furnished aldehyde (+)-106. A total of 9 synthetic transformations were required for the Smith to Paterson transition. This set the stage for what the Novartis researchers described as “the most challenging phase of the whole campaign”, attachment of the C(1)-C(6) subunit 120. The Paterson first-generation synthesis called for a final union employing the reagent-controlled aldol addition of the (+)-diisopinocampheylboron enolate of a C(1)-C(6) ketone with aldehyde (+)-106 (Scheme 17). In similar fashion, the Novartis team would construct the C(6)-C(7) bond using methyl ketone 120 (Scheme 21) as the enolate coupling partner. In the event, this transformation required 6.6 equivalents (!) of the 120-derived enolate relative to (+)-106 and furnished at best 60% yield of the desired C(7) epimer, together with 15% of the undesired isomer. Unfortunately, this reaction also proved to be quite capricious. On several occasions none of the desired adduct was observed; instead multiple products derived from an initial isomerization of the C(8)-C(9) olefin of (+)-106 arose. While the mechanism for this isomerization remains undetermined, this problem was solved by replacing solid (+)-B-chlorodiisopinocampheylborane (DIP-Cl) with a commercially available 70% solution in hexane. On production scale, the instability of the aldol adduct/boron complex to the original workup conditions also proved to be detrimental to the overall efficiency of the reaction. This puzzle was eventually solved simply by subjecting the crude reaction mixture directly to reverse-phase silica gel chromatography. Ultimately, the large-scale boron-mediated aldol union proceeded with a reproducible yield of 65%, albeit with a diastereo-selectivity of only 4:1. Equally disappointing, the transformation still required greater than a 6-fold excess of the C(1)-C(6) component 120
What remained was the directed reduction of the C(5) ketone and global deprotection to afford (+)-discodermolide with a longest linear sequence of 26 steps (33 total steps), and an overall yield of 0.65%. While far from optimal, the Novartis-Smith-Paterson hybrid sequence was exploited for the production of over sixty grams of the natural product. This amazing feat required the efforts of 43 chemists over a period of 20 months. The material thus generated proved to be sufficient for evaluation of (+)-discodermolide in early-stage human clinical trials.
The most recent total synthesis of (+)-discodermolide comes from the research group of Panek,xlik whose approach to the polypropionate framework is built largely upon the stereoselective crotylsilane-based C-C bond construction protocol developed in the Panek Laboratory.lxxvii Here, control of the stereochemical relationships, both absolute and relative, depends upon the choice of the crotylsilane reagents and Lewis acids used for the particular transformation. Accordingly, retrosynthetic disconnections of the C(6)-C(7) and C(14)-C(15) σ-bonds led to alkyl iodide (+)-125, aldehyde (−)-127, and methyl ketone (−)-129, which in turn would be generated via treatment of crotylsilanes (+)-121, (−)-ent-121, and (−)-128, respectively, with an appropriately functionalized Roche ester-derived aldehyde (+)-27 (Scheme 22).
Synthesis of the Panek C(15)-C(24) subunit (+)-125 began with a highly diastereoselective Felkin-controlled additionlxxv of crotylsilane (+)-121 to aldehyde (+)-122, followed by acidic workup (Scheme 22). Protection of the resulting diol as the di-tert-butylsilylene acetal and ozonolytic olefin cleavage furnished aldehyde (+)-123, which in turn was subjected to a second stereoselective crotylation event, this time utilizing silane (−)-ent-121. Six transformations including installation of the diene employing the boron-mediated procedure of Tsai and Mattesonlxxviii then furnished alkyl iodide (+)-125.
In similar fashion, crotylation of benzyl aldehyde (+)-27 with silane (−)-ent-121 began the synthesis of the C(7)-C(14) subunit (−)-127. A four step conversion to TMS-alkyne (−)-126, followed by hydrozirconation,lxxix iodination, and palladium(0)-mediated methylation, completed the construction of the C(13)-C(14) trisubstituted olefin, wherein the Z-vinylsilane functionality would serve as a masked vinyl iodide.lxxx Working next at the opposite terminus of the fragment, debenzylation and oxidation were followed by a second Corey-Fuchs homologation,lxix this time with subsequent formylation to afford the C(7)-C(14) aldehyde (−)-127. Production of C(1)-C(6) fragment (−)-129 began with a highly selective Felkin-controlledlxxv crotylation of aldehyde (+)-122 (Scheme 22). In this case, the use of silane (−)-128, which bears an additional methyl group, ultimately afforded the desired methyl ketone (−)-129 after acetalization and oxidative olefin cleavage.
In the Panek approach, construction of the C(6)-C(7) bond was effected via a dialkylboron-mediated, substrate-controlled acetate aldol reaction between methyl ketone (−)-129 and aldehyde (−)-127 (Scheme 23). Precedent had demonstrated that the desired 1,5-anti sense of stereoinduction should pertain.lxxxi,lxxxii Indeed, a variety of dialkylboron triflate reagents and solvent conditions all led to high levels of diastereoselectivity favoring the desired 1,5-anti isomer (−)-130, with the best case proving to be that exploiting dibutylboron triflate as the boron component. With the C(5) stereogenicity in place, Evans-Tishchenko anti-selective reductionlxxxiii of β-hydroxy ketone (−)-131 induced the requisite C(7)-S configuration. Upon basic hydrolysis of the resulting β-hydroxybutyrate and silica gel chromatography, Panek and coworkers noted an unexpected acetal migration, which they turned to their advantage via a two-step oxidation/methylation sequence involving the newly liberated primary alcohol to furnish the corresponding methyl ester (−)-131. Protection of the C(7) hydroxyl as a MOM ether, partial reduction of the alkyne, and iododesilylationlxxxiv then set the stage for the next major fragment union in the form of a palladium(0)-catalyzed cross-coupling as Smith (Sections 3.2 and 3.6) and Marshall (Section 3.4) had employed. Initial attempts to effect the desired C-C bond formation with the C(11) hydroxyl protected as a TBS ether however met with failure, under both boron- and zinc-mediated conditions (not shown). After considerable experimentation, replacement of the C(11) TBS ether with a MOM ether, as in (+)-132, provided a viable reaction partner for the requisite cross-coupling (Scheme 23).
In the event, treatment of vinyl iodide (+)-132 with the boronate derived from alkyl iodide (+)-125 in the presence of catalytic PdCl2(dppf) and thallium ethoxidelxvii led to construction of the discodermolide carbon skeleton. A four-step sequence requiring installation of the C(19) carbamatel and global deprotection completed the synthesis, which comprised a longest linear sequence of 27 steps (42 total steps) and an overall yield of 2.1%.
The potential chemotherapeutic application of (+)-discodermolide has inspired a number of research programs directed towards the design, synthesis, and evaluation of analogues.lxxxi,ixb, ixc,lxxxv,lxxxvi,lxxxvii,lxxxviii The goals of such efforts typically fall into three major categories: (1) Identification of the minimum structural elements required for potent tumor cell growth suppression; (2) Development of an understanding of the structure-activity relationship, ideally one that would enable the rational design of additional more potent congeners; and (3) Structural simplification to lessen the production cost, thus facilitating pharmaceutical development. Highlights of various analogue programs are presented in the sections to follow.
In 1994, Schreiber and coworkers reported the biological evaluation of the two antipodes of discoder-molide.ixb Surprisingly, both the natural (1) and the unnatural (−) enantiomer (Figure 3, 134) inhibited cell proliferation, with IC50 values of 6 nM and 72 nM, respectively, in the MG63 human osteoblast cell line. The mechanism of inhibition for the two agents was distinguished via flow cytometric analysis, which demonstrated that cells treated with the natural (+)-discodermolide were arrested at the G2/M phase of the cell cycle, while (−)-discodermolide treated cells were arrested in S phase. Subsequent experiments employing radiolabeled discodermolides showed that the natural (+) congener exhibited specific binding to a then-unidentified receptor, while no specific binding was noted for (−)-discodermolide, even at micromolar concentrations. Additionally, the unnatural (−)-antipode did not compete for (+)-discodermolide binding, further suggesting that the two agents interact with distinct cellular targets. Several months later, in early 1996, concurrent reports from Schreiber, et al. xxv and the ter Haar laboratory,xxiv identified microtubules as the cellular receptor for (+)-discodermolide.
In the course of these investigations, Schreiber and coworkers devised and synthesized a number of structurally-varied analogues,ixc with the aim of further elucidating the interaction between (+)-discodermolide and microtubules. The Schreiber team found that substantial truncation of the carbon skeleton was severely detrimental (Figure 3); removal of either the C(1)-C(7) subunit, as in 135, or the C(16)-C(24) subunit, as in 136, resulted in compounds that exhibited no cell growth inhibitory activity. Subtle simplification was however tolerated, as the 16-normethyl congener (+)-137 retained low-nanomolar potency, while epimerization of the C(16) stereogenic center (as in 138) led to diminished activity. Likewise, some modification was acceptable at C(17), as demonstrated by the acetyl congener 139. Conversely, epimerization of C(17), as in 140, resulted in complete loss of cell growth inhibition.
Examining next the C(1)-C(6) subunit, neither the lactone nor even the lactone oxidation state were found to be required for effective cell growth inhibition. Indeed, thiophenyl acetals (+)-141 and (+)-142 (Figure 4) retained nearly identical potency to the natural product (+)-discodermolide (1). The 16-normethyl thiophenyl acetal counterpart (+)-143 was also highly active. In their search for an appropriate scaffold upon which to develop a discodermolide-like affinity binding probe, Schreiber and colleagues attached a long-chain substituent in a manner that did not significantly interfere with the ligand-receptor interaction. To this end, three compounds were developed (144, 145, and 146), which revealed the effects of extension at C(24), C(17), and C(16), respectively. Appendages at the C(24) or C(17) positions resulted in only a moderate decrease in cytotoxicity, however extension of the C(16) methyl was not tolerated.
The Schreiber analogues began to elucidate the complex structure-activity relationships of (+)-discodermolide with tubulin. First, the stereogenicity of C(16) and C(17) proved to be of great importance, but the C(16) methyl substituent could be removed with little impact. Second, Schreiber demonstrated that either the lactone region could be modified, or the terminal diene could be extended, without severely diminishing activity. These observations laid the groundwork for the development of future analogues of (+)-discodermolide, most notably by the research groups of Smith, Paterson, and Curran, and also by researchers at the Harbor Branch Oceanographic Institute, Novartis Pharmaceuticals, and Kosan Biosciences, Inc.
As the group responsible for the isolation of the natural product,viii Gunasekera and coworkers at Harbor Branch were in a unique position to use naturally occurring (+)-discodermolide as a starting point for the development of semi-synthetic analogues (Figure 5). Their initial efforts focused on the preparation of acetylated congeners (147–152),lxxxva,b while a later report disclosed a series of saturated congeners (153–156) resulting from platinum oxide promoted hydrogenation of (+)-discodermolide.lxxxvd In vitro cell growth proliferation assay against several cancer cell lines revealed considerable structure-activity relationship vis-à-vis the four pendant hydroxyl groups, as well as the various olefins embedded in the discodermolide backbone.
For example, acetylation of either the C(3) hydroxyl, the C(7) hydroxyl, or both concurrently, as in (+)-147, (+)-148, and (+)-149, respectively, imparts an increase in cytotoxicity relative to the parent (+)-discodermolide (Table 1).lxxxixConversely, incorporation of acetyl substituents at either the C(11) hydroxyl, as in (+)-150 or (+)-151, or at the C(17) hydroxyl, as in (+)-152 led to a dramatic reduction in potency. Subsequent to this report, the Harbor Branch team also prepared a series of α,β-unsaturated lactone derivativeslxxxvd (not shown), employing the acetylated congeners as a starting point. The results reinforced the observation that free hydroxyl groups at the C(11) and C(17) positions are required for cytotoxicity. With respect to the saturated variants, fully saturated congener (−)-153 was completely inactive against the cell lines tested, while monoene (−)-154, which retains only the C(13)-C(14) trisubstituted olefin of (+)-discodermolide, exhibited modest nanomolar cytotoxicity in all but one cell line (Table 1). An improvement in activity was realized in the A549, P388, and MIP-100 cell lines upon reincorporation of the C(8)-C(9) cis-olefin, as in 155, but activity in the remaining cell lines suffered. Diminished potency was also noted upon removal of the C(7) hydroxyl from monoene (−)-154 to give (−)-156.
The Harbor Branch team also prepared a series of α,β-unsaturated discodermolide derivatives, employing the acetylated congeners as a starting point (Figure 6, Table 2).lxxxvd The most potent compound isolated during these studies was the same (+)-2,3-anhydrodiscodermolide 157 that had been observed previously by Smith and coworkers as a low level byproduct in their global deprotection (See Section 4.7.1). Also produced were α, β-unsaturated analogues 158–161, which comprise varying patterns of acetylation. In this series, incorporation of the acetyl moiety at C(11) [as in (+)-158] diminished the activity somewhat, and acetylation at C(17) [as in (+)-159] completely abolished activity, following the same trend noted in the saturated lactone series (Section 4.3). Not surprisingly, (+)-160 and (+)-161 were essentially inactive as well. In a separate set of experiments, two truncated derivatives of discodermolide (162 and 163) were also generated, but these compounds likewise displayed no cell growth inhibition.
Some twelve years after the isolation of (+)-discodermolide from the marine sponge Discodermia disso-luta,viii researchers at Harbor Branch reported the isolation and structural elucidation of several new naturally occurring discodermolide congeners from sponge samples of the Discodermia genus (164–168, Figure 7).lxxxvc Evaluation of the newly discovered congeners (Table 3) revealed that epimerization or removal of the C(2) methyl substituent [(+)-164 and (+)-165] moderately diminished the activity. Additionally, the linear discodermolide methyl ester (+)-166 and the C(19)-descarbamoyl analogue (+)-167 retained only modest potency relative to (+)-discodermolide, while the cyclic analogue (+)-168 was inactive.
Shortly after the completion of their first-generation total synthesis of (+)-discodermolide,xlig Paterson and coworkers disclosed the synthesis of three stereoisomers (169–171) and two truncated analogues (172 and 173) of the natural product (Figure 8).lxxxi A follow-up report detailed the production of three additional discodermolide congeners (174–176).lxxxvi All of these compounds were produced by exploiting intermediates and/or side products generated during their discodermolide syntheses. To date, biological activity has been reported for only four of the eight compounds (Table 4).lxxxix Here, potency was reduced, on average, by over an order of magnitude for C(5) epimer (+)-169 and C(7) epimer (+)-170, while deletion of the C(15) to C(24) subunit completely abolished activity. Remarkably, triol (+)-173, in which the entire C(1)-C(7) portion has been replaced with a hydroxymethyl substituent, exhibits antiproliferative activity comparable to the C(5) and C(7) epimeric congeners 169 and 170. This result contrasts with the Schreiber observation, wherein a very similar truncated congener 135 (Figure 3), missing only the hydroxymethyl substituent, was completely inactive.
In support of their discodermolide program, which ultimately resulted in the advancement of the natural product into Phase I clinical trials, researchers at Novartis Pharmaceuticals also designed, synthesized, and evaluated a number of synthetic analogues. During this program, variation was incorporated at each of the three major subunits of discodermolide [specifically, the C(1)-C(7), C(8)-C(14), and C(15)-C(24) regions]. The compounds and the data to result from this effort were first reported in a 2002 meeting of the American Association for Cancer Research,xc and also in a series of patents published in 2002 and 2003.xci
With respect to the C(1)-C(7) lactone subunit, the Novartis group found that the completely unsubstituted lactone congener 177 (Figure 9, Table 5) displayed cell growth inhibitory activity rivaling that of the natural product, and the sodium salt of discodermolide seco-acid 178 retained modest potency. Replacement of the C(1)-C(7) moiety with a variety of other substituents, as in carbamate 179, phenyl ether 180, or amides 181 and 182, while not completely debilitating, did result in lessened activity. Lactone 183 and truncated congener 184 were essentially inactive.
The next set of analogues from the Novartis team explored the effect of saturation, truncation, or replacement of the C(21)-C(24) terminal diene (185–190, Figure 10). On the whole, these modifications were well tolerated (Table 6), an observation which was also noted by Schreiber (Section 4.2), Harbor Branch (Section 4.3), and Smith (Section 4.7.2) in similar systems. Dihydro congeners 185, 186, and 187 were particularly potent, exhibiting single-digit nanomolar activity in four of the five tested cell lines. Triene 188, which lacks carbons C(23) and C(24), displayed good to excellent antiproliferative activity across the panel. Replacement of the diene with either a phenyl or benzyl ether, as in 189 and 190, respectively, had a major lowering effect on potency only in the MIP-101 cell line. Two final compounds described by Novartis incorporated modification of the critical central region of (+)-discodermolide. Originally reported by Smith and coworkers (see Section 4.7.2), 14-normethyldiscodermolide 191 retains low-nanomolar potency in four of the five cell lines tested at Novartis. On the other hand, compound 192, in which the N-methyl amide was intended to mimic the C(13)-C(15) substructure of the natural product, displayed sub-micromolar activity only in the A549 cell line.
A final series of compounds to be described by the Novartis group were evaluated and reported in collaboration with Harbor Branch.lxxxve During the final step of the Novartis campaign to produce sixty grams of (+)-discodermolide, an HCI-mediated global desilylation, several additional compounds in trace quantities were subsequently purified and characterized by Gunasekera and coworkers. Of the six derivatives isolated (Figure 11, Table 7), five (193–197) can be seen to have arisen from an acid-promoted C(13)-C(9) olefin-assisted cyclization with concomitant displacement of the allylic C(7) hydroxyl group. The multitude of products formed suggests the intermediacy of a tertiary carbocation, which either undergoes E1-type elimination [as in (−)-193 and (+)-194], or alternately is trapped by water to afford 195–197. The sixth compound, (+)-198, resulted from direct displacement of the C(7) hydroxyl by the C(11) substituent.
Upon evaluation, the bicyclic discodermolide congeners were all less potent than the natural product (+)-1. However, one compound retained modest cell growth inhibitory activity, specifically (+)-194, which bears an exomethylene at C(14).
In a joint research program at the University of Pittsburgh, Curran (Chemistry Department) and Day (Pharmacy) engaged in the design and synthesis of simplified discodermolide analogues.lxxxviii The simplest compounds produced, which possessed the general structure defined by skeleton 199 (Figure 12), exhibited at best minimal cell growth inhibitory potency (on the order of double digit micromolar).Many of these compounds did, however, retain properties unique to (+)-discodermolide, such as the ability to hypernucleate tubulin polymerization at low temperature.lxxxviiid The most potent antiproliferatives to result from the Curran/Day work (Figure 12, Table 8), although still only in the low micromolar regime, were those that retained a C(1)-C(6) cyclic subunit and the C(21)-C(24) terminal diene,lxxxviiid such as (+)-200a, (+)-200b, and (+)-201c.
Following the successful effort to produce one gram of totally synthetic (+)-discodermolide, the discodermolide program at Penn broadened to include the production of structurally related analogues. This venture was significantly strengthened by a trio of very productive collaborations, both academic and industrial. Horwitz and coworkers at the Albert Einstein School of Medicine contributed much of the cytotoxicity data that is presented in this section. Moreover, the Horwitz group continues to actively pursue elucidation of the discodermolide binding domain of β-tubulin. In 2002, Smith and coworkers also initiated a productive partnership with Myles at Kosan Biosciences, which resulted in many of the analogues and most of the biological data described in this section. Finally, a collaboration with Vogelstein at The Johns Hopkins University (Howard Hughes Medical Institute) permitted in vivo evaluation of (+)-discodermolide and several analogues both as stand-alone chemotherapeutics and as the chemical component of the treatment paradigm known as Combination Bacteriolytic Therapy (COBALT).
The design of the initial set of analogueslxxxviia was guided largely by knowledge gained during the gram-scale effort; the first compounds created were drawn directly from minor byproducts isolated during the latter stages of that endeavor, while many of the later analogues would address the synthetic difficulties encountered therein. Additionally, several (+)-discodermolide congeners were created simply to probe the structure/activity relationship. Tactically, the Smith/Kosan groups addressed the discodermolide skeleton at four key points: (1) the C(1)-C(6) lactone, (2) the C(14) methyl substituent, (3) the C(19) carbamate moiety, and (4) the terminal diene subunit.
The first analogue to be isolated arose as a minor byproduct of global desilylationlxxxviia during the final stage of the Smith gram-scale discodermolide synthesis. While the deprotection proceeded reliably and in high yield (>90%), the scale of those reactions permitted the isolation of significant quantities of a secondary product. Single-crystal X-ray analysis revealed this compound to be a,β-unsaturated lactone (+)-157 (Figure 13), the result of an acid-mediated dehydration event. Interestingly, despite the simplified structure, subsequent biological evaluation against a number of human cancer cell lines established that 2,3-anhydrodiscodermolide (+)-157 was generally a more potent inhibitor of cell growth than was (+)-discodermolide (Table 9). Smith and coworkers therefore embarked on a systematic investigation of the structure-activity relationship of the lactone region.lxxxviie The results indicated that the lactone moiety could indeed be further simplified, while retaining considerable potency (Table 9); both (+)-2-normethyl-2,3-anhydrodiscodermolide (203) and (+)-2,4-normethyl-2,3-anhydrodiscodermolide (204) show activity comparable to (+)-157, and thus superior to (+)-discodermolide (1). The α,β-unsaturation also proved to be unnecessary, as (+)-2-normethyl-3-deoxydiscodermolide 205 possessed cell growth inhibitory activity on par with the unsaturated congeners, while the unsubstituted, saturated lactone congener (+)-206 (also prepared by Novartis) was less potent only in the NCI/ADR cell line. Removal of the C(7) hydroxyl group, as in (+)-207 and (+)-208, was moderately detrimental to cytotoxicity.
An interesting discovery was the potency of (+)-4,5-epi-2,3-anhydrodiscodermolide 209. This observation led to the hypothesis that only the lactone carbonyl is involved in a critical interaction, and that the function of the ring substituents is simply to direct the orientation of the carbonyl. In other words, the 4,5-epi congener (+)-209 was able to competently preserve the lactone position, and thus retain potent bioactivity. Under the notion that the conformationally more rigid five-membered ring might provide a more effective scaffold, three y-lactone discodermolide congeners were synthesized and evaluated by our colleagues at Kosan (Figure 13). Upon evaluation, butyrolactone 210, which bears the C(5) and C(7) stereogenicity (using discodermolide numbering) found in the natural product, exhibited cell growth inhibitory activity superior to discodermolide in the drug-sensitive cell lines, while maintaining potency against the multi-drug resistant cell line (Table 9). In contrast, the 5,7-bis-epimer 211 was significantly less potent.
Smith, et al. speculated that the lactone region could perhaps be simplified further while preserving significant cell growth inhibitory activity. To explore this idea, several discodermolide congeners in which the entire C(1)-C(7) subunit has been replaced with arylethyl substituents (Figure 14) were designed and synthesized.lxxxviib Pleasingly, despite the removal of five stereogenic centers, the majority of these analogues displayed sub-micromolar activity against the drug-sensitive cell lines; moreover, several retain activity in the resistant line (Table 10). The simple phenol derivative, (+)-213, proved to be the most potent of the aryl lactone-replacement congeners, exhibiting antiproliferative potency not far removed from that of (+)-discodermolide. Taken together, these observations could have a significant impact on both the cost and ease of synthesizing a discodermolide derivative as a pharmaceutical.
Seeking to retain the synthetic simplicity of the aryl analogues shown in Figure 14, Smith, et al. returned to the hypothesis that, with respect to the C(1) – C(7) subunit, it is the placement and orientation of the C(1) car-bonyl that drives potency. With this in mind, a series of coumarin and lactam-derived compounds (225–230),lxxxviig was designed and synthesized to probe further the geometric requirements for antiproliferative activity. Again, the placement of the lactone carbonyl proved critical, with the 7-substituted coumarin (+)-225 and the 23,24-dihydro counterpart (+)-226 exhibiting cell growth inhibitory activity equivalent to (+)-discodermolide (1) in all three cell lines tested, while the 6-coumarin (+)-227 was significantly less potent. Notably, analogues (+)-225 and (+)-226 represent the structurally simplest congeners reported to date that retain the low-nanomolar antiproliferative activity exhibited by (+)-discodermolide. On the other hand, the three lactam analogues proved somewhat diminished in potency relative to the coumarins, although lactam (−)-228, which bears a free NH, is more active than the two N-substituted congeners (+)-229 and (+)-230.
The C(13)-C(14) Z-trisubstituted olefin embedded in the parent (+)-discodermolide represented a significant synthetic challenge. Indeed, the efficient installation of this structural motif was (and remains) among the most challenging aspects of the total synthesis. Deletion of the C(14) methyl group, generating a Z-disubstituted olefin, served to simplify the structure, and, more importantly, to simplify the chemistry required for construction. Adding to the merit of this strategy, Schreiber and coworkers had noted that deletion of the nearby C(16) methyl group had no negative impact on activity (See Section 4.2).Gratifyingly, totally synthetic (+)-14-normethyldiscodermolidelxxxviia (191, Figure 16) retained potent cell growth inhibition in three of the four cell lines tested (Table 12). Unlike (+)-discodermolide (1), however, the 14-normethyl congener (+)-191 is essentially inactive at submicromolar levels in the multi-drug resistant NCI/ADR line. This deficiency notwithstanding, the relative ease of synthesis and the potent cytotoxicity of (+)-191 made the 14-normethyl scaffold an attractive starting point for the design of a number of analogues.
This investigation began with production of simplified lactone-replacement analogues (231–235, Figure 16),lxxxviic mimicking the modifications made in the 14-methyl series. Interestingly, while dehydration of the lactone of (+)-discodermolide (1) to form (+)-157 (Figure 13, Table 9) led to a more potent compound in the three drug-sensitive cell lines, identical modification of the 14-normethyl scaffold, as in (+)-231, results in a significant decrease in tumor cell growth inhibition (Table 12). Further simplification by removal of the C(2) methyl group [(+)-232] restores potency to the levels displayed by the parent (+)-14-normethyldiscodermolide (191) in the MCF-7 and SKOV-3 lines, while activity in the A549 cell line is relatively depressed. An even more striking contrast between the discodermolide and the 14-normethyl scaffolds appears when the hydroxyphenylethylene unit replaces the lactone. While the 14-methylated congener of this compound [(+)-213, Figure 14, Table 10] displays considerable potency, no cytotoxicity is observed with the 14-normethyl variant (+)-233 at levels up to 5 µM. Replacement of the phenol with an aniline moiety, as in (+)-234, likewise leads to an inactive congener, as does substitution of the lactone with a pyridone ring [(+)-234].
Smith and coworkers next sought to explore the effect of saturation and truncation of the C(21)-C(24) terminal diene of (+)-14-normethyldiscodermolide (191).lxxxviia,c To this end, a series of four analogues were prepared (236–239, Figure 17). In vitro evaluation of the tumor cell growth inhibitory activity of this series again demonstrated the impotency of the 14-normethyl variants of (+)-discodermolide (1) in the multidrug resistant NCI/ADR cell line, as all four congeners proved essentially inactive at submicromolar levels (Table 13). In fact, the most truncated analogues, (+)-236 and (+)-237, display no appreciable cytotoxicity in any of the cell lines tested. Against MCF-7 and SKOV-3 cells, reincorporation of C(23) in the form of triene (+)-238 restored moderate cytotoxicity, while (+)-239, which retains the full carbon skeleton of (+)-14-normethyldiscodermolide, but with a saturated terminus, displayed potency rivaling that of the natural product. Similar results were noted by Novartis in systems with the C(14) methyl intact (Section 4.5). Taken together, these results clearly exhibit a trend of decreasing activity as the C(24) terminus is truncated, while isosteric replacement of the terminal olefin with a saturated counterpart preserves cell growth inhibitory activity.
In an effort to find modifications which might improve either the activity or the physicochemical properties of discodermolide, the influence of substitution at the C(19) carbamate was explored next. The most extensive series of carbamate-substitution congeners was generated employing the synthetically simplified 14-normethyl scaffold (Figure 18).lxxxviia,d The cell growth inhibitory activity data gathered during this study (Table 14) demonstrated that appending the carbamate in the 14-normethyl series (240–250), by and large, had little effect on cytotoxic potency across the panel of cell lines tested, with dimethylaniline congener (+)-245 exhibiting the best activity. A significant loss in potency was noted with phenylsulfone congener (+)-250, while complete removal of the carbamate, leaving a free hydroxyl group at C(19) [(+)-251, not depicted), likewise lowered the activity. Curiously, carbamate substitutions applied to the 2,3-anhydro scaffold, as in 252–255 (Figure 18) were detrimental to potency.lxxxviiad The dimethyl aniline congener (+)-256, however, exhibited activity comparable to that of 2,3-anhydrodiscodermolide 157. Moving to the parent discodermolide skeleton, the dimethylaniline substituent as in (+)-259 again proved to be the most promising addition to the carbamate.lxxxviid However, the dimethylaniline moiety is known to be labile in vivo, primarily due to the action of cytochrome P450. Thus, the less labile imidazole (+)-260 and morpholine (+)-261 were designed and synthesized; unfortunately the resulting cytotoxicity did not equal that shown by (+)-259.
In collaboration with the Vogelstein group at The Johns Hopkins University (Howard Hughes Medical Institute), (+)-discodermolide (1) and three of the most interesting analogues [(+)-191, (+)-157, and (+)-203, Figure 19] generated by the Smith group were evaluated for antitumor efficacy in an in vivo mouse model, specifically employing HCT-116 colon carcinoma xenografts. Each of these agents served to slow the growth of the treated tumors upon a single administration of the drug. The most potent of these agents, 2,3-anhydrodiscodermolide (+)-157, completely halted tumor growth for a full 20 days. Importantly, little general toxicity was observed during these studies.
Based on these single-agent studies, 2,3-anhydrodiscodermolide (+)-157 was next employed as the chemical component of the Vogelstein therapeutic paradigm COBALT (combination bacteriolytic therapy), which has proven remarkably effective in human xenograft mouse models for the treatment of fast-growing, poorly vascularized, solid tumors.xcii,xciii This treament protocol entailed administration of a combination of (+)-157 and the genetically-modified, non-toxic anaerobic bacteria C. novyi-NT, in this case to mice bearing HCT-116 xenograft tumors. Remarkably, the treated tumors exhibited massive hemorrhagic necrosis within one day following treatment.lxxxviif These lesions gradually scarred over and healed during the next three weeks; ultimately, complete tumor regression was observed, after only a single administration of the combination therapy.
As the library of discodermolide analogues grows, so too does our understanding of the structure-activity relationship of this most interesting antitumor agent. For example, the data currently available suggest that the conformational restrictions imparted by the Z-olefins at C(8)-C(9) and, particularly, C(13)-C(14) are critical to maintain potent cell growth inhibitory activity. Likewise, the perturbation caused by manipulation of the C(11) and/or C(17) hydroxyl groups, either in terms of stereogenicity or functionality, proved quite detrimental to activity, a result that is not surprising given the role that these substituents play in governing the conformation of (+)-discodermolide through non-bonded interactions.xxviii,xxix Conversely, some discodermolide substructures could indeed withstand significant modification or group deletion. For example, the C(19) carbamate and the C(23)-C(24) olefin could be completely removed while maintaining potency; saturation of the terminal diene was also acceptable. Alternatively either the carbamate or the diene could be appended with a variety of substituents with little impact on inhibitory potency. Removal of the C(14) or C(16) methyl groups also had little effect on activity in most of the cell lines evaluated. Perhaps most notably, the data reveal that the C(1)-C(7) region could be considerably simplified, while retaining extremely potent cell growth inhibitory activity. The latter modifications, while providing useful structure-activity information, also serve to ease somewhat the synthetic difficulty associated with production of highly active discodermolide-related compounds.
A number of curious structure-activity effects were noted in analogues incorporating multiple modifications that independently had little impact on cytotoxicity. For example, while removal of the C(19) carbamate, truncation of the diene, and simplification of the C(1)-C(7) lactone all have little or no detrimental effect on cytotoxicity when the C(14) methyl group is intact, identical changes to the 14-normethyl scaffold render the resulting discodermolide congeners completely devoid of activity at the sub-micromolar level. Likewise, the 2,3-anhydro skeleton seem to be quite sensitive to substitution at the carbamate, while the 14-normethyl and the parent discodermolide systems tolerate changes to this region.
With respect to the underlying biology, a number of factors account for the potent cytotoxicity exhibited by (+)-discodermolide. The most striking is the unparalleled degree of tubulin hypernucleation promoted by this agent, even at low temperatures and in the absence of helper proteins. Discodermolide is also not a substrate for P-glycoprotein, and thus retains activity against paclitaxel-resistant and multi-drug resistant cell lines. For reasons not yet fully understood, discodermolide and paclitaxel actually form a synergistic drug combination, wherein each is much more potent in the presence of low concentrations of the other than would be predicted from the individual IC50 values. Furthermore, it is now known that discodermolide possesses a second mode of cytotoxicity, specifically the induction of an accelerated senescence phenotype, which is not observed with paclitaxel or other members of this class of microtubule stabilizing agents. Separately, the potential utility of microtubule stabilizing natural products, including (+)-discodermolide, as neuroprotective agents is only beginning to be explored.
From the synthetic perspective, methods for the production of (+)-discodermolide have become increasingly shorter and more efficient. Indeed, the production of sixty grams of the natural product by Novartis showcased the increasing potential of total synthesis to provide clinically relevant quantities of complex natural products. In a broader sense, the Novartis work is simply the continuation of an accelerating trend in pharmaceutical development; over 60% of the small-molecule drugs introduced to the market between 1981 and 2002 either were, or were derived from, naturally occurring compounds.xciv Furthermore, our investigation of nature’s reserve of interesting, medicinally useful compounds has barely scratched the surface. Going forward, it is likely that many potentially useful lead compounds will be identified that possess synthetically demanding architectures. Thus, future therapeutic breakthroughs will depend, at least in part, on the ability of synthetic chemists to rise to the challenge.
The discodermolide synthetic program in the Smith Laboratory was supported by the National Institutes of Health (National Cancer Institute and Institute of General Medical Sciences) through grants CA-19033 and GM-29028, by the U.S. Army through grant DAMD17-00-1-0404, and by Kosan Biosciences, Inc.
Amos B. Smith, III (born August 26, 1944) attended Bucknell University, where he was awarded the University’s first combined four-year B.S.-M.S. degree in Chemistry in 1966. After a year in medical school (University of Pennsylvania), he earned his Ph.D. degree (1972) and completed a year as a Research Associate at Rockefeller University. In 1973, he joined the Department of Chemistry at the University of Pennsylvania; currently, he is the Rhodes-Thompson Professor of Chemistry, a Member of the Monell Chemical Senses Center, and the Associate Director of the Penn Center for Molecular Discovery (PCMD). From 1988 to 1996 he served as Chairman of the Department of Chemistry and in 1998 he became the first Editor-in-Chief of the American Chemical Society’s journal, Organic Letters. In 2001, he was appointed as an Honorary Member of the Kitasato Institute (Tokyo, Japan). Smith's research interests, recorded in more than 530 publications, encompass three diverse areas: natural product synthesis, bioorganic chemistry and materials science. To date his Laboratory has achieved the total synthesis of more than 90 architecturally complex natural products. Smith has been a Visiting Professor at Columbia, Cambridge (UK) and Auckland (NZ) Universities and has received a number of awards, including the ACS Arthur C. Cope Scholar Award (1991), the Alexander von Humboldt Research Award for Senior U.S, Scientist (1992), the ACS Award for Creativity in Organic Chemistry (1997), the Centenary Medal, Royal Society of Chemistry (2002), the Yamada-Koga Prize (2003), the University of Pennsylvania first Provost’s Award for Distinguished Teaching and Mentoring of Ph.D. Students (2004), the Order of the Rising Sun, Gold Rays with Neck Ribbon from the Government of Japan (2004), Fellow in American Academy of the Arts and Sciences (2006), and the Royal Society of Chemistry Simonsen Medal (2008).
Brian Scott Freeze was born in Youngstown, OH in 1974. After receiving both a B.S in Biology and a Masters in Chemistry from Youngstown State University, he proceeded to earn a Ph.D. in 2005 under the direction of Professor Amos B. Smith, III at the University of Pennsylvania. His thesis research focused on the total synthesis of (+)-discodermolide and the production of novel analogues. He is currently employed as a medicinal chemist at Millennium Pharmaceuticals, Inc. in Cambridge, MA.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.