Introduction

Through aeons of evolution, nature has gifted us with a seemingly limitless source of important secondary metabolites. Such compounds are often astoundingly intricate in terms of their molecular architecture, with stereochemically elaborate scaffolds that dwarf structures conceived by mankind. Unsurprisingly, such extraordinary structures demand effective methodologies and strategies, along with hard work and perseverance, to ensure a successful outcome from a suitably focused synthetic campaign. Furthermore, the vanishingly low isolation yields of such natural products can preclude their full stereochemical assignment, rendering total synthesis a valuable tool for structural elucidation.1, 2, 3, 4, 5

Among the vast chemical space carved out by nature are the polyketides, typified by their dazzling array of functionality and stereochemistry, providing a testing intellectual challenge for the synthetic chemist. Enticed by these intriguing structures, which generally have impressive biological activities,6 our group has had a longstanding interest in the development of novel synthetic methods and strategies that are both robust and, where required, flexible. In this context, the efficiency of our suite of versatile boron-mediated aldol reactions has proved invaluable for the controlled installation of the highly oxygenated frameworks of these captivating natural products.7, 8, 9

In this account, we provide an overview of recent research endeavors that have culminated in the total synthesis of several challenging polyketide natural products with promising anticancer activity in our group: spirastrellolide A methyl ester (1), leiodermatolide (2), rhizopodin (3) and chivosazole F (4) (Figure 1). In particular, we highlight the unexpected obstacles encountered and subsequent discoveries that resulted in the successful total syntheses of these highly challenging targets.

Figure 1
figure 1

Structures of spirastrellolide A methyl ester (1), leiodermatolide (2), rhizopodin (3) and chivosazole F (4).

Full size image

Spirastrellolide A methyl ester

The spirastrellolides constitute an extraordinary family of spiroacetal macrolides first isolated by Andersen and co-workers10 in 2003 from extracts of the Caribbean sponge Spirastrella coccinea. The most abundant congener, spirastrellolide A (1) (isolated as the corresponding methyl ester), exhibits striking structural complexity, containing 20 stereocenters, a 38-membered macrolactone and a nine-carbon side chain featuring a (Z,E)-1,4-diene.11, 12, 13, 14, 15 The macrocycle itself contains a tetrahydropyran (A ring), a bicyclic 6,6-spiroacetal (BC rings) and a tricyclic 5,6,6-spiroacetal (DEF rings) featuring a chlorine atom at C28. Additionally, spirastrellolide A was found to exhibit potent antimitotic properties via selective protein phosphatase 2A inhibition (IC50=1 nM).10, 12 Beyond the obvious potential as a novel anticancer lead, such phosphatase inhibitors have also shown therapeutic promise in tackling obesity, autoimmune conditions and neurodegenerative disorders.16 The combination of the synthetic challenge posed by their architectural complexity and promising biological activity has rendered the spirastrellolides the focus of intense research efforts from numerous groups.15 Despite this, only five completed syntheses have been reported to date,17, 18, 19, 20 two of which are from our group.21, 22, 23, 24

Our efforts toward spirastrellolide A methyl ester began soon after disclosure of the originally proposed structure and our synthetic approach evolved concurrently with structural determination studies on this moving target.25 A flexible endgame was a strict requirement as a consequence of the ambiguity surrounding the C46 hydroxyl stereocenter. Specifically, our initial strategy in face of these imposed requirements involved a modular approach to macrocycle formation, resulting in the successful assembly of the complete ABCDEF ring system, followed by late-stage side-chain attachment to facilitate preparation of both possible C46 diastereomers.26

With advanced intermediate 5 (Scheme 1) in hand after a sustained campaign of dedicated efforts,22, 26 synthesis completion appeared tantalizingly close. Unfortunately, selective removal of the C40 silyl-protecting group to enable side-chain incorporation proved to be a major obstacle. In the end, a global deprotection, followed by protecting group adjustment, was required. Oxidation to the corresponding aldehyde 6 then proceeded smoothly and set the scene for homologation. At this point, a variety of organometallic addition reactions were trialed unsuccessfully. We surmised that these failures were likely to be a reflection of the steric constraints imposed on the C40 aldehyde by the proximal cage-like macrocycle. After exhaustive experimentation, it was found that a simple Wittig olefination reaction could be used to access a terminal alkene, thereby allowing side-chain incorporation via olefin cross-metathesis.27, 28 After considerable experimentation, the cross-metathesis with dicarbonate 7 required relatively forcing conditions (refluxing in benzene), due to the steric constraints imposed by the macrocycle. The resulting allylic carbonate 8 then allowed a π-allyl Stille cross-coupling with stannane 9 to afford the bis-acetonide protected natural product, which underwent a global deprotection to afford the first total synthesis of spirastrellolide A methyl ester (1).22 Notably, the 46-epi diastereomer of 1 showed distinctly different NMR spectra due to the influence of the proximate macrocycle.

Having successfully completed the target molecule, and thereby validating the configurational assignment, we next sought to improve our synthesis, paying specific attention to avoiding unnecessary redox steps and protecting group manipulations. In particular, the need for a divergent side-chain installation strategy was now deemed unnecessary with the stereochemistry of the natural product now unambiguously assigned. Additionally, we sought to capitalize on the availability of key fragments from our first-generation approach, giving rise to the revised retrosynthetic analysis in Scheme 2. Notably, we looked to establish the C1-C47 carbon backbone in 10 (from allylic carbonate 11 and stannane 12) in its entirety before macrolactonization, thereby simplifying incorporation of the (E,Z)-skipped diene side chain. Building on earlier work, the BC-spiroacetal moiety would be installed through 4-methoxybenzyl (PMB) deprotection/in situ spiroacetalization of a Z-enone arising from coupling of the C1-C16 alkyne fragment 13 and C17-C40 aldehyde 14.26 Disconnection across C24-C25 via an sp3-sp2 Suzuki coupling29 then reveals two intermediates utilized previously, C17-C24 vinyl iodide 15 and C25-C40 bis-spiroacetal 16.30

In our first-generation synthesis, the construction of the C26-C40 DEF bis-spiroacetal 17 via an acid-mediated deprotection/spiroacetalization cascade of 18 was a major bottleneck.31 The problem stemmed from formation of the undesired furan 19 via competing elimination (Scheme 3a). Even after extensive optimization, we could only generate a modest amount of the required C26-C40 DEF bis-spiroacetal 17. Thus, we needed to revise our strategy to achieve a reliable multigram supply of this essential fragment. First, we removed the appended γ-lactone in a bid to avoid competitive furan formation. Additionally, we noted that the bis-spiroacetal could arise from a tetraol linear precursor. In particular, we recognized that the sense of asymmetric induction via the Sharpless asymmetric dihydroxylation required to install the C37/C38 and C26/C27 hydroxyls was the same. This led to an adventurous double dihydroxylation/spiroacetalization cascade as in 20 to 21, which, if successful, would provide an elegant and efficient synthesis of the DEF bis-spiroacetal ring system (Scheme 3b).

The required linear precursor 20 (Scheme 4) was readily prepared from aldehyde 22 and ketone 23, notably using an Oehlschlager–Brown syn-choroallylation32 and our lactate aldol methodogy33 to set up the required stereocenters.24 A boron-mediated aldol reaction facilitated the fragment union to form β-hydroxyketone 24, which led onto the required linear precursor 20 via a four-step sequence.30 At this stage, we attempted the pivotal double asymmetric dihydroxylation.34 This initially afforded bis-hemiacetal 25, which to our delight spirocyclized under mild acidic conditions to afford the DEF bis-spiroacetal 21. Fortuitously, we discovered that other spirocyclic isomers of 21 could be resubmitted under acidic conditions to afford the required DEF bis-spiroacetal cleanly. This was only made possible by the increased stability of the DEF bis-spirocycle circumventing furan formation, giving us the opportunity to use thermodynamic equilibration rather than kinetic control. A final bis-silylation delivered the protected fragment 26 efficiently. Most importantly, this route facilitated a dependable multigram scale synthesis of the crucial C26-C40 bis-spiroacetal moiety.

Armed with an efficient and scalable route towards C1-C16 alkyne 13, C17-24 vinyl iodide 15 and now the C26-C40 bis-spiroacetal 26,26, 30 we set out to improve the fragment coupling sequence (Scheme 5).35 Preparation of C17-C40 aldehyde 14 commenced with a primary triethylsilyl (TES) deprotection, oxidation and methylenation to provide the corresponding C25-C40 alkene 16. Hydroboration of 16 followed by an in situ sp3-sp2 Suzuki cross-coupling with vinyl iodide 15 forged the C24-C25 bond and furnished diene 27 cleanly.29, 36 The final two stereocenters of the C17-C40 fragment were set up via a diastereoselective substrate-controlled double hydroboration sequence; installing the C17 and C23 hydroxyl groups and affording the required 23,24-anti stereochemistry in 28. Protecting group manipulations then yielded an advanced triol, which was subjected to a selective triple oxidation of the two primary alcohols with concomitant lactonization to afford the required C17-C40 aldehyde 14.

Our initial coupling strategy to form the C16-C17 bond hinged upon an the addition of an alkynyllithium species to the C17 aldehyde.26 However, this transformation now proved to be capricious owing to competing addition to the γ-lactone moiety. Instead, a Nozaki–Hiyama–Kishi coupling37, 38 between iodoalkyne 29 and aldehyde 14, to our delight, chemoselectively and reliably forged the C16-C17 bond (Scheme 6). The BC-spiroacetal formation commenced with a Lindlar reduction of the alkyne 30 and oxidation to the Z-enone. Subsequent bis-PMB deprotection under controlled conditions set the scene for a concomitant acetalization to cleanly forge the BC-spiroacetal ring system, now affording 31 with all the requisite ABCDEF rings in a stereodefined manner. With the carbon and oxygen skeleton for the macrocycle now in hand, our attention turned towards side-chain installation and the final macrolactonization. A selective primary TBS ether deprotection, partial reduction of the γ-lactone and vinylation afforded allylic alcohol 32, which was then treated with triphosgene to both temporarily mask the diol as well as providing the requisite leaving group for the π-allyl Stille cross-coupling. Pleasingly, the planned cross-coupling between allylic carbonate 11 and vinyl stannane 12 proceeded efficiently, and was a major improvement over our previous cross-metathesis route in the presence of the full macrocycle. With only the macrolactonization and global deprotection left, the finish line was now in sight. Once again, this transformation proved to be significantly more challenging than initially anticipated!

Frustratingly, subjecting seco-acid 10 to increasingly forcing conditions (including refluxing in toluene) for macrolactonization39, 40 not only failed to furnish the cyclized product, but returned degraded starting material. We initially surmised that the side chain was perhaps impeding the macrolactonization; however, diol 33 corresponding to the truncated macrocycle also failed to cyclize when subjected to previously established macrolactonization conditions (Scheme 7a). This unexpected difficulty was in stark contrast to the highly efficient macrolactonization (>95%) observed in our first-generation route (Scheme 7b), which we attributed to a degree of favorable conformational preorganization in the seco-acid 34. Comparison of the seco-acid 10 with that used previously highlighted only one seemingly minor structural difference—the (very distal) C23-TES ether. Therefore, we hypothesized that unfavorable conformational effects, presumably imposed by the additional silyl-protecting group, were operating to bias the free acid away from ring closing with the C37 alcohol. As such, we treated seco-acid 10 with pyridinium p-toluenesulfonate in methanol to effect controlled mono- and bis-TES ether cleavage. Our hypothesis was proven to be correct; submitting either of the mono- or bis-desilylated products (35 and 36) to standard Yamaguchi macrolactonization conditions now afforded macrocycles 37 and 38 in excellent yield (Scheme 7c). A final global deprotection completed our second-generation synthesis of spirastrellolide A methyl ester (1) in 23 linear steps and 6% overall yield from C26-C40 bis-spiroacetal 26. When compared with the first-generation synthesis (25 steps and 1% overall yield), it is pleasing to note the improvement in efficiency, both in terms of step count and yield. Moreover, we discovered that we were incredibly lucky in our first-generation synthesis—where the troublesome C23-TES ether was unintentionally cleaved in the BC-spiroacetalization step, which greatly assisted the crucial downstream macrolactonization reaction. An important lesson was learned here that protecting groups can have subtle and unpredictable conformational effects in such complex substrates!

Leiodermatolide

In 2008, leiodermatolide (2) was isolated from the lithistid sponge Leiodermatium sp. collected off the coast of Florida by the Wright group.41 Spectroscopic analysis illuminated the planar structure of 2 and revealed a 16-membered macrolactone containing a Z,Z-diene and a pendant carbamate group, as well as an E,E-diene on the side chain terminating in a δ-lactone. The assigned structure highlighted the presence of nine stereocenters; six of which lie in the macrocycle and three in the terminal δ-lactone.42 Biological evaluation showed that leiodermatolide exhibited potent anticancer activity, in particular against a range of drug-resistant cancer cell lines. While leiodermatolide-treated cells exhibited physiological responses often typified by tubulin-binding compounds, in vitro studies failed to show evidence for any direct tubulin interaction. As such, it was suggested that leiodermatolide acted via an indirect mechanism orthogonal to other known tubulin-targeting anticancer drugs, indicative of a promising anticancer drug candidate.

Our involvement with leiodermatolide was borne from its initially inconclusive stereochemical assignment. In collaboration with the Wright group, extensive NMR spectroscopic analysis, molecular modeling and computational DP4 NMR predictions43 allowed us to refine the structure to a single diastereomer for the C1-C16 macrocycle and the C21-C25 δ-lactone with >99% probability. Unfortunately, the distal nature of the C21-C25 δ-lactone relative to the macrocycle precluded a conclusive determination of the stereochemistry between these two stereoclusters, leading to four candidate stereoisomers for the natural product. To definitively pin down the stereochemistry of 2, we embarked on a synthetic campaign geared towards confirming the three-dimensional structure of the macrocycle followed by the full natural product. A synthesis-enabled stereochemical elucidation was a notion shared with other research groups,44, 45 which, to date, has resulted in one other group successfully synthesizing leiodermatolide.46, 47

As the absolute configuration was unknown, we arbitrarily targeted ent- 2 and its diastereomer for initial studies. Our initial approach towards 2 hinged upon a late-stage sp2-sp2 Suzuki coupling across C17-C18 to allow the flexible appendage of both enantiomers of the C18-C25 δ-lactone to the macrocycle, as shown in Scheme 8a. The C18-C25 δ-lactone 39 could be readily synthesized from either enantiomer of 40. We anticipated that the C1-C17 macrocycle 41 could be constructed from a linchpin bis-halide fragment 42, leveraging the more reactive vinyl iodide to selectively engage in a Stille cross-coupling with C1-C11 vinyl stannane 43.

In executing this approach (Scheme 8b), we discovered that the bis-TBS protection of the C7 and C9 hydroxyl groups required relatively forcing conditions to effect the second silylation at C7.48 This observation indicated the possibility of realizing a site-selective C9 carbamate installation in the endgame. Our resulting synthesis of the C1-C17 macrocycle 41 confirmed our relative stereochemical assignment through spectroscopic correlations.48 However, the specific rotation recorded for the macrocycle was opposite in sign to (–)-leiodermatolide, tentatively suggesting that we may have embarked in the wrong enantiomeric series. Additionally, there were two key issues we needed to address in the evolution of our synthetic strategy. First, while the semireduction of vinyl dibromide 44 to (Z)-vinyl bromide 45 proceeded smoothly, subsequent attempts at converting it into the vinyl stannane proved problematic. This involved cleavage of the C7 and C9-TBS ethers to afford diol 46, followed by stannylation under Wulff–Stille conditions49 to form stannane 43, albeit in a modest yield (Scheme 6b). Furthermore, despite preliminary results suggesting otherwise, our vision of a late-stage site-selective carbamate installation proved unrewarding; treatment of the macrocycle 47 with trichloroacetyl isocyanate50 resulted in a 3:2 mixture of regioisomeric products 41 and 48 that favored the undesired C7 carbamate 48. Moreover, attempts at realizing the key Suzuki coupling to afford the full leiodermatolide carbon skeleton proved fruitless; vinyl bromide 41 was found to be unreactive under a variety of palladium-catalyzed conditions.51

This intelligence gathering exercise prompted us to revise our synthetic strategy towards 2, as highlighted in Scheme 9, and we instead looked towards forming the fully elaborated macrocycle via a late-stage macrolactonization. As the C11-C12 bond was reliably installed via a Stille coupling, we sought to disconnect the molecule into the C1-C11 vinyl stannane ent- 43 and the C12-C25 δ-lactone 49. The C12-C25 fragment itself can then be constructed from vinyl iodide 50 and δ-lactone ent-39, using a Suzuki coupling to forge the C17-C18 bond. Despite disappointing initial results, we remained optimistic about effecting a regioselective carbamate formation, thereby minimizing protecting group manipulations.

Our revised synthesis of the C1-C11 stannane ent- 43 commenced from the Weinreb amide 51 derived from (R)-Roche ester (Scheme 10). By trapping the kinetic Z-enolate of the derived ketone with Comins’ reagent52 followed by a Suzuki-type methylation of vinyl triflate 52, we successfully formed the trisubstituted alkene 53.53 Using methodology developed by our group, a lactate aldol reaction between (R)-lactate-derived ethyl ketone (R)-40 and aldehyde 54 readily afforded the required anti adduct 55 with excellent diastereoselectivity.33, 54, 55 A four-step sequence installed the requisite alkyne and removed the lactate auxiliary. The required 1,3-anti reduction on ynone 56 proved problematic, using the Evans–Tishchenko56 protocol failed outright. Unfortunately, the Evans–Saksena reduction57 on the same substrate gave poor diastereoselectivity. This was rationalized based on the small size of the alkyne substituent, reducing its preference to occupy the equatorial position in the transition state. Based on this hypothesis, we looked towards increasing the steric bulk by preparing the Z-vinyl iodide 57.58 Luckily, this drastically improved the selectivity of the 1,3-anti reduction via the Evans–Saksena protocol, completing the required stereotetrad in the C1-C11 fragment. Next, lithiation followed by trapping with tributyltin chloride proceeded smoothly to give the C1-C11 stannane 58. Gratifyingly, this route was a significant improvement over our initial approach (20% yield over 14 steps, versus 6% yield over 14 steps).48, 59

The C12-C17 vinyl iodide was constructed again using our lactate aldol methodology, furnishing 50 in four steps (via aldol adduct 59) from ethyl ketone (S)-40 (Scheme 11).33, 54, 55 Synthesis of the C18-C25 δ-lactone commenced with a boron-mediated aldol reaction between ketone (R)-40 and propanal, affording the required anti adduct 60. The δ-lactone was constructed through a BF3·OEt2-mediated Mukaiyama aldol reaction between ketone 61 and silyl ketene acetal 62.8, 60 This, followed by an acid-mediated lactonization, delivered the δ-lactone 63, where the matched stereoinduction from 1,2-Felkin and 1,3-Evans polar models are mutually reinforcing.61 Subsequent silylation afforded the protected lactone 64, where a two-step sequence revealed the required vinyl boronate ent-39 in anticipation for the key cross-coupling.

The planned Suzuki coupling could be effected between vinyl iodide 50 and vinyl boronate ent-39. Advantageously, we discovered that the fragment union could also be readily achieved with excellent geometrical control via the Heck reaction to deliver 65 (Scheme 12),62 saving two steps in converting the alkene 64 to the vinyl boronate ent- 39 (vide supra). A two-step procedure revealed the required C12-C25 vinyl iodide 49, which underwent a facile Stille coupling63, 64 with stannane 58 to establish 66, corresponding to the full carbon skeleton of leiodermatolide. Finally, a series of redox and protecting group manipulations revealed the seco-acid 67, which was efficiently macrocyclized under our preferred Yamaguchi conditions39 to generate the 16-membered macrolactone.

With a global deprotection revealing the des-carbamoyl derivative of leiodermatolide 68, all that was required was the pivotal regioselective C9 carbamoylation. We surmised that the steric hindrance around C7 should heighten the reactivity of the C9 alcohol, a rationale supported by molecular modeling studies. As previously alluded to, treating the truncated macrolactone 47 with trichloroacetyl isocyanate50 favored the formation of the undesired C7 carbamate 48, with extensive experimentation failing to overturn this result. Interestingly enough, we observed that esterification or silylation proceeded with high selectivity at the C9 position. This hinted that it was indeed the more reactive position, with the carbamoylating agent behaving anomalously. Leveraging this finding, a sequence involving bis-silylation, selective C9 desilylation, followed by treatment with trichloroacetyl chloride and C7 desilylation successfully led to (–)-leiodermatolide (2) in 23 steps and 3.2% overall yield.59 Careful comparison with the authentic sample provided by the Wright group confirmed that they were identical in all respects. Serendipitously, this three-dimensional structure corresponds exactly to the one out of four stereoisomers arbitrarily rendered in our isolation paper.42 At this point, we could embark on a program of SAR studies and further biological evaluation of this promising anticancer lead structure.65

Rhizopodin

Rhizopodin (3) is an architecturally complex macrocyclic polyketide first isolated in 1993 by Reichenbach and co-workers66 from the myxobacterium Myxococcus stipitatus. By binding with and inhibiting actin polymerization, rhizopodin mediates potent antiproliferative activity as well as strong cytostatic effects against a range of cancer cell lines.67 This selective interaction with actin also enabled its structural elucidation, with X-ray crystallographic studies of the bound rhizopodin–actin complex revealing an intriguing C2-symmetric macrodiolide.68 From a structural perspective, 14 of the 18 stereocenters are embedded in the 38-membered macrolide core, together with two oxazole rings and two diene motifs, with the remaining four stereocenters located on the two side chains.69 The ornate architecture and promising anticancer profile of rhizopodin has rendered intensive research towards its total synthesis. Although several groups have reported the synthesis of various substructures,70, 71, 72, 73, 74 there has only been two completed total syntheses of the target structure itself.75, 76, 77

Our proposed synthesis (Scheme 13) of rhizopodin (3) centered on structural simplification into the truncated monomer 69 and known side-chain fragment 70.78 This disconnection provided a degree of flexibility, with macrocycle formation possible via direct or sequential esterification, followed by bidirectional aldol coupling with ketone 70 to incorporate the requisite side chain(s). Oxazole formation was envisaged via amide bond formation between C14-C22 acid 71 and C8-C13 amino alcohol 72 followed by dehydration, while diene installation was proposed using a Stille coupling of vinyl iodide 73 and a suitable C8 stannane.

As is often the case with complex polyketide synthesis, the strategic incorporation of orthogonal-protecting groups was of crucial importance. Initially, we envisaged incorporating PMB ethers to chemoselectively unmask the required alcohols for the macrolactonization and side-chain attachment. However, we found that an oxidative PMB ether cleavage using dichloro-5,6-dicyanobenzoquinone resulted in the concomitant oxidation of the C5 allylic methyl ether, with alternative Lewis acidic cleavage degrading our advanced intermediates.77 As such, we opted for a carefully selected combination of silyl-protecting groups. Notably, attempts at deprotecting the C16-OTBS ethers in the endgame resulted solely in eliminated product. Frustratingly also, attempts at deprotecting a primary C22-OTIPS ether to allow side-chain installation, in the presence of a secondary C16-OTES ether, resulted in simultaneous cleavage of both silyl groups. These difficulties ultimately forced us to opt for a riskier gamut of silyl-protecting groups in acid 71 and aldehyde 73 (vide infra).

Synthesis of the C14-C22 carboxylic acid 71 commenced with a Brown allylation onto Roche ester-derived aldehyde 74 (Scheme 14a).79 The remaining stereocenters in this fragment were generated first via a Mukaiyama aldol reaction between aldehyde 75 and silyl ketene acetal 76, setting up the C18 stereocenter, and a subsequent diastereoselective reduction80 of the cyclic ketone after methanolysis of dioxinone 77. From β-hydroxylactone 78, subsequent protections and oxidation afforded the C14-C22 acid 71. The amino alcohol coupling partner 72 required for the oxazole formation was formed from propargyl alcohol 79 (Scheme 14b). A Sharpless asymmetric epoxidation (yielding epoxide 80)81, 82 followed by amidation and regioselective epoxide opening gave oxazoline 81. A final sequence of methylation and hydrolysis then delivered the amino alcohol 72.

The final C1-C7 fragment 73 required for the macrocycle was obtained by an enantioselective Mukaiyama aldol reaction between aldehyde 82 and Chan’s diene (83) (Scheme 14c).83, 84 Subsequent methanolysis of dioxinone 84 followed by a Narasaka reduction85 generated the free diol. Protecting group manipulations and a final methylation of the free C5-OH then afforded the required Stille coupling partner 73.

Fragment assembly commenced with an amide bond formation between carboxylic acid 71 and amino alcohol 72 (Scheme 15). Using modified Robinson-Gabriel conditions developed by Wipf and Graham,86 oxazole 85 was formed cleanly. Subsequent stannylation afforded vinyl stannane 86, which was coupled with vinyl iodide 73 via a Stille cross-coupling63 to give the truncated monomer in anticipation for the key macrocyclization step. At this stage, we discovered that a series of oxidation state adjustments and protecting group manipulations were critical for the success of the macrocycle formation. While conditions required for methyl ester hydrolysis concomitantly unmasked the required C18-OH, Yamaguchi macrolactonization conditions39 disappointingly afforded a mixture of oligomers, primarily corresponding to the monomeric truncate. As such, we were forced to adopt a stepwise approach to access both coupling partners for the macrolactonization. A controlled reduction to the aldehyde 87 therefore was performed, meaning that this key intermediate could be subjected to either a controlled C18-OTMS desilylation (88) or a Pinnick oxidation to afford seco-acid 89.

Following these maneuvers, a selective esterification between alcohol 88 and seco-acid 89 served to complete the linear carbon skeleton in 90 (Scheme 16). This was followed by a similar sequence of desilylation, oxidation and macrolactonization to close the required macrocycle 91. The C2-symmetry of the molecule presented the opportunity of performing a bidirectional side-chain installation in the endgame. This required a selective C22-22′ primary TBS ether cleavage, a capricious operation owing to the presence of multiple secondary silyl-protecting groups of similar lability. In the end, carefully controlled exposure of the protected macrocycle 91 to HF/py selectively afforded the C22/22′ diol 92 which, following oxidation, underwent a double boron-mediated aldol addition with ketone 70.78 Drawing from our reidispongiolide synthesis,78 a sequence involving a controlled dehydration,87 followed by a conjugate reduction and global deprotection concluded our synthesis of rhizopodin (3) in 29 steps in 0.2% overall yield.77 The eventual success of this project required judicious fine-tuning of the protecting group strategy and redox steps, emphasizing the need for perseverance based on a flexible synthesis plan.

Chivosazole F

Following their discovery of rhizopodin, Reichenbach and co-workers88 and Höfle and co-workers89 reported the isolation of chivosazoles A-F from the myxobacterium Sorangium cellulosum in 1995. The chivosazoles are a structurally unprecedented class of polyene macrolides, with each member of the family differing in terms of the substitution at C11 and C20. Notably, the chivosazole family displayed potent inhibitory activity against filamentous fungi, yeast and a panel of human cancer cell lines. This bioactivity stems from its selective inhibition of actin polymerization. Intriguingly, the lack of structural homology to other known actin binders suggests that the chivosazoles may have a distinct mode of action.90, 91 What ignited our interest in the chivosazoles as a synthetic target was their astounding array of structural features (Scheme 17). Specifically, all congeners as typified by chivosazole F (4) possess a 31-membered macrolactone, containing 10 stereocenters and an oxazole moiety. However, the most impressive feature is the set of conjugated polyenes with alternating geometry in the macrocycle: a (Z,E,Z,E)-C2-C9 tetraene, a (Z,E)-C12-C15 diene and an (E,E,Z)-C23-C28 triene regions.92 These polyene regions demanded careful handling of sensitive late-stage intermediates and mild reaction conditions, necessary to suppress both potential olefin isomerization and degradation pathways. Perhaps as a reflection of the challenges imposed by this demanding target, only two total syntheses of chivosazole F (4), including our approach described below, have been reported to date.93, 94

Our synthetic approach needed to address the delicate nature of the chivosazole structure; in particular, the isomerization-prone (2Z,4E,6Z,8E)-tetraene. Therefore, we sought to minimize the number of endgame transformations. To this end, we envisaged a highly convergent approach towards accessing the full carbon skeleton by using site-selective cross-couplings. This broadly disconnects the full carbon skeleton to reveal the C14-C35 northern hemisphere and the C1-C13 southern hemisphere of the natural product.

The success of this strategy crucially relied on the judicious choice of coupling handles and cross-coupling conditions. Building on initial intelligence gathering studies, we discovered that the Stille cross-coupling provided the most efficient means of fragment union. We also anticipated that a late-stage macrolactonization might generate the macrocycle. This analysis revealed four constituent fragments—the C1-C5 fragment 93, the C6-C13 fragment 94, the C14-C26 fragment 95 and the C27-C35 fragment 96.

Recognizing the 19,22-syn relationship, synthesis of the C14-C26 bis-halide linchpin commenced with an asymmetric boron-mediated aldol reaction from methyl ketone 97 and aldehyde 9895, 96 to afford β-hydroxyketone 99 (Scheme 18a). This was then subjected to Evans–Tishchenko reduction56 to establish the remaining stereocenter. The oxazoline ring was cyclized using diethylaminosulfur trifluoride97 following amide formation from carboxylic acid 100 and amino alcohol 101.98 From oxazoline 102, oxazole formation using MnO2 proved incompatible with the pendant vinyl iodide functionality, suggesting that this oxidation step should be conducted postfragment assembly. Beginning from the ethyl ketone derivative 103 of (S)-Roche ester and known aldehyde 104,99, 100 a boron-mediated aldol reaction101 readily installed the C31 and C32 stereocenters in β-hydroxyketone 105, with an Evans–Tishchenko reduction again used to set the final C30 stereocenter (Scheme 18b). A six-step sequence revealed aldehyde 106, which was subjected to a Stork–Zhao olefination, deprotection and stannylation to afford the required stannane 96.49

The C1-C13 southern hemisphere contains what is arguably the most delicate polyene region of the chivosazoles. A vinylogous Mukaiyama aldol reaction102 between the chiral silyl ketene aminal 107 (derived from imide 108) with aldehyde 109 forged the two stereocenters in the C7-C13 fragment 110 (Scheme 19).103 Subsequent Stork–Zhao olefination of aldehyde 111 installed the terminal (6Z)-vinyl iodide in 112, which then engaged in a site-selective Stille cross-coupling with stannane 93 to afford the C1-C13 southern hemisphere 113 in preparation for exploring the planned fragment coupling sequence.

With the two hemispheres in hand, we looked towards effecting the site-selective Stille coupling between the stannane 114 derived from 113 and bis-halide 102. Unfortunately, not only did this fail to effect the required coupling, it also highlighted the propensity for the tetraenoate 114 to isomerize under Pd(0) conditions (Scheme 20a). Similarly, model studies investigating the esterification of 115 with vinyl stannane 116, with the goal of effecting a macro-Stille ring closure, afforded the isomerized (2E,4E)-stannane 117 under Yamaguchi conditions (Scheme 20b). To avoid handling the isomerization-prone (2Z) olefin, we next investigated the possibility of achieving a late-stage macro-olefination with a pendant phosphonate ester at C30 in 118. The revised synthesis of the southern hemisphere thus involved a Stille coupling with vinyl iodide 94 and stannane 120 (Scheme 20c).

Using optimized Stille cross-coupling conditions, and rationalizing chemoselective coupling on steric and electronic grounds, we were able to append C3-C5 stannane 119 onto the C6-C13 vinyl iodide 94 to yield the C1-C13 tetraene 120 (Scheme 21). Crucially, addition of tBu3P104 was required to prevent isomerization of the (6Z)-alkene. These conditions also allowed for the successful site-selective formation of the C13-C14 bond between bis-halide 102 and vinyl stannane 120, as well elaborating the resulting vinyl bromide 121 with the C27-C35 fragment 118 (derived from vinyl iodide 122), with complete control of alkene geometry throughout the process. This success led us to ponder whether we could turn this into a one-pot process. Remarkably, with sequential addition of each fragment (i. 119, ii. 94, iii. 102 and iv. 118), we were able to assemble the full carbon skeleton of the chivosazoles in 123 in one pot in 56% yield (82% per coupling step). At this advanced stage, the (4E,6Z,8E)-triene was found to be highly prone to isomerization on attempting to adjust the oxidation state at C3 ahead of the planned Horner–Wadsworth–Emmons-type macro-olefination. Furthermore, model studies on the planned Ando-olefination105 gave poor control over the desired 2Z geometry. This series of disappointing and incredibly frustrating setbacks forced us to return to the drawing board.

The challenges imposed by the delicate triene necessitated us to reconfigure our choreography of fragment coupling to an end-stage macro-Stille cyclization (Scheme 22). Furthermore, to access the (2Z) geometry, an alternative olefination strategy was required. The anticipated lability of the target molecule also prompted us to switch from an acetonide to a silylene-protecting group for the 32,34-diol to facilitate a mild final deprotection. These alterations meant that our constituent fragments towards assembling 4 would involve a C3-C5 aldehyde 124, a revised C7-C13 stannane 125 and a revised C27-C35 phosphonate 126.

The revised C27-C35 phosphonate 126 was made from diol 127,94 an intermediate used in our previous routes. Building on the prior work, fragment coupling could be conducted in a stepwise manner (via the C7-C26 vinyl bromide 128) or a one-pot process (i. 102 ii. 125 iii. 126) to deliver efficiently the advanced fragment 129 (Scheme 23). Gratifyingly, using the Still–Gennari-type phosphonate afforded useful selectivities towards the desired 2Z geometry for the Horner–Wadsworth–Emmons olefination with aldehyde 124.106, 107 At this stage, a double oxidation of the C7-OH and the oxazoline was carried out using MnO2, notably accomplishing the challenging aromatization on a delicate advanced fragment. Subsequently, the resulting aldehyde 130 was elaborated via a Stork–Zhao olefination to furnish the full carbon skeleton and also the seco precursor for the ring-closing intramolecular Stille reaction. To our delight, the critical macrocyclization delivered the protected natural product with complete retention of olefin geometry. A final global deprotection concluded our total synthesis of chivosazole F (4) in 20 steps and 2.5% overall yield. Our success in this arduous campaign hinged upon careful initial analysis and planning, which fortunately, allowed for a highly convergent approach and a succinct endgame sequence. While we recognized the potential lability of such advanced polyene fragments, we could not have anticipated the frustration it brought. In this case, it truly stressed the importance of a flexible, modular strategy and the ability to adapt the strategy as required.

Conclusions

Our recent synthetic endeavors towards these highly challenging classes of complex polyketides not only showcases the versatility of our group’s aldol methodology but also highlights the trials and tribulations we overcame in a sustained campaign to achieve these enticing targets. In our total synthesis of spirastrellolide A methyl ester, we discovered that the subtle, unexpected structural effects imposed by distal-protecting groups proved to be highly consequential in the critical macrolactonization. Similarly, for rhizopodin, a carefully choreographed sequence of protecting group incorporation and selective deprotection, was pivotal to achieving the target. Our campaign towards leiodermatolide underlines the need to reassess fragment coupling strategies when required. This is a common theme and important lesson—and was certainly a defining obstacle in our campaign towards chivosazole F. In the end, a carefully orchestrated sequence of fragment coupling steps proved to be vital for success.

In this account, the highlighted setbacks and accompanying explanations of strategy evolution serve to illuminate the unanticipated difficulties that can make or break a total synthesis. Overall, we are provided with a humbling reminder that despite continual advances in the field of chemical synthesis, there is still much to be learned from tackling a structurally complex natural product.

Dedication

Dedicated to Professor KC Nicolaou.

scheme 1

Endgame sequence for the first-generation synthesis of spirastrellolide A methyl ester (1).

scheme 2

Revised retrosynthesis of spirastrellolide A methyl ester (1).

scheme 3

(a) First-generation approach towards the C26-C40 DEF bis-spiroacetal 17. (b) Revised strategy towards the C26-C40 DEF bis-spiroacetal 21.

scheme 4

Revised synthesis of the C26-C40 DEF bis-spiroacetal 26.

scheme 5

Synthesis of the C17-C40 aldehyde 14.

scheme 6

Synthesis of the full C1-C47 carbon and oxygen skeleton of spirastrellolide A methyl ester 10.

scheme 7

(a) Truncated seco-acid 33 failed to macrocyclize when subjected to established macrolactonization conditions. (b) Macrocyclization conditions in our first-generation synthesis. (c) Endgame and total synthesis of spirastrellolide A methyl ester (1).

scheme 8

(a) Initial approach towards leiodermatolide (ent-2). (b) Summary of our first-generation synthesis towards the C1-C17 macrocycle 41.

scheme 9

Revised retrosynthesis for leiodermatolide (2).

scheme 10

Synthesis of the C1-C11 vinyl stannane 58.

scheme 11

Synthesis of the C12-C17 vinyl iodide 50 and C18-C25 δ-lactone ent-39.

scheme 12

Fragment union and completion of (–)-leiodermatolide (2).

scheme 13

Retrosynthetic analysis of rhizopodin (3). Disconnection (1) refers to an esterification/macrolactonization, and disconnection (2) refers to an aldol/dehydration/reduction sequence.

scheme 14

(a) Synthesis of the C14-C22 carboxylic acid 71. (b) Synthesis of the C8-C13 amino alcohol 72. (c) Synthesis of the C1-C7 vinyl iodide 73.

scheme 15

Synthesis of the truncated C1-C22 monomers.

scheme 16

Fragment union and completion of the total synthesis of rhizopodin (3).

scheme 17

Initial synthetic strategy towards chivosazole F (4) and the four proposed fragments.

scheme 18

(a) Synthesis of the C14-C26 bis-halide linchpin 102. (b) Synthesis of the C27-C35 vinyl stannane 96.

scheme 19

Synthesis of the C1-C13 southern hemisphere fragment 113.

scheme 20

(a) Initial attempts at fragment union under Stille conditions failed to deliver the product and resulted in isomerization of the tetraenoate. (b) Esterification of alcohol 115 to the C1-C5 acid 116 resulted in concomitant isomerization of the C2 olefin. (c) Our revised synthetic approach to chivosazole F (4).

scheme 21

Using the site-selective Stille coupling strategy to form the chivosazole backbone.

scheme 22

Final strategy adopted towards the total synthesis of chivosazole F (4).

scheme 23

Revised fragment coupling and completion of the total synthesis of chivosazole F (4).