Total synthesis of peloruside A via kinetic lactonization and relay RCM cyclization reactions (and identification of iso-peloruside A)

Abstract

A convergent total synthesis of peloruside A (1) is described. The key strategic features are a diastereoselective lactonization to generate a C5-C9 valerolactone from the C₂-symmetric ketone 4, which comprises C1–C9 of 1, and a relay ring closing metathesis (RRCM) reaction to produce a dehydrovalerolactone 20, which embodies C13–C19. A new isomer of 1, the valerolactone iso-peloruside A (iso-1), was identified.

The highly cytotoxic marine macrolide (+)-peloruside A (1) was isolated from the New Zealand sponge, Mycale hentscheli, by Northcote and co-workers.^[1] It showed LD₅₀s ranging from 6–18 nM toward H441, SH-SY5Y, and P388 cancer cell lines^[2] and has been further explored from a preclinical perspective. Like paclitaxel (cf. Taxol^®), peloruside A is a microtubule-stabilizer, arresting cells in the G2–M phase of the cell cycle, but it targets a different tubulin binding site than paclitaxel.^[3] As a consequence peloruside A (1) is a candidate for use against paclitaxel resistant cell lines.^[4] The therapeutic potential, low natural abundance (3 mg/170 g wet sponge), and architectural complexity have generated significant interest in the development of chemical syntheses of 1. To date, four total syntheses of peloruside A have been reported (by the research groups of De Brabander, Taylor, Ghosh, and Evans)^{[5],[6],[7],[8]} and related synthetic studies have been described.^{[9],[10],[11]}

Our interest in peloruside A (1) as a target was greatly heightened when we identified a plan that seemed ideally matched with strategies and technologies developed earlier in our group. In particular, it was attractive to apply a diastereoselective kinetic lactonization of a pseudosymmetric azelaic acid derivative^[12] and to capitalize on the versatility of relay ring-closing metathesis (RRCM)^[13] reactions.

As summarized in Scheme 1 our plan for the synthesis was to effect a late stage aldol coupling (cf. a) between the aldehyde acceptor 2 and methyl ketone donor 3 (or the complementary ketone/aldehyde pair 2′/3′). We planned to install the Z-trisubstituted and doubly allylically branched Δ^16,17-alkene in 3 (or its aldehyde analog 3′) via RRCM (cf. b) of the silaketal^[14] 5a or ester^[15] 5b (peloruside A skeleton numbering is used throughout). We envisioned the main aldehyde fragment 2 (or its ketone analog 2′) to arise from the C₂-symmetric azelaic ester precursor 4 (cf. c). As further detailed below, reduction of the C5-ketone in 4 from either of its homotopic faces would give an alcohol that, through engagement of its pro-S rather than pro-R diastereotopic ester group, would selectively provide a valerolactone/monoester (i.e., 11 in Scheme 2).^[16] Terminus differentiation and C9–C10 bond formation would then provide access to 2 (or 2′).

Scheme 1.

Retrosynthesis outline for accessing peloruside A (1). BPS = tert-butyldiphenylsilyl; MOM = methoxymethyl; PMB = para-methoxybenzyl; RCM = ring-closing metathesis; TBS = tert-butyldimethylsilyl.

Scheme 2.

Synthesis of main fragment 2 via the key diastereoselective lactonization of pseudosymmetric carbinol 9. Reagents and conditions: (a) i) DIBAL-H, Et₂O, −78 °C; then (EtO)₂P(O)CH(Na)CO₂ⁿHex, 80%; ii) SAD, 0 °C, 88%; (b) i) 30 mol% HI, THF, 0 °C, 83%; ii) Me₃OBF₄, proton sponge, 0 °C to RT, 84%; (c) i) 1,2-ethanedithiol, BF₃·OEt₂, 0 °C; ii) MOMCl, i-Pr₂NEt, DCM, RT, 90% over 2 steps; iii) I₂, NaHCO₃, acetone, H₂O, 0 °C, 91%; iv) Otera’s catalyst,^[22] MeOH, toluene, 90 °C, 77%; (d) Raney-Nickel, H₂, EtOH, RT, 91%; (e) TMG, C₆H₆, RT; TFA, 98% (dr 12:1). (f) i) L-Selectride^®, THF, −78 °C, 87%; ii) prenyl bromide, indium powder, DMF, 55 °C, 80%; iii) AlCl₃, NaI, CH₃CN, DCM, 0 °C, 92%; iv) (p-MeO)-PhCH(OMe)₂, CSA, DCM, 4Å mol sieves, 87%; v) MOMCl, i-Pr₂NEt, DCM, RT, 99%; (g) i) LAH, THF, 0 °C, 96%; ii) BPSCl, ImH, DMAP, DMF, RT, 92%; iii) DIBAL-H, DCM, −78 °C, 95%; iv) DMP, NaHCO₃, DCM, RT; v) Zn(BH₄)₂,^[23] DCM, 0 °C (dr 3:1), 62% (2 steps); vi) TBSOTf, 2,6-lutidine, DCM, RT, 89%; (h) i) HF·pyridine, THF, pyridine, RT, 91%; ii) DMP, NaHCO₃, DCM, RT; iii) NaClO₂, NaH₂PO₄, t-BuOH, H₂O, Me₂C=CHMe, RT; iv) CH₂N₂, RT, 87%, 3 steps; v) O₃, pyridine, DCM, MeOH, −78 °C, 75%. CSA = camphorsulfonic acid; DCM = dichloromethane (CH₂Cl₂); DMAP = 4-dimethylamino-pyridine; DMF = dimethylformamide; DMP = Dess-Martin periodinane; dr = diastereomeric ratio; ImH = imidazole; LAH = lithium aluminum hydride; OTf = triflate; SAD = Sharpless asymmetric dihydroxylation; TFA = trifluoroacetic acid; THF = tetrahydrofuran; TMG = tetramethylguanidine.

The synthesis of fragment 2 is presented in Scheme 2. The main features are: i) synthesis of the C₂-symmetric ketodiester 4, ii) diastereoselective conversion, via desymmetrizing kinetic lactonization, of the derived alcohol 9 to the δ-valerolactone 11, and iii) chemoselective elaboration (including prenylation) of the lactone vs. ester carbonyl groups [C9 vs. C1] in 11 to give fragment 2.

1. The tetrol 7 was prepared from the ethylene ketal of dimethyl acetone dicarboxylate (6)^[17] by utilizing a one-pot DIBAL-H reduction of both esters to the intermediate 1,5-dialdehyde, followed by in situ double Horner-Wadsworth-Emmons homologation with (EtO)₂P(O)CH(Na)CO₂ⁿHex.^[18] Two-fold Sharpless asymmetric dihydroxylation (SAD)^[19] of the bis-methyl ester analog was initially studied,^[16] but this transformation was plagued by incomplete conversion to the bis-methyl ester variant of tetrol 7. NoD ¹H NMR analysis^[20] of the aqueous component of the biphasic mixture indicated the accumulation of the intermediate alkenediol. The polar nature of this half-oxidized intermediate likely led to its undesired partitioning into the aqueous layer and, consequently, away from the active hydroxylation species. The substantial water solubility of the desired tetrol was a further complication. This prompted our selection and use of the less polar and more well-behaved bis-n-hexyl ester series, which allowed isolation of 7 in 88% yield. Exposure to a catalytic amount of aqueous hydriodic acid then promoted ketal-metathesis by engagement (and protection) of the C2- and C8-hydroxyl groups to give a spirocyclic ketal as a single diastereomer. Installation of the methyl ethers found at C3 and C7 in peloruside A (1) was achieved with Meerwein’s salt to provide 8. Transketalization of 8 with ethanedithiol (BF₃·OEt₂),^[21] bis-MOM ether protection of the C2/C8-diol, dithiolane removal (I₂, aq NaHCO₃), and transesterification with methanol (Otera’s catalyst) smoothly gave 4 (cf. Scheme 1). The sequence from 6 to 4 proceeded in over 30% yield.

2. While many reagents sufficed to chemoselectively reduce the C5-ketone in 4, most were complicated by premature lactonization of the resulting hydroxy bisester 9, with not only modest but also irreproducible levels of diastereocontrol. Hence, neutral conditions for this reduction that also minimized handling of 9 were sought. Use of hydrogen gas over Raney nickel provided a convenient solution; simple filtration and solvent removal provided the C₁-symmetric alcohol 9 (91%), which contains the new non-stereogenic but chirotopic^[24] C5-carbinol center. The key diastereoselective cyclization of this substrate^[16] proceeded with preferential engagement of the pro-S ester group at C9 to give lactone 11 (98%, dr 12:1). Use of tetramethylguanidine (TMG, 2.0 equiv) to promote this lactonization gave reproducibly high levels of diastereoselectivity. Interestingly, a labile adduct of TMG with the product 11, perhaps involving a tetrahedral intermediate in which the lactone carbonyl group remained engaged with TMG, was generated. We found it best to cleave this adduct by treatment with anhydrous trifluoroacetic acid (NMR analysis) prior to aqueous workup. The sense of the kinetic diastereocontrol observed in this lactonization was anticipated^[12] on the basis of conformational analysis; the equilibrium concentration of the reactive preclosure conformer 10-r, which could engage the C1-ester and lead to the minor (and undesired) C5-epimer of 11, should be lower than that of the reactive preclosure conformer 10-s, a necessary intermediate enroute to 11. Analogous considerations equally apply to the pair of tetrahedral intermediates derived from 10-r and 10-s, should it be the case that the rate-limiting step in the lactonization is not the initial closure implied by 10. A distinguishing feature of this symmetry-enabled approach is the rapidity and efficiency with which the C1-C9 portion of the peloruside skeleton was established.

3. Chemoselective reduction of the lactone rather than ester functional group in 11 was achieved with L-Selectride^®[25] to provide a lactol, which was treated with prenyl bromide/indium.^[26] This installed the gem-dimethylated C10-moiety, while simultaneously inducing relactonization, now of the C5-OH with the C1-ester.^[25] Use of the prenyl unit was designed to permit potential access to either the aldehyde 2 or methyl ketone 2′, but as things later developed, implementation in the former role proved more valuable. Bis-MOM ether cleavage (AlCl₃, NaI), PMP-acetal formation, and MOM reprotection of the C2-OH gave 12 as a single epimer at C9 (assigned to be of S configuration,^[27] consistent with α_MOM-chelation controlled addition). A series of functional/protecting group manipulations [cf. (g), Scheme 2] served to carry 12 efficiently to 13, the precursor to main fragment 2. Key among these was the oxidation/reduction of the C8-carbinol, which had just been revealed by a highly regioselective reductive cleavage by DIBAL-H of the PMP-acetal, to effect the required inversion of configuration at C8. Finally, but again efficiently, the C1-methyl ester was reinstated and the C11-aldehyde generated by ozonolysis [cf. (h), Scheme 2] to complete the synthesis of 2.

4. To explore a metathesis-based synthesis of the hindered Z-alkene in 3 (Scheme 3), we first prepared and examined the RCM reaction of the simple diene^[14] 22a (Scheme 3), which was an epimeric mixture at C15. This proved to be a poor RCM substrate (e.g., <20% yield of 16 using 45 mol% of G2).^[28] We turned to the relay RCM substrate 5a,^[29] which was made by sequential loading of the alcohols 14^[13] and 15 onto Ph₂SiCl₂. RRCM proceeded to give 16 in 92% yield. For the preparation of 5a as essentially one diastereomer, enantiomerically enriched β-hydroxynitrile 15 was accessed by addition of lithioacetonitrile to the enal derived from allylic oxidation (SeO₂) of (R)-citronellene, followed by enzymatic (kinetic) resolution of the resulting C15-epimeric carbinols (Novozyme 435).^[13],[30] An added advantage of adopting the RRCM strategy here was that the two epimeric carbinols (15 and C15-epi-15) were easily discriminated by the lipase, allowing for efficient resolution; the alternative substrate 23, the precursor to 22a, has two nearly isosteric groups and was a poor substrate for lipase resolution.

5. A complementary approach involving the ester 5b and its RRCM product, the lactone 20 was also studied. Preparation of 5b is inherently easier than that of the silaketal 5a, since it involves simple cross coupling of two functional groups with complementary reactivity, namely the alcohol 15 with the acid 19. Moreover, a sample of nonracemic 19 proved to be easier to access than one of 14. The terminal alkene in the 2-propenylated compound 18^[31] was efficiently isomerized to the more stable, internal, 1-propenyl analog using RhCl₃/EtOH;^[32],[33] there was no evidence of erosion of the high dr, which implies that the branched α-carbon effectively impedes further isomerization by β-hydride elimination of the C_α methine hydrogen. This sequential allylation/2- to 1-propenyl isomerization/ethylene cross-metathesis degradation^[34] resulted in net vinylation of, in this instance, an Evans acyloxazolidinone auxiliary (i.e., 24). We recommend consideration of this strategy in instances when ethenylation of carbanionic species is needed. Hydrolysis of the oxazolidinone 18 gave the acid 19, which upon DCC activation was coupled with 15 to provide 5b. This ester smoothly underwent RRCM to give lactone 20 in 70% yield.

6. Efficient elaboration of either 16 or 20 to the differentially protected diol derivative 17 was straightforward following desilylation (of 16) or reduction (of 20) and sequential TBS and PMB installations on the 1° and 2° hydroxyls, respectively. The choice of sodium borohydride in acetic acid as the reagent for the reduction of 20 was dictated by its base lability, which otherwise led to epimeric mixtures of the reduced diol product. Base instability was an even bigger obstacle in the final conversion of the nitrile in 17 to the methyl ketone in fragment 3. A solution emerged in the form of a modified Blaise reaction, wherein the reagent formed in situ from BrCH₂CO₂allyl/Zn°/Cp₂TiCl₂^[35] converted the nitrile to an intermediate β-aminoenoate, the hydrolysis of which was carefully monitored at pH 3 (ca. 2 days) to afford the β-ketoester 21. Attempts to effect hydrolysis or dealkylation and decarboxylation of the methyl ester analog of 21 to 3 were compromised by competitive β-elimination of PMB-OH. We solved this complication by use of the allyl ester 21, which could be smoothly decarbalkoxylated with Pd(PPh₃)₄ (HCO₂H, Et₃N).^[36] The non-basic conditions associated with this conversion of nitrile to methyl ketone render this strategy applicable to other base-sensitive nitrile substrates.

   The final stage of the synthesis is presented in Scheme 4. The main features are vii) the Paterson boron aldol coupling^[37] of fragments 2 and 3, viii) the regioselective macrocyclization of the diol acid 28, and ix) the removal of protecting groups from the penultimate intermediate 30, including our observation of a byproduct we name here iso-peloruside A (iso-1).

7. A number of model experiments designed to establish the feasibility of either the 2′/3′ or 3/2 aldol donor/acceptor pairing were carried out.^[38] These eventually guided us to the use of the dicyclohexylboron enolate derived from ketone 3, which added to the aldehyde 2 (1.0 equiv, −78 to −20 °C) to form the C11–C12 bond and produce the ketoalcohol 25 (64%) with essentially full control of the new stereocenter at C11. Presumably, the 1,5-stereoinduction keyed by the β(C15)-stereocenter in the ketone donor 3^[39] is enhanced by (i.e., matched with) that from the β(C9)-stereocenter of the α,α-gem-dimethylated aldehyde acceptor 2^[40] to impart essentially perfect diastereoselection during this powerful cross-coupling reaction.^[41] Reduction of the C13-ketone with Me₄NBH(OAc)₃ proceeded to give the 1,3-anti-diol 26 (dr 95:5). Methylation with Meerwein’s salt proceeded with near perfect regioselectivity. Installation of a MOM ether at C11 and removal of both PMB ethers at C9 and C15 proceeded smoothly to complete the preparation of 27.

8. Conversion of the methyl ester in 27 to the acid 28, in preparation for the required macrolactonization, was very slow with LiOH (aq MeOH) and was complicated by partial epimerization at C2 when KOSiMe₃ (THF) was employed. These issues were solved by use of the more reactive and less basic LiOOH. Yamaguchi cyclization of diol acid substrate 28 was uneventful; the free C9-hydoxyl gave no evidence of interference (lactonization substrates in all previous peloruside A syntheses and studies have had the ketone oxidation state at C9). Oxidation of 29 to the C9-ketone 30 was smoothly effected with the Dess-Martin periodinane.

9. The silyl ether and MOM ether protecting groups were sequentially removed by treatment first with HF·pyridine in THF, buffered by additional pyridine, and second with 4N aq HCl in THF (RT). The first stage (removal of three silyl groups) was monitored by LC/MS analysis. The primary C24-OH was generated quickly at room temperature, the C5-tert-butyldiphenylsilyl ether (OBPS) was cleaved in about a day at 60 °C, and the hindered C8-OTBS required an additional day to completely release the C5/C8/C24-triol (having the C5-OH engaged to C9 as a hemiketal). Even though it required strongly acidic conditions (aq HCl, RT), the final MOM deprotection was planned with confidence, given the precedent from the earlier total syntheses. Without further purification, the C5/C8/C24-triol was treated with aq HCl at room temperature for 3 h to deliver (+)-peloruside A (1) in 49 % overall yield from 30 (over 2 steps). We also observed and isolated (via careful SiO₂ chromatography, MeOH/DCM:5/95) a minor (5%) and slightly more polar byproduct formed during this experiment. We deduced its structure as that of the valerolactone derivative iso-1 based on ¹H NMR and ESI MS/MS analyses. The masses of fragments observed from the latter (and indicated on the structure of iso-1 in Scheme 4) were informative in suggesting the absence of a macrolactone in this new isomeric substance. In particular, the fragmentation of the C10-C11 bond via retro-aldol reaction was evidenced by the complementary pair of sodiated ions at m/z 327 and 267 and further cleavage of the C14–C15 bond in the latter gave the ion observed at m/z of 125. Additionally, ¹H NMR data of iso-1 bore strong similarities with the earlier described valerolactone 12. iso-Peloruside A (iso-1) presumably arose from intramolecular translactonization within 1 by the C5-hydroxyl in the minor [and unobserved (¹H NMR)] open form of the C9-hemiketal to the C1-carbonyl group.

Scheme 3.

Synthesis of fragment 3 via the relay ring-closing metathesis reactions of substrates 5a and 5b. Reagents and conditions: (a) i) 15, Ph₂SiCl₂, pyridine, RT; ii) 14, pyridine, 41%; (b) G2, PhCH₃, N₂, 65 °C, 92%; (c) i) TBAF, THF, RT, 93%; ii) TBSCl, Et₃N, DMF, RT, 84%; iii) Cl₃CC(=NH)OPMB, CSA, DCM, RT, 76%; (d) i) RhCl₃·3H₂O, EtOH/H₂O, 80 °C, 98% (15:1 = E:Z); ii) LiOH, H₂O₂, THF/H₂O, 0 °C, 99%; (e) 15, DCC, DMAP, DCM, 0 °C to RT, 82%; (f) G2, DCM, 45 °C, 70%; (g) i) NaBH₄, AcOH, EtOH, 0 °C to RT, 82%; ii) TBSCl, Et₃N, DMAP, DCM, RT, 99%; iii) Cl₃CC(=NH)OPMB, CSA, DCM, RT, 77%; (h) i) Zn°, Cp₂TiCl₂, allyl 2-bromoacetate, THF, 60 °C; ii) pH 3 buffer, i-PrOH/H₂O/THF, RT; (i) Pd(PPh₃)_4, HCO₂H, Et₃N, THF, RT, 61% (over 3 steps). (j) CH₂=CH₂ (15 psi), G2, PhH, 65 °C, quantitative. G2 = second generation Grubbs initiator, [Ru=CHPh(Cl)₂(PCy₃)(H₂IMes)]; [Ru*] = Ru(Cl)₂(H₂IMes)L_n; TBAF = tetrabutylammonium fluoride; DCC = dicyclohexylcarbodiimide.

Scheme 4.

Synthesis of peloruside A (1) via aldol coupling of fragments 2 and 3, macrolactonization of seco-acid 28, and deprotection. Reagents and conditions: (a) c-Hex₂BCl, Et₃N, 3, Et₂O, −78 °C to −40 °C; 2, Et₂O, −78 °C to −20 °C, 64% (along with 34% of recovered 3); (b) NMe₄BH(OAc)₃, AcOH/MeCN, −30 °C, 95% (dr 95:5); (c) i) Me₃OBF₄, 2,6-di-t-Bu-pyridine, DCM, 0 °C to RT, 82% (after two cycles); ii) MOMCl, i-Pr₂NEt, DCM, 0 °C to RT, 98%; iii) DDQ, DCM/pH 7 buffer, 0 °C to RT, 94% (after two cycles); (d) LiOH, H₂O₂, THF/MeOH/H₂O, 0 °C to RT, ca. 100%; (e) 2,4,6-trichlorobenzoyl chloride, i-Pr₂NEt, PhCH₃, 0 °C to RT; DMAP, PhCH₃, RT, 54%; (f) DMP, NaHCO₃, DCM, 0 °C to RT, 83%; (g) i) HF/pyr, THF, pyr, RT to 60 °C; ii) HCl (4 N), THF, RT; 49% 1 and 5 % iso-1 (over 2 steps).

In conclusion this total synthesis of peloruside A (1) capitalizes on the kinetic lactonization reaction of the pseudosymmetric azelaic acid derivative 9 (to 11, Scheme 2) and a relay ring closing metathesis reaction to access the trisubstituted Δ^16.17-alkene subunit (5b to 20, Scheme 3). Modified Blaise and net enolate ethenylation sequences are also noteworthy. Formation of the isomeric valerolactone iso-1 during final HCl-catalyzed deprotection of the MOM-ethers at C2 and C11 was observed for the first time.