Structural Revision and Total Synthesis of Azaspiracid‐1, Part 1: Intelligence Gathering and Tentative Proposal

Abstract

Phantom molecules from Nature: Degradation studies of azaspiracid-1 (see originally proposed, incorrect structure), the notorious marine neurotoxin isolated from contaminated mussels, revealed that the double bond of ring A was located at C7–C8 and not at C8–C9 and also provided further information on the overall stereochemistry of the molecule. Structure 1-i (Figure 1, absolute stereochemistry and relative stereochemistry between ABCDE and FGHI domains unknown) was proposed in 1998, on the basis of NMR spectroscopic analysis, for a poisonous substance called azaspiracid-1, which was isolated from the mussel family Mytilus edulis.1 A recent total synthesis2, 3 of 1-i and its FGHI epimer proved that this assignment was incorrect.4 Herein we report synthetic and degradation studies that led to the revised structure 1-c or ABCD-epi-1-c as the most likely depiction of the molecular architecture of this biotoxin (Figure 1). The only doubt that remained was the absolute configuration of the ABCD domain of the molecule, and this question was resolved by total synthesis, as is described in the following paper in this issue.5 Originally proposed structure 1-i (relative stereochemistry between ABCDE (C1–C27) and FGHI (C28–C40) domains and absolute stereochemistry unknown) and newly proposed structure 1-c or ABCD-epi-1-c of azaspiracid-1. Upon realizing that the originally proposed structure 1-i of azaspiracid-1 was in error,2, 3 our plan for determining its true structure called for the degradation of the natural material into smaller fragments, whose chemical synthesis may narrow down the location of the structural discrepancy between the synthesized (originally proposed) and actual structures. To this end, azaspiracid-1 was derivatized by exposure to TMSCHN2 in MeOH, and the methyl ester obtained was treated with NaIO4 in aqueous MeOH at ambient temperature, resulting in the cleavage of its C20C21 bond and in the formation of the corresponding aldehyde 3 (C1–C20 fragment) and lactone 4 (C21–C40 fragment) in quantitative yield (Scheme 1). Compound 3 was further elaborated by treatment with NaBH4 to afford, supposedly, hydroxy methyl ester 5 (≈90 % yield), followed by hydrogenation to give the fully saturated compound 6 (≈90 % yield). On the other hand, cleavage of the exocyclic olefinic bond within 4 with O3–Me2S to the corresponding ketone, followed by saponification of the E-ring lactone with aqueous NaOH in MeOH and esterification of the resulting carboxylic acid with TMSCHN2 gave methyl ester 7 (≈90 % yield). Finally, treatment of compound 7 with NaIO4 furnished amino acid 8 in high yield (≈90 %). These degradatively derived intermediates (i.e. 4–6 and 8, Scheme 1) were to serve several purposes. First, spectroscopic comparisons with synthetic materials would immediately pinpoint any errors in each domain of the molecule. Second, comparison of amino lactone 4 with the two synthetic FGHI epimers of the same structure (i.e. 4) would clarify the relative stereochemistry between the ABCDE and FGHI domains of azaspiracid-1. Third, optical rotation comparisons of the degradative and synthetic samples may reveal the absolute configuration of each fragment and, consequently, of the entire structure. Chemical degradation and derivatization of azaspiracid-1 (1-i: originally proposed structure) to C1–C20 alcohol 6, C21–C40 lactone 4, and C26–C40 carboxylic acid 8. Reagents and conditions: a) TMSCHN2, MeOH, 25 °C, 1 h; b) NaIO4, MeOH/H2O (4:1), 25 °C, 1 h, ≈100 % over two steps; c) NaBH4, MeOH, 25 °C, ≈90 %; d) Pd/C (5 %), H2, MeOH, 25 °C, ≈90 %; e) O3, Me2S, MeOH, −78 °C; f) NaOH (0.1 n), MeOH/H2O (4:1), 25 °C, 2 h; g) TMSCHN2, MeOH, 25 °C, 15 min, ≈90 % over three steps; h) NaIO4, MeOH/H2O (4:1), 25 °C, 30 min, ≈90 %. TMS=trimethylsilyl. One drawback of the adopted degradation scheme was the loss of stereochemical information regarding the C20 and C21 hydroxy-bearing centers. However, if we assume that the C21 hemiketal functionality would rest in its thermodynamically most stable conformation, the only stereocenter in question would be that at C20. As 1-i had already been eliminated as the true structure, the C20-epi-azaspiracid-1 (C20-epi-1-i) structure and its FGHI epimer (C20-epi-FGHI-epi-1-i) as logical targets moved to the front as possible alternatives (see Figure 2). Their total syntheses along the same lines as those described for 1-i,2, 3 however, also proved them to be incorrect.6 Structures of C20-epi-1-i and C20-FGHI-epi-1-i (synthesized and proven incorrect). Having established that the azaspiracid-1 structural problem did not reside exclusively at the C20 stereocenter, the focus of our investigation shifted next to the synthesis of the two diastereomeric EFGHI amino lactones 4 and FGHI-epi-4 (Scheme 2). Thus, the previously synthesized dihydroxy dithiane 92 was selectively monoacetylated (at the primary position) with Ac2O in the presence of 2,4,6-collidine (85 % yield), and the resulting hydroxy dithiane intermediate was deprotected by exposure to PhI(OCOCF3)27 to afford the corresponding lactol, whose oxidation with NIS–TBAI8 completed the synthesis of the required allylic acetate lactone 10. The latter compound, 10, underwent Stille coupling with stannanes 11 and ent-11 according to a previously described procedure3 to afford compounds 12 and 13, respectively. These substrates were then smoothly transformed into the targeted amino lactones 4 and FGHI-epi-4, respectively, via intermediate compounds 14 (for 4) and 15 (for FGHI-epi-4) according to our previous sequence3 and as summarized in Scheme 2: 1) HF⋅py-induced selective TES removal; 2) NIS-initiated ring closure; 3) nBu3SnH–Et3B-facilitated reductive deiodination; and 4) TAS-F-mediated desilylation. The only deviation from the previously employed sequence3 was the use of TAS-F9 instead of TBAF, as the former reagent afforded superior results in the final deprotection step. Most gratifyingly, the 1H NMR spectral data of one of these synthetic materials, namely FGHI-epi-4, matched those of the sample derived from natural azaspiracid-1 (Scheme 1), whereas its diastereomeric amino lactone 4 exhibited slightly different and diagnostically distinct signals to those of the naturally derived substance. Based on these results, we concluded that the relative stereochemical arrangement for the EFGHI domain of azaspiracid-1 is that depicted by structure FGHI-epi-4 (absolute stereochemistry unknown).10 Chemical synthesis of C21–C40 lactone 4 and its FGHI epimers, FGHI-epi-4 and determination of the relative stereochemistry within the EFGHI domain. Reagents and conditions: a) Ac2O (5.0 equiv), 2,4,6-collidine (10.0 equiv), CH2Cl2, 25 °C, 16 h, 85 %; b) PhI(OCOCF3)2 (1.5 equiv), MeCN/pH 7 buffer (4:1), 0 °C, 10 min, 81 %; c) NIS (10.0 equiv), TBAI (2.0 equiv), CH2Cl2, 25 °C, 40 min, 74 %; d) 10 (3.0 equiv), [Pd2dba3] (0.9 equiv), AsPh3 (0.9 equiv), LiCl (18 equiv), EtNiPr2 (12 equiv); then 11 or ent-11 (0.03 M in THF, syringe pump addition), NMP, 45 °C, 4 h; e) HF⋅py (excess), THF/py (1:1), 0→25 °C, 2 h; f) NIS (10.0 equiv), NaHCO3 (30 equiv), THF, 0 °C, 16 h, 26 % for 14 and 38 % for 15 over three steps; g) Et3B (1.0 M in hexanes, 0.2 equiv), nBu3SnH/toluene (1:2), 0 °C, 5 min; h) TAS-F (5.0 equiv), DMF, 0 °C, 20 h, 70 % for 4 and 74 % for FGHI-epi-4 over two steps. NIS=N-iodosuccinimide, NMP=N-methylpyrrolidinone, TBAI=tetra-n-butylammonium iodide, dba=dibenzylideneacetone, TAS-F=tris(dimethylamino)sulfur (trimethylsilyl)difluoride. The next goal was to establish the absolute stereochemistry of azaspiracid-1, and as we had confirmed the structural integrity of the FGHI domain 8 of the molecule and had secured its optical rotation from degradation studies (see Scheme 1), we decided to synthesize the carboxylic acid 8 for comparison purposes. This objective was met expeditiously, starting with the previously prepared intermediate 16,2 as shown in Scheme 3. Thus, a two-step oxidative protocol involving dihydroxylation (NMO–OsO4 cat.) of the terminal olefin of 16, followed by NaIO4-induced cleavage of the resulting 1,2-diol, provided aldehyde 17 in 80 % overall yield. Oxidation of this aldehyde (17) with NaClO2, followed by TAS-F-mediated desilylation resulted in the formation of the desired amino acid 8 in 35 % yield over the two steps. As expected, the NMR spectroscopic data of synthetic 8 matched those of the degradatively obtained material. Furthermore, the two samples exhibited comparable rotations, but with opposite signs (synthetic: [α]=+49.6 (c=0.4, MeOH); natural: [α]=−59.0 (c=0.016, MeOH)). Furthermore, the 1H NMR spectrum of the (R)-PGME (phenyl glycine methyl ester) derivative 18 (see Scheme 3 for preparation of the PGME derivatives) of synthetic carboxylic acid 8 was identical to the (S)-PGME derivative of the naturally derived acid 8 (and (S)-PGME derivative 19 derived from synthetic 8 was identical to the (R)-PGME derivative of the naturally derived 8), thereby establishing the absolute stereochemistry of the FGHI domain of azaspiracid-1 as the antipode of compound 8 that we had prepared as shown in Scheme 3. By virtue of the chemistry depicted in Scheme 2, the absolute stereochemistry of the entire EFGHI domain of azaspiracid-1 could now be shown as that in Scheme 3 (structure FGHI-epi-4). Synthesis of FGHI amino acid 8 and determination of the absolute stereochemistry of EFGHI domain of azaspiracid-1. Reagents and conditions: a) OsO4 (0.1 equiv), NMO (3.0 equiv), acetone/H2O (3:1), 25 °C, 3 h, 92 %; b) NaIO4 (3.0 equiv), MeOH/pH 7 buffer (2.5:1), 25 °C, 1 h, 87 %; c) NaClO2 (10.0 equiv), NaH2PO4 (10.0 equiv), 2-methyl-2-butene (excess), tBuOH/H2O (4:1), 25 °C, 30 min, 93 %; d) TAS-F (5.0 equiv), DMF, 0 °C, 16 h, 38 %; e) (R)-PGME or (S)-PGME (5.0 equiv), EDC (5.0 equiv), HOBt (5.0 equiv), NaHCO3 (10 equiv), DMF, 25 °C, 16 h, 75 % for 18 and 63 % for 19. NMO=N-methylmorpholine N-oxide, PGME=phenyl glycine methyl ester, EDC=1-(3-dimethylaminopropyl)-3-ethylcarbodiimide, HOBt=1-hydroxybenzotriazole. With the establishment of both the relative and absolute stereochemistries of the EFGHI domain of azaspiracid-1, our attention was then turned to the remaining segment, namely the ABCD (C1–C20) framework of the natural product, in which we had, by now, suspected the structural discrepancy to be embedded. As outlined in Scheme 4, a synthetic sample of 21, a substance corresponding to the degradatively derived compound from the natural product (structure 5, Scheme 1) was prepared from the previously synthesized intermediate 202 by desilylation (TBAF, 88 % yield). The spectroscopic data for these two samples (5 and 21) differed significantly, thereby confirming our suspicions for the location of the structural discrepancy (see Table 1). Furthermore, the main differences between the two sets of spectra were associated with protons located on ring A, namely 6-H, 7-H, 8-H, and 9-H (see numbering on structure 21, Scheme 4). Despite these clear differences, the final structural deconvolution of this region of azaspiracid-1 remained elusive, primarily as a result of the very similar 2D NMR spectral data for the naturally derived and synthetic materials whose comparison implied only subtle connectivity and spatial differences. Finally, it was a discovery by C. Hopmann and the late Professor D. John Faulkner that shed light on this issue. Indeed, the structural elucidation of lissoketal (22, see Scheme 4),11 a marine natural product isolated by these workers in 1997, provided the first clue to the azaspiracid-1 puzzle. This secondary metabolite 22 bears a close resemblance to the proposed structure of azaspiracid-1 (1-i) except for the fact that the endocyclic double bond in ring A resides between C7 and C8, rather than between C8 and C9. Intriguingly, close examination of the 1H NMR chemical shift values (see Table 1) and the 2D NMR correlation pattern reported for lissoketal (22) showed remarkable similarities to those observed for natural azaspiracid-1.1 In particular, a weak HMBC correlation between C10 and 7-H as well as a weak COSY correlation between 6-H and 9-H observed for the natural azaspiracid-1 are in line with observations reported by Hopmann and Faulkner11 and, therefore, point to the lissoketal ring A structural arrangement. Furthermore, relocating the endocyclic double bond from C8C9 to C7C8, places the 6-H in a doubly allylic environment, thereby providing a possible explanation for the significant downfield shift for this proton exhibited in the 1H NMR spectrum of the naturally derived intermediate 5 (δ=4.79 ppm) relative to that observed for the synthetic material 21 (δ=4.51 ppm).12 With the NMR spectroscopic parameters associated with rings B, C, and D for the naturally derived (5) and synthetic (21) samples in good agreement, the alternative structure 23 (see Scheme 4) was proposed for the hydroxy methyl ester derived from the natural product through intermediate 3 (see Scheme 1). To test this hypothesis, a chemical synthesis of the newly proposed structure 23 was required. Chemical synthesis of C1–C20 alcohol 21 for comparison with naturally derived material, structure of lissoketal (22), and second proposed structure, 23, for the ABCD domain of azaspiracid-1. Reagents and conditions: a) TBAF (2.0 equiv), THF, 25 °C, 2 h, 88 %. TBAF=tetra-n-butyl ammonium fluoride. H 5[a] (natural) 21 22 5[b] 23 δ(1H) [ppm] 6-H 4.79 4.50 4.34 4.79 4.81 7-H 2.48 2.06 5.80 5.62 5.61 2.03 2.04 8-H 5.74 6.07 5.80 5.74 5.77 9-H 5.62 5.79 2.57 2.48 2.51 1.91 2.03 2.18 16-H 3.78 3.94 – 3.78 3.84 17-H 4.21 4.27 – 4.21 4.15 19-H 4.34 4.57 – 4.34 4.51 41-H 0.92 0.98 – 0.92 0.95 The synthesis of the newly proposed structure 23 was initiated from the previously described compound 242 and proceeded as outlined in Scheme 5. Thus, the free hydroxy group in 24 was protected as a TES ether (95 % yield), the pivaloate ester was cleaved with DIBAL-H (96 % yield), and the liberated primary alcohol was oxidized under Swern conditions to afford aldehyde 25 (95 % yield). Addition of vinyl Grignard reagent to 25 (76 % yield) followed by acetylation (85 % yield) yielded allylic acetate 26 as a mixture of diastereoisomers (≈2:1). Ireland–Claisen rearrangement13 of 26 followed by methylation of the resulting carboxylic acid with diazomethane gave methyl ester 27 in 52 % overall yield. Selective removal of the TES group (HF⋅py, 85 % yield) from 27, Swern oxidation of the resulting alcohol (90 % yield) followed by treatment with KHMDS–TMSCl led to TMS enol ether 28, whose oxidation to the corresponding enone (55 % overall yield), reduction with NaBH4, and acetylation (71 % overall yield) furnished allylic acetate 29. Finally, exposure of 29 to catalytic amounts of [Pd2dba3]⋅CHCl3 in the presence of nBu3P and NaBH4,14 followed by treatment with TBAF led to the targeted intermediate 23 (42 % overall yield) through intermediate 30 (which was chromatographically separated from its Δ8,9 isomer that emerged as the minor product during the reduction step (≈2:1)). Chemical synthesis of the second proposed structure 23 for the ABCD domain. Reagents and conditions: a) TESCl (1.5 equiv), imidazole (3.0 equiv), 4-DMAP (0.1 equiv), CH2Cl2, 0 °C, 12 h, 95 %; b) DIBAL (2.0 equiv), CH2Cl2, −78 °C, 1 h, 96 %; c) (COCl)2 (5.0 equiv), DMSO (11 equiv), CH2Cl2, −78 °C, 30 min, −60 °C, 1 h; then Et3N (22 equiv), −78→−25 °C, 95 %; d) CH2CHMgBr (1.6 equiv), Et2O, 0→25 °C, 1 h, 76 %; e) Ac2O (5.0 equiv), py (10.0 equiv), 4-DMAP (0.1 equiv), CH2Cl2, 0 °C, 3 h, 85 %; f) LDA (1.5 equiv), THF, −78 °C, 10 min, TBSCl (1.5 equiv), HMPA (1.5 equiv), −78→25 °C, 24 h; g) CH2N2 (excess), Et2O, 25 °C, 30 min, 52 % over two steps; h) HF⋅py (10.0 equiv), py/THF (1:1), 0 °C, 2 h, 85 %; i) (COCl)2 (10.0 equiv), DMSO (22 equiv), CH2Cl2, −78 °C, 30 min, −60 °C, 1 h; then Et3N (44 equiv), −78→−25 °C, 90 %; j) KHMDS (1.5 equiv), THF, −78 °C, 1 h, TMSCl (1.6 equiv), −78 °C, 30 min; k) Pd(OAc)2 (5.0 equiv), DMSO, 72 h, 55 % over two steps; l) NaBH4 (3.0 equiv), CeCl3⋅7 H2O (1.0 equiv), MeOH, −50 °C, 1 h; m) Ac2O (20 equiv), py (40 equiv), 4-DMAP (0.1 equiv), CH2Cl2, 0→25 °C, 12 h, 71 % over two steps; n) [Pd2dba3]⋅CHCl3 (0.2 equiv), nBu3P (0.4 equiv), NaBH4 (10.0 equiv), dioxane/H2O (9:1), 0→25 °C, 12 h; o) TBAF (3.0 equiv), THF, 0→25 °C, 3 h, Δ7,8/Δ8,9 (2:1), 42 % over two steps. TES=triethylsilyl, DMSO=dimethyl sulfoxide, TBS=tert-butyldimethylsilyl, LDA=lithium diisopropylamide, KHMDS=potassium bis(trimethylsilyl)amide. Much to our disappointment, however, the spectral data of synthetic 23 did not match those of the degradatively derived 5 (see Table 1), sending us, one more time, back to the drawing board. Apparently, the location of the double bond in ring A was not the only problem with the originally proposed structure (1-i) of azaspiracid-1. A more profound structural change, perhaps having to do with the skeletal stereochemistry, was needed to accommodate the structure of the natural product. Indeed, the fully hydrogenated synthetic product 31 (see Scheme 6) obtained from synthetic 21 (10 % Pd/C, H2, 95 % yield) proved to be different from the corresponding hydrogenated fragment 6 obtained from 5, which was derived by degradation and elaboration of the natural product as already described above (see Scheme 1). Preparation of fully saturated synthetic ABCD fragment 31. Reagents and conditions: a) Pd/C (10 %), H2, CH3CO2Et, 25 °C, 1 h, 95 %. At this juncture, and having pinned down the location of the double bond within ring A, we were still faced with 128 possible structures for the ABCD domain of azaspiracid-1 arising from its seven stereogenic centers. However, our intelligence gathering at this stage extended to an additional piece of information regarding the thermodynamic stability of the degradatively derived ABCD fragments (e.g. 5) and of the natural product itself. Their ABC double spiroketal bridge was noted to be stable under acidic conditions, whereas our corresponding synthetic materials so far revealed their fleeting nature under similar conditions. This intriguing observation, subtle as it was, proved to be crucial in shaping our next hypothesis for solving the mystery of azaspiracid-1, as we will describe in the following communication in this issue.5 4: Rf=0.30 (silica gel, EtOAc/MeOH 9:1); [α]=+36.4 (MeOH, c=0.14); IR (film): max=3416, 2954, 2919, 1727, 1430, 1261, 1096, 1044, 797, 673 cm−1; 1H NMR (600 MHz, CD3OD/0.5 % CD3COOD): δ=5.35 (s, 1 H), 5.31 (s, 1 H), 4.52 (d, J=10.1 Hz, 1 H), 4.26 (br s, 1 H), 3.81 (d, J=1.8 Hz, 1 H), 2.85–2.70 (m, 2 H), 2.64–2.56 (m, 1 H), 2.49 (dd, Hz, 1 H), (m, 1 H), (d, Hz, 1 H), (d, Hz, 1 H), (m, 2 H), (m, 3 H), (m, 2 H), (d, Hz, 1 H), (m, 1 H), (m, 2 H), (m, 2 H), (d, Hz, 3 H), (m, 1 H), 0.95 (d, Hz, 3 H), (d, Hz, 3 H), (d, Hz, 3 H), (d, Hz, 3 NMR MHz, for (silica gel, EtOAc/MeOH 9:1); (MeOH, IR (film): 1261, cm−1; 1H NMR (600 MHz, CD3OD/0.5 % CD3COOD): (s, 1 H), (s, 1 H), (d, Hz, 1 H), (br s, 1 H), (d, Hz, 1 H), (d, Hz, 1 H), Hz, 1 H), (dd, Hz, 1 H), (m, 1 H), 2.48 (d, Hz, 1 H), (m, 1 H), (d, Hz, 1 H), (m, 1 H), (m, 2 H), (m, 1 H), (dd, Hz, 1 H), (d, Hz, 1 H), (m, 1 H), (m, 2 H), (m, 3 H), (d, Hz, 3 H), (m, 2 H), (d, Hz, 3 H), 0.98 (d, Hz, 3 H), (d, Hz, 3 H), 0.95 (d, Hz, 3 NMR MHz, for (silica gel, [α]=+49.6 (MeOH, IR (film): cm−1; 1H NMR (600 MHz, (br s, 1 H), (br s, 1 H), (br s, 1 H), (s, 2 H), Hz, 1 H), (m, 1 H), (m, 1 H), (m, 2 H), (m, 1 H), (m, 1 H), (m, 2 H), (m, 1 H), (m, 2 H), (m, 1 H), (d, Hz, 3 H), (d, Hz, 3 H), (d, Hz, 3 NMR MHz, for (silica gel, IR (film): cm−1; 1H NMR MHz, Hz, 1 H), (m, 2 H), 5.62 (dd, Hz, 1 H), (m, 1 H), (m, 1 H), (m, 1 H), (m, 1 H), (m, 1 H), (s, 3 H), (m, 1 H), (m, 4 H), (m, 1 H), (m, H), (m, 1 H), (m, 1 H), 0.98 (d, Hz, 3 NMR MHz, for In of D. John Faulkner