Total synthesis and complete configurational assignment of amphirionin-2

Amphirionin-2 is a linear polyketide metabolite that exhibits potent and selective cytotoxic activity against certain human cancer cell lines. We disclose herein the first total synthesis of amphirionin-2 and determination of its absolute configuration. Our synthesis featured an extensive use of cobalt-catalyzed Mukaiyama-type cyclization of γ-hydroxy olefins for stereoselective formation of all the tetrahydrofuran rings found in the natural product, and a late-stage Stille-type coupling for convergent assembly of the entire carbon backbone. Four candidate diastereomers of amphirionin-2 were synthesized in a unified, convergent manner, and their spectroscopic/chromatographic properties were compared with those of the authentic material. The present study culminated in the reassignment of the C5/C7 relative configuration, assignment of the C12/C18 relative configuration, and determination of the absolute configuration of amphirionin-2.


Introduction
Marine polyketides are an important source of new chemotherapeutic agents for the treatment of cancer. 1 As such, the structure, synthesis, and biological function of this class of natural products have gained signicant interest from the chemical community. 2 Marine polyketides are mostly noncrystalline, scarcely available substances from natural sources, and their complex structures are characterized mainly by NMR spectroscopic analysis. Integrated with quantum chemical calculations that enable the prediction of chemical shis and 3 J H,H values, 3,4 NMR-based structural assignment of stereochemically complex natural products has become more feasible than ever. Unfortunately, however, congurational assignment of remote stereogenic centers between which only negligible, if any, stereoelectronic and/or steric interactions exist, is still beyond the reach of NMR spectroscopic analysis and computational simulations. 5 Orchestration of chemical synthesis, NMR and other spectroscopic techniques and, where appropriate, chromatographic analysis is indispensable for achieving complete congurational assignment of complex natural products. [6][7][8] Amphirionin-2 (putative structures 1 and 2, Fig. 1) is a linear polyketide metabolite, isolated from cultured cells of the marine benthic dinoagellate Amphidinium sp. KCA09051 strain. 9 Amphirionin-2 exhibited potent cytotoxic activity against the human colon carcinoma Caco-2 cell line and the human nonsmall cell lung adenocarcinoma A549 cell line with IC 50 values of 0.1 and 0.6 mg mL À1 , respectively, whereas it showed only moderate cytotoxicity against the human cervix adenocarcinoma HeLa cell line (20% growth inhibition at 1 mg mL À1 ). Furthermore, amphirionin-2 displayed in vivo antitumor activity against murine tumor P388 cells (T/C 120% at 0.5 mg kg À1 ).
The gross structure of amphirionin-2 was determined on the basis of extensive 2D-NMR analyses. The relative congurations of two unique hexahydrofuro [3,2-b]furan moieties were individually characterized based on NOESY correlations. The relative congurations of C4/C5 and C5/C7 were deduced from conformational analyses based on J values and NOESY correlations. The absolute conguration of C5 was assigned on the basis of a modied Mosher analysis. 10 However, the relative conguration between two remote stereogenic centers C12 and C18 could not be correlated by means of NMR-based structure analysis. Thus, the complete stereochemical assignment of amphirionin-2 needs to await its total synthesis.
Here we describe a unied, convergent total synthesis of amphirionin-2 and its three diastereomers for the rst time to determine the absolute conguration of this natural product in an unambiguous manner.

Results and discussion
Our synthetic blueprint toward 1 is summarized in Scheme 1A. The target structure 1 could be derived from 3 by a reductive cleavage of the le-end tetrahydrofuran ring. We envisaged that all the tetrahydrofuran rings found in 3 would be synthesizable by an extensive use of cobalt-catalyzed Mukaiyama-type cyclization of g-hydroxy olens. As shown in Scheme 1B, Inoki and Mukaiyama have reported that the reaction provides a diastereoselective access to 2,5-trans-2-hydroxymethyl tetrahydrofuran derivatives V from g-hydroxy olens I in the presence of appropriate cobalt(II) chelate complexes under O 2 atmosphere (hereaer referred to as Mukaiyama cyclization), 11 and its mechanism involves radical intermediates II, III, and IV. 11,12 Later, the Hartung group has demonstrated that the carboncentered radical intermediate IV can be trapped with various radical terminators to deliver 2,5-trans-tetrahydrofuran derivatives VI (hereaer referred to as Hartung-Mukaiyama cyclization). 13 We envisioned that 3 should be synthesized via a Stille-type reaction 14 of vinylstannane 4 and iodoolen 5. This late-stage fragment assembly would also enable an access to diastereomer 2 from 4 and ent-5 (latter not shown). Vinylstannane 4 would be accessible from olens 6 and 7 through an olen cross-metathesis 15 and subsequent Hartung-Mukaiyama cyclization of the derived internal olen. We were aware of the uncertainty of this retrosynthetic disconnection because Mukaiyama-type cyclization has mainly been applied to terminal olens at early stages of total synthesis 16 and its versatility toward internal olens remained ambiguous. Moreover, Hartung-Mukaiyama cyclization has rarely been utilized in complex molecule synthesis. 16g Nevertheless, it appeared worthwhile to pursue this approach that allows for a convergent access to the tricyclic ether skeleton of 4. Olens 6 and 7 were traced back to g-hydroxy olens 8 and 9 by considering Mukaiyama cyclization, respectively. Meanwhile, iodoolen 5 would be derived from g-hydroxy olen 10 via a Hartung-Mukaiyama cyclization. In turn, 10 would be available from ghydroxy olen ent-9 by means of a Mukaiyama cyclization.
Completion of the total synthesis of 1 and 2 is illustrated in Scheme 4. Stille-type reaction of vinylstannane 4 (1 equiv.) with iodoolen 5 (1.1 equiv.) was non-trivial and required optimization of reaction conditions. Initial experiments showed that the reaction under palladium catalysis ([Pd 2 (dba) 3 $CHCl 3 ]/ Ph 3 As with or without CuI) provided (E,E)-diene 3 in only low yield and resulted in signicant side reactions, including isomerization of the C15-C16 double bond and homodimerization of 5 (Table S3, ESI †). It was eventually found that the reaction was best performed by using CuTC in NMP 33 at room temperature, giving 3 in 83% yield with essentially no erosion of the conguration of the double bonds. The conguration of the diene moiety of 3 was conrmed to be E,E by NOESY correlations and a coupling constant ( 3 J H15,H16 ¼ 15.6 Hz). Cleavage of the TBDPS ether of 3 with TBAF delivered alcohol 24 in 96% yield. Aer iodination (I 2 , Ph 3 P, imidazole, 76%), the derived iodide 25 was exposed to excess zinc dust in acetic acid to furnish 1 (97%). The diastereomer 2 was synthesized from 4 and ent-5 in the same manner.
The 1 H and 13 C NMR spectra of 1 and 2 revealed that both compounds were not identical with natural amphirionin-2 (for assignment of 1 H and 13 C NMR signals, see Tables S4 and S5, ESI †). These results indicated the necessity of re-examination of the original structural assignment of the natural product. The 1 H NMR chemical shis of the C1-C12 moiety of synthetic 1 and 2 were signicantly deviated from those of the corresponding moiety of the natural product, whereas the 13 C NMR signals of synthetic 1 and 2 were similar to those of the authentic material and inconsistencies were limited to the C5-C10 moiety. Signicantly, 1 and 2 were distinguishable from each other by 1 H NMR analysis despite the C12 and C18 stereogenic centers being separated by six carbon-carbon bonds. Careful comparison of the 1 H NMR spectra of 1 and 2 revealed subtle differences in signals assigned for H-11, H-16, H-17, and H-18 (Fig. S2, ESI †). With respect to these protons, the 1 H NMR chemical shis of 2 rather than 1 were in better agreement with those of natural amphirionin-2. It is known that stereoelectronic and/or steric interactions between two stereogenic centers separated by two or more methylene units are negligible by NMR spectroscopy. 5 In the present case, the C13-C16 conjugated diene would be responsible for unusual long-range stereochemical interactions between the C12 and C18 stereogenic centers. 34 We considered that the relative conguration of C12/C18 of natural amphirionin-2 might be same as that of synthetic 2, and that the relative conguration of C4/C5 and/or C5/C7 of the original stereochemical assignment should have been incorrectly assigned. Re-Examination of NOESY correlations and 3 J H,H values of natural amphirionin-2 suggested that the relative conguration of C5/C7 of the natural product might be opposite to that of the proposed structures 1 and 2 (Fig. S1, ESI †).
inconsistency around 2.04-2.10 ppm in the 1 H NMR spectra of synthetic 27 and 28 and natural amphirionin-2 was due to the 5-OH signal, which was rmly assigned on the basis of COSY experiments. The 5-OH signal disappeared upon addition of a drop of D 2 O (Fig. S3 and S4, ESI †). Therefore, we determined that the 1 H NMR spectrum of 27 matched that of natural amphirionin-2. The 13 C NMR spectra of synthetic 27 and 28 were completely indistinguishable from each other.
Moreover, chiral HPLC analysis (Chiralpak IB N-5: 4.6 mm I.D. Â 250 mm; eluent: 10% i-PrOH/n-hexane; ow rate: 1.0 mL min À1 ; UV detection: 254 nm) demonstrated that the retention time of 27, 28, and natural amphirionin-2 was 8.9, 7.7, and 8.8 min, respectively (Fig. S5, ESI †). Co-Injection of synthetic 27 and authentic amphirionin-2 resulted in a single peak under the above analytical conditions. Thus, we concluded that the relative conguration of amphirionin-2 is same as that Finally, we evaluated the cytotoxic activity of our synthetic 1, 2, 27, and 28 against a small panel of human cancer cell lines, including the non-small cell lung adenocarcinoma A549, the cervix adenocarcinoma HeLa, the acute T cell leukemia Jurkat, and the chronic myelogenous leukemia K562 cell lines by WST-8 assay (Fig. 3). A549 cells showed biphasic response to synthetic 1, 2, 27, and 28. The viability of A549 cells decreased to 39-65% at 10 mM, increased to 63-73% at 30 mM, and then underwent to 11-38% at 100 mM. The sensitivity of A549 cells toward synthetic 27 was more moderate than that expected from the results reported in the isolation paper. 9 This apparent discrepancy would be ascribable to the difference of the source of cells and/or experimental conditions, as a similar difference in potency was observed for the positive control doxorubicin: IC 50 0.6 mM (this work) versus IC 50 0.07 mM (ref. 9). A similar biphasic response was observed for HeLa cells upon exposure to our synthetic compounds; the cell viability declined to around 60% at 3 mM, restored to 86-98% at 30 mM, and dropped to 10-51% at 100 mM. These results suggested the possibility that our synthetic compounds would have at least two different mechanisms of action in A549 and HeLa cells. In contrast, Jurkat and K562 cells responded to our synthetic compounds in a dosedependent manner. Jurkat cells were found to be more sensitive than K562 cells toward these compounds. Overall, our synthetic 1, 2, 27, and 28 showed cell line-dependent cytotoxic activity, whilst their stereochemistry did not have signicant correlation with their cytotoxic potency.

Conclusions
A unied total synthesis of four candidate stereoisomers 1, 2, 27, and 28 of amphirionin-2 was completed in 17 linear steps from diol 11 or 19 linear steps from benzyloxyacetaldehyde. The salient feature of the present work is an extensive use of cobaltcatalyzed Mukaiyama-type cyclization of g-hydroxy olens in stereocontrolled construction of all the tetrahydrofuran rings of amphirionin-2. The present study illuminates the versatility of Hartung-Mukaiyama cyclization of g-hydroxy olens, and also expands the reaction scope by demonstrating the synthesis of complex 2,5-trans-substituted tetrahydrofurans from internal olens (e.g., 16E/16Z / 17). A late-stage CuTC-mediated Stilletype reaction for convergent assembly of two hexahydrofuro[3,2b]furan moieties is another important feature of our synthesis, which enabled rapid access to four candidate stereoisomers.
The 1 H and 13 C NMR spectroscopic data of originally assigned structures 1 and 2 showed non-identity of these compounds with natural amphirionin-2 and suggested the necessity to reassign the C5/C7 relative conguration. Eventually, the relative conguration of this natural product was fully established on the basis of the 1 H and 13 C NMR spectroscopic data and chiral HPLC chromatograms of 27 and 28 with authentic reference. The absolute conguration was determined by comparing the specic rotation value and CD spectroscopic data of 27 and 28 with those of the authentic material. Thus, it was concluded that the absolute conguration of amphirionin-2 is shown by the structure 27.

Conflicts of interest
There are no conicts to declare.