An Unexpected Transannular [4+2] Cycloaddition during the Total Synthesis of (+)‐Norcembrene 5

Abstract

We report a concise and versatile total synthesis of the diterpenoid (+)‐norcembrene 5 from simple building blocks. Ring‐closing metathesis and an auxiliary‐directed 1,4‐addition are the key steps of our synthetic route. During the synthesis, an unprecedented, highly oxidized pentacyclic structural motif was established from a furanocembranoid through transannular [4+2] cycloaddition.

They may look alike… During studies toward the total synthesis of the norditerpene natural product norcembrene 5, one diastereomer of a key intermediate (α‐OAc at C13) underwent clean and spontaneous Diels–Alder cycloaddition at room temperature (see scheme). However, the corresponding β isomer did not undergo a Diels–Alder reaction, but was oxidized to the corresponding furanone, which was further transformed into norcembrene 5.

Norcembrenolides are a large and diverse family of norditerpene natural products. The main source for the isolation of norcembrenolides are gorgonian soft corals found in the Western Atlantic Ocean.1 Most of these natural products were isolated from soft corals of the genus Sinularia (family: Alcyoniidae). Representative norcembrenolides 1–8 exhibit a variety of different functional‐group patterns embedded in a macrocyclic ring, which is present in all congeners (Figure 1).2 Several of these compounds show biological and pharmacological activity, such as cytotoxicity or antiviral properties. Gyrosanin A2d (2) tested positive for cytotoxicity against P‐388 cancer cell lines (mouse lymphocytic leukemia). Sinuleptolide2c (3), norcembrenolide/5‐episinuleptolide2c (4), scabrolide E2e (5), leptocladolide A2f (6), and 7E‐leptocladolide A2f (7) showed strong to moderate cytotoxic activity both against KB (human oral epidermoid carcinoma) and Hepa59T/VGH (human liver carcinoma) cancer cell lines. Sinuleptolide (3) furthermore exhibited antiviral activity against HCMV (human cytomegalovirus) cells.2d Despite their unique molecular structure featuring a 14‐membered cembrane ring, a bridging dihydrofuranone, and a lactone or ester motif, only a few accomplished (semi)syntheses of norcembrenolides have been reported so far.2b, 3

Figure 1.

Representative norcembrenolide natural products.

Norcembrene 5 (1) was first isolated in 1985 by the groups of Fenical and Clardy.2a Its absolute configuration was not determined so far, and both enantiomeric structures of norcembrene 5 were reported in later isolations of this compound, all referring to the original publication.2c, 4 Therefore, its absolute configuration remained ambiguous.

Herewith, we report the total synthesis of norcembrene 5 as well as the elucidation of its absolute configuration. Comparison of the optical rotation of our synthesized material with the literature value revealed opposite signs (synthetic +51.4 vs. −77 for the isolated material), thus establishing the natural product as (−)‐norcembrene 5.2a Retrosynthetically, (+)‐norcembrene 5 (1) can be assembled from furanocembranoid 9 (Scheme 1). In a biomimetic3a oxidation/transannular cyclization cascade, the furan moiety is cleaved to the 3‐furanone motif present in the natural product. Methanolysis of the butenolide completes the transformation of 9 into 1. The macrocycle of 9 is constructed by ring‐closing metathesis (RCM) from triene 10, which is accessible from aldehyde 11 and selenolactone 12 by an aldol reaction. The stereocenter of 11 is introduced through auxiliary‐directed 1,4‐addition to compound 13, available from 2‐allylfuran (14) by cross‐metathesis.

Scheme 1.

Retrosynthetic analysis of (+)‐norcembrene 5 (1).

Our synthesis began with the preparation of enantiomerically pure aldehyde 11 (Scheme 2). Cross‐metathesis of 2‐allylfuran5 (14) with acrylic acid and the Hoveyda–Grubbs II catalyst afforded unsaturated carboxylic acid 15 in high yield.6 The Evans auxiliary for the stereoselective 1,4‐addition was attached by amidation of the lithiated oxazolidinone with the acid chloride formed in situ from 15 to furnish enone 13 in 81 % yield.7 Diastereoselective installation of the isopropenyl moiety was performed using freshly recrystallized CuBr⋅SMe₂ as the copper source.8 Under optimized conditions, compound 16 was obtained in quantitative yield as a single diastereomer. Subsequent reductive cleavage of the auxiliary directly afforded aldehyde 11 in 77 % yield.

Scheme 2.

Synthesis of enantiomerically pure aldehyde 11. DIBAL‐H=diisobutylaluminum hydride, DMF=N,N‐dimethylformamide.

In the next sequence, the macrocycle of the furanocembranoid scaffold was established (Scheme 3). First, aldehyde 11 was assembled with selenolactone 12 (available in three steps from (S)‐(−)‐glycidol) through an aldol reaction and subsequent oxidation/elimination under the conditions reported by Mulzer and co‐workers.9 The two diastereomeric butenolides 17 a and 17 b were obtained in a ratio of 3.2:1, and were separated and used independently for further transformations. Acetylation of the secondary alcohol followed by Vilsmeier–Haack formylation10 of the furan ring furnished aldehydes 19 a and 19 b, both in high yield. Subsequent olefination of the aldehyde functionality initially caused synthetic drawbacks. After extensive experimentation, we found that the diastereomers 19 a and 19 b behaved differently in the Wittig olefination. Compound 19 a was successfully converted into 10 a by using KHMDS as a base for ylide formation in 72 % yield. By contrast, for the conversion of diastereomer 19 b, LiHMDS gave 10 b in better yield (64 %) than the use of KHMDS (47 %).

Scheme 3.

Synthesis of furanocembranoids 20 a/b by a ring‐closing metathesis reaction. DMAP=4‐dimethylaminopyridine, HMDS=hexamethyldisilazide, LDA=lithium diisopropylamide.

With compounds 10 a and 10 b in hand, ring‐closing metathesis as the key step of the synthetic route could be performed. Initial studies for this transformation involving the continuous addition of Grubbs II catalyst to the starting material in refluxing benzene resulted in very inconsistent product yields varying from 9 to 53 %. After screening a number of different reaction parameters, two distinct changes eventually gave reasonable access to the desired macrocycles 20 a and 20 b. First, 1,4‐benzoquinone was added to the reaction mixture in catalytic amounts. This additive has been reported to prevent isomerization during olefin metathesis,11 but we suppose that it operated as scavenger for decomposition species formed by the catalyst at high temperatures. Second, the catalyst was added in several portions to the reaction mixture instead of being added continuously. Under these optimized conditions, furanocembranoids 20 a and 20 b could be prepared in 58 and 69 % yield, respectively.

For further functionalization, the trisubstituted double bond of the macrocycle needed to be hydrated stereo‐ and regioselectively (Markovnikov). Since direct hydration, such as Mukaiyama hydration, failed, a two‐step sequence consisting of Upjohn dihydroxylation12 followed by deoxygenation of the secondary alcohol was applied (Scheme 4). Under similar conditions as established by Theodorakis and co‐workers, furanocembranoid 20 a was treated with a mixture of OsO₄ and NMO to enable site‐selective dihydroxylation of the C−C double bond2b to give diol 21 a as a single diastereomer in 46 % yield. After the preparation of 21 a, we observed slow but spontaneous conversion of this compound into another product when stored in solution. Thorough characterization by 2D NMR spectroscopy revealed that 21 a underwent a transannular [4+2] cycloaddition between the furan and the butenolide moiety. In this transformation, pentacyclic compound 22, the exo‐Diels–Alder product, was formed as a single diastereomer, as confirmed by NOESY experiments. This unprecedented highly oxygenated and congested structure contains a quaternary carbon center and eight other stereocenters (five of them contiguous). This motif is of interest for further structure–activity relationship (SAR) studies in the future. For better characterization, we heated 21 a in benzene at reflux to promote full conversion into 22. Despite this unexpected transannular reaction we were able to convert diol 21 a into 9 a by deoxygenation of the secondary alcohol moiety. Treatment of 21 a with Et₃SiH and BF₃⋅OEt₂ was fast enough to furnish alcohol 9 a in 47 % yield,2b before 21 a was able to undergo the undesired transannular [4+2] cycloaddition. Similarly, 9 a was prone to undergo transannular [4+2] cycloaddition to give pentacycle 23 as a single product featuring identical stereochemical relationships.

Scheme 4.

Site‐selective hydration of furanocembranoid 20 a and [4+2] cycloaddition products 22/23 with key NOESY correlations. NMO=4‐methylmorpholine N‐oxide.

This transannular [4+2] cycloaddition is of general interest, since similar transannular cycloadditions of these natural products have been reported previously and seem to reflect a general reactivity trend in this natural product family. In their synthesis of intricarene, Trauner and co‐workers relied on a transannular [5+2] oxidopyrylium cycloaddition of the similar furanocembranoid bipinnatin J to obtain the natural product.13 Another example is the transannular [2+2] photocycloaddition in an approach to bielschowskysin by West, Roche, and co‐workers through dearomatization of a furan ring with a concomitant transannular [2+2] cycloaddition reaction.14b These previously reported transannular cyclizations are considered to occur in the actual biosynthesis of these natural products, thus contributing largely to the architectural diversity in the molecular scaffolds present in furanocemranoid diterpenes. Therefore, it is highly likely that in future isolation efforts, novel natural product congeners with the molecular architecture of 22 and 23 will be discovered from Sinularia‐type soft corals, thus comprising an additional case of “natural product anticipation”.15

The two‐step sequence for hydration of the C−C double bond was also applied to diastereomeric furanocembranoid 20 b (Scheme 5), since direct Mukaiyama hydration proved to be unsuccessful as described for 20 a. However, applying the same reaction conditions for dihydroxylation (OsO₄/NMO) as for 20 a only gave unsatisfactory yields. The best result that could be obtained was a 24 % yield of diol 21 b despite screening different temperatures, solvent systems, and concentrations. Other reagents for dihydroxylation, such as AD‐mix, K₂OsO₄/NMO, or RuCl₃/NaIO₄, showed no reaction at all. Nevertheless, the synthesis was continued, and by treatment of 21 b with Et₃SiH/BF₃⋅OEt₂, alcohol 9 b was afforded in 36 % yield (61 % brsm). To our surprise, neither compound 21 b nor compound 9 b showed any tendency to undergo transannular cycloaddition at all (see Figure 2 for further details). Oxidation of the furan moiety of 9 b, followed by 5‐exo‐trig cyclization of the tertiary alcohol, was triggered by treatment with the Jones reagent to afford norcembrenolide 24 in high yield.2b, 16 Finally, a sequence of deprotonation of the butenolide and elimination of the acetate group via intermediate 25, followed by methanolysis of the lactone gave access to (+)‐norcembrene 5 (1) in good yield. The absolute configuration of 1 was established with the help of single‐crystal X‐ray analysis of furanocembranoid 20 b. Since the C1 stereocenter was introduced earlier in the synthesis through the enantioselective 1,4‐addition, the configuration of all other stereocenters could be deduced by correlation.

Scheme 5.

Preparation of (+)‐norcembrene 5 (1) from furanocembranoid 20 b.

Figure 2.

Comparison of the Gibbs free energies of compounds 21 a/b and the corresponding products of transannular [4+2] cycloaddition 22/22 b via 21 a/b ^≠ as based on DFT calculations.

Since there was no obvious explanation for the different reactivity of compounds 21 a/9 a (transannular [4+2] cycloaddition) and their diastereomers 21 b/9 b (no transannular reaction), we performed DFT calculations for further clarification (Figure 2; for experimental details and calculated structures, see the Supporting Information). We calculated the Gibbs free energies of diol 21 a and its reaction product 22 via transition state 21 a ^≠, as well as the Gibbs free energies of diol 21 b and its (not observed) reaction product 22 b via transition state 21 b ^≠. The first interesting result was the clearly higher free energy and therefore higher reactivity of 21 a in comparison to 21 b (ΔG=11.9 kcal mol⁻¹) in the ground state. We also observed a lower free energy of transition state 21 a ^≠ as compared to the possible transition state 21 b ^≠ (ΔG=4.47 kcal mol⁻¹). These results imply activation energies of 22.8 kcal mol⁻¹ for the transition of 21 a to 21 a ^≠ (black dash) and 39.2 kcal mol⁻¹ for the transition of 21 b to 21 b ^≠ (red dash). In conclusion, the activation barrier for diol 21 b is almost twice that for 21 a. It appears that for this reason, the transannular Diels–Alder reaction of 21 a takes place, whereas compound 21 b fails to undergo the transannular reaction. Since compounds 9 a, 9 b, and 23 have the same molecular scaffolds and only differ in their substitution pattern from 21 a, 21 b, and 22, analogous behavior can be presumed.

In conclusion, we have established the absolute configuration of (−)‐1 and completed a concise synthetic route to its enantiomer (+)‐norcembrene 5 (1) in 13 steps (requiring isolation of the product) from 2‐allylfuran (14). An optimized ring‐closing metathesis reaction was applied as a key step to assemble the 14‐membered carbocyclic scaffold. Our route is very versatile and could be adopted for the synthesis of further furanocembranoids and norcembrenolides. In addition, we demonstrated the formation of the unprecedented and highly congested pentacyclic structures 22 and 23 through transannular [4+2] cycloaddition. The frequent occurrence of transannular ([5+2] and [2+2]) cycloaddition reactions in furanocembranoid (bio)synthesis strongly suggests that the molecular scaffold obtained from our transannular [4+2] cycloaddition may be yet another case of natural product anticipation.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

M. Breunig, P. Yuan, T. Gaich, Angew. Chem. Int. Ed. 2020, 59, 5521.