Totally Synthetic by Paul H. Docherty, 25 August 2009

Total Synthesis of Sorangicin A

Smith

A. B. Smith, III, S. Dong, J. B. Brenneman, R. J. Fox, J. Am. Chem. Soc. 2009, 131, 12109-12111.

It’s not often that I have to draw such an exotic structural motif into a target, but sorangicin A is more than just a structural challenge - it’s an antibiotic, with broad-spectrum activity against both Gram-positive and Gram-negative strains. The synthetic challenge was taken up by a variety of groups, including those of Crimmins and Lee, but Amos Smiths finished the first total synthesis.

Such large macrolides can be best analysed with retrosynthesis. Splitting the molecule into four fragments, lactonisation is the most obvious disconnection, that can be performed with DMAP. Two isolated alkenes were installed using Julia olefination, whilst a Stille coupling can be used for the formation of the Z,E,Z-triene. That leaves the simple matter of constructing the fragments...

I’m starting with a look at the dioxabicyclo[3.2.1]octane unit first: The chiral glycolic aldehyde derived from l-gulonic acid γ-lactone was treated with a chromium Schiff base and Danishefskys diene, promoting a hetero-Diels-Alder reaction and formation of a second stereocenter. For the derivatisation of the enone in a Noyori three-component coupling protocol, bromostyrene was first lithiated, and the metal then switched for zinc, followed by addition of the enone, resulting in a Michael addition type product. This was then quenched by addition of methyl iodide, which required a bit of optimisation. Initially, ten equivs of MeI in HMPA were used, giving a reasonable yield of the product as a single stereoisomer; however, dimethylation also occured. The yield was improved by mediating the bascity of the intermediate by addition of CuI•PBu₃ just prior to the methyl iodide. A very interesting synthetic step, and an astonishing yield over so many steps.

For the cyclisation of the dioxabicyclo[3.2.1]octane unit, the groups used a stereoselective reduction and some protection group transformations to yield the key intermediate. The precursor was treated with base to form a terminal epoxide as expected. This was then attacked by the other hydroxyl group, resulting in the formation of a bicycle. Nice approach. Completion of this fragment required only formation of a vinyl iodide via Takai olefination, and an oxidative clevage of the styrene to reveal an aldehyde. Interestingly, they used a two-step protocol - a Sharpless dihydroxylation, followed by addition of periodate. Presumably the more direct ozonolysis was unselective, and attacked the freshly installed vinyl iodide.

The tetrahydropyran-containing unit was created using a more familiar approach. Indeed, this piece was built by using a Suzuki-Miyaura coupling, whilst the pyran section was created by a boron aldol reaction, then forming a ketal at C-23. The extraneous methanol was removed, leaving the desired pyran in an impressive brevity.

Work on the dihydropyran (DHP) fragment was conducted via a coupling of an enone with the ready elaborated sidechain. The reaction was sluggish, as the enol functionality was again working against the group. However, forming a cuprate of the vinyl bromide solved this, allowing a highly selective and reasonably high yielding reaction (60%, 20:1 d.r.). Btw: A Myers alkylation provided the single stereogenic center in the vinyl bromide, whilst a hetero-Diels-Alder reaction using Danishefskys diene and the same Jacobsen catalyst gave the DHP.

Removing the silane subsequently to an oxidation forms a ketone rather than the required DHP. A regioselective enolisation using LDA allowed selective formation of the correct alkene. Triflation using Comin’s reagent then set them up for some slightly unusual palladium chemistry - a catalytic reduction using tributyltinhydride as the terminal reductant. Very interesting and very selective.

Unfortunately, the C-10 secondary alcohol was missassigned in the original paper. Inversion was required, but a Mitsunobu reaction failed. Ring-bound inversions are subject to the complex conformation thermodynamics, and the allylic nature doesn’t help either. In the end, they performed a Ley oxidation / Luche reduction sequence. The primary silanol was deproteced and the sulfone required for the Julia olefination appended by Mitsunobu reaction followed by oxidation of the sulfide.

In another Julia coupling, the group coupled the DHP-diol fragment with the dioxabicyclo[3.2.1]octane aldehyde to complete the left-hand side. However, and to no massive surprise, the olefination required a lot of optimisation. The choice of base, and more specifically, counterion, was highly influential - tBuLi in HMPA/DMF was the winner, with a yield of 39%. Next up was the second Julia coupling, this time with the now epimerised DHP. The lessons learned in the first coupling were useless - this time, KHMDS in DME prevailed, delivering 86% product solely as the desired trans- isomer.

Now, the carbon skeleton needed only the triene unit - definitely the most fragile moiety. However, a Stille coupling with a dienyl stannane was very effective - especially as a ‘less risky’ dienyne route was prone to isomerisation. An 88% yield for that step is magnificent. Notable is the use of excess Ph₂PO₂NBu₄, which prevented the isomerisation. The last major step is the cyclisation. An effective protocol for the macrolactonisation was found using a modified Mukaiyama reagent, containing a tetrafluoroborate counterion. Smith says that this non-nucleophillic counterion was key in preventing isomerisation of the triene, and allowed ring closure in an impressive 85%.

Global deprotection was again, non-trivial. They needed to remove both a t-butyl ester and a MOM group in the presence of an acid. It’s amazing to see how much solvent choice influences the reactivity of acids. TMSOTf was reactive enough to remove both groups, but isomerised the triene. So a stepwise protocol was used, in which the less reactive TBSOTf was used to cleave the ester, then mineral acid used to remove the MOM group. This must have been exceptionally frustrating to conduct, but it delivered deprotection in 70% over the two steps, and the natural product. A very interesting total synthesis!