One of the most difficult bits of the synthesis of the 3D compound was the formation of the cis tri-substituted double bond:
This unit was prepared by a Suzuki coupling of the vinyl iodide and the corresponding right hand side fragment. The synthesis of the vinyl iodide was the problem step.
The obvious route is to use a Wittig reaction. A variant of this reaction was developed by Zhao, he used it to prepare 1-iodoalkyl ylides by treating a phosphonium salt with strong base to give the phosphorane which was then iodinated (I2) to give the iodo phosphonium salt. Further base treatment and reaction with an aldehyde produces the cis-vinyl iodide exclusively in reasonable yield (40-55%). So we tried the following reaction:
The yield varied between 21-31% after chromatography with the Z/E ration about 10:1 – 15:1 (traces of the E olefin were separated at the next step). So this reaction was not reproducible in any sense of the word except that in the event the lousy reproducibility was reproduced! I did not observe any des-iodo olefin, suggesting that the formation of the iodo ylide from ethyltriphenylphosphonium iodide via ylide iodination had been completed before it was added to aldehyde. So in order to be careful with material I restricted the maximum batch size to 2Kg of aldehyde. Some factors contributing to the poor yield were; instability of the aldehyde, complicated work-up and apparent instability of the vinyl iodide to the work-up conditions, just to name a few.
The method called for iodination using elemental iodine making a darkish-red reaction mixture almost black. While colour change is not usually a reliable indicator of reaction success it certainly helps to see if anything is going on. During process optimisation work I found that N-iodosuccinimide can be used to replace iodine without detriment. While this makes the reaction easier to handle and increased the reproducibility of the process, it did not contribute to an increase in yield. I also observed the formation of the methyl ketone (a) during the work-up accordingly it was changed to a non-aqueous work-up. Consequently this by-product and the aforementioned stability problems completely disappeared! This allowed a scale-up to 3kg per reaction and the yield over 9 reactions on a scale of 3kg per reaction of aldehyde was 31% ± 0.5%. Also isolated was the epoxide (b) as a 1:1 mixture of diastereoisomers in the same yield as the desired product. The formation of the epoxide was also noted by Smith who proposed a mechanism of its formation.
Alternative approaches were investigated in an attempt to minimize this major by-product, however, they were unsuccessful. For example, employing a method described by Shen [Perkin Trans. I, 1995, 1331] (where the initially formed betaine intermediate was deprotonated with a second equivalent of base and then iodinated) produced the des-iodo olefin. Utilizing Hanessian’s phosphonates in this process also resulted in only des-iodo olefin. There are other ways of preparing this structural unit, for example Panek described a very nice route: The key steps are hydrozirconation-iodination of a (1-alkynyl)trimethylsilane followed by Negishi-type cross-coupling. The resultant (Z)-vinyl silane is iododesilylated and subjected to a second cross-coupling reaction to give the trisubstituted olefin. I looked at this route, and it works well, however the yield over the 4 steps was not really better than the Zhao method, and it added 4 more steps to an already long synthesis. So we stuck with the Zhao protocol.
Apart from triphenylphosphine oxide we found significant amounts of triphenylphosphine to be present and interestingly this was not observed during reaction monitoring. Triphenylphosphine was isolated after the chromatography on silica-gel. This suggested the presence of an intermediate which decomposed during the contact with silica-gel delivering the observed triphenylphosphine. If one accepts that the mechanism proposed by Smith is correct the protonated form of intermediate (c) can also collapse to an unstable iodo epoxide with the elimination of triphenylphosphine. This may especially relevant during a non-aqueous work-up.
The following anecdote illustrates the pitfalls into which one can stumble. During the scale-up of this process the first reaction did not deliver the required yield of 30%, only some 18% of vinyl iodide was isolated along with starting material. This result created some consternation until it was realised that the sodium hexamethyldisilazide, used as a base in this sequence, is supplied as a 35% solution in THF. As we carried out the reaction in winter it was quite cold and somehow the THF solution of sodium hexamethyldisilazide apparently separated in the drum, we assumed it was homogeneous! For the next reaction the drum was shaken vigorously! The result 32% yield! A label was applied “shake well before use”.
