Update 01/10/11 - It seems that I never actually gave the references for the original papers. The synthesis was actually published in three back-to-back JACS papers - the first is J. Am. Chem. Soc., 1981, 103, 3210, and you can read on from there. I also found the relevant synarchive page to be helpful when writing this.
I hadn't planned to cover this synthesis, Woodward's last, so early in this series, but as a review on the use of thiopyrans as templates in polypropionate syntheses was recently published in Chem. Commun. it seems timely to mention it now. Woodward once said in a talk at CIBA in India that
"Much of the art of directed synthesis involves the design of ways to place constraints on molecular motion, with the aim of bringing about desired changes and suppressing others"
A popular way of doing this, as has been said before, is through the use of cyclic templates, a tactic used extensively by chemists of the Woodward and Corey eras. The ease with which desulfurisation can be accomplished using Raney Nickel makes thianes and thiopyans uniquely suitable as temporary rings which can be cleaved mildly and selectively later on. This property made them the cornerstone of Woodward's approach to erythromycin A where they were used to set 8 of the 10 stereocentres found on the macrocyclic ring.
The route began with the alkylation of the (tetrahydro-4H-thiopyran-4-one derived) hemidithioacetal with the mesylate shown. This was followed by cleavage of both dimethylacetals with a bit of aqueous acetic acid, which left the dithioacetal unperterbed, and is a good example of why I both love and hate these compounds. Initially the group used an enantioenriched form of the thiopyran as the starting material, leading to an inseparable 1:1 mixture of diastereomeric ketoaldehydes. Interestingly, when investigating conditions for the aldol-type cyclisation of these compounds they found that L-proline produced two (now separable) diastereomers in greatly diminished ee (10-12%). Conversely, D-proline gave the same mixture in a similar yield but with a much larger ee (80-86%). Intrigued by this unexpected result the group prepared the racemic ketoaldehyde (again as a 1:1 mixture of diastereomers used 'as is') and subjected them to 'aldolization' with a catalytic amount of D-proline. This lead to the same 1:1 separable mixture of diastereomeric β-hydroxyketones, both formed with modest asymmetric induction (ee 36-38%). The desired diastereomer was then separated and dehydrated to give an enone from which the required enantiopure building block could be obtained by fractional crystallisation from hexane-benzene. This proved exceedingly efficient, with the desired enatiomer obtained in >95% ee and the mother liquor containing material which was almost racemic, allowing the synthesis to proceed. Although the overall yield of this compound was only 10-12% from the starting hemidithioacetal, the sequence was practical and scalable enough that the group were not held back by this potential bottleneck and could continue with the synthesis.
A few minor adjustments then gave the acetonide which would be used twice to construct the backbone of the natural product. I notice that there's no mention of a co-oxidant in the dihydroxylation, so this could be another instance where the group again used stoichiometric osmium tetroxide (a great example is Woodward's cortisone synthesis, where one step uses 68.5g OsO4!). However, as the Upjohn procedure (i.e. the use of catalytic OsO4 withNMO as the terminal oxidant) was reported in 1976, they don't really have any excuse here!
The group then took this bis-thiane and developed it into two quite different looking building blocks. The first, which retained the bicyclic core, was obtained simply by MOM deprotection (in the presence of the acetonide... I wouldn't have relied on that working.) and Swern oxidation of the newly revealed secondary alcohol. The second building block was obtained by desulfurisation of both rings using Raney Nickel, which also removed the benzyl protecting group. The primary alcohol was then dehydrated to give the terminal alkene according to the Grieco protocol, and this was converted to the corresponding aldehyde by ozonolysis.
The aldehyde and ketone fragments thus prepared were then joined using an aldol reaction using the rather hindered mesityl lithium as the base. Use of less hindered bases lead to epimerisation of the aldehyde α-position leading to complex mixtures of diastereomers. The aldol product was then oxidised to the diketone and converted to the enone. In order to selectively effect 1,2 reduction of this to the required diastereomer it was first converted to the benzyl mercaptan adduct before treatment with LiAlH4 and acetate protection. Finally, desulfurisation was carried out, removing the last of the temporary rings which had performed admirably in directing events so far. Also lost was the protected thio and the benzyl protecting group, whose alcohol was converted as before to the aldehyde, with loss of one carbon. The final carbon atoms in the macrocyclic backbone were then introduced by treatment of this aldehyde with the thioester enolate shown. Unfortunately, this proved highly selective for the isomer having the incorrect configuration at the thioester α-position. Eventually the group found that this centre could be epimerised to give almost exclusively the desired diastereomer when the compound was treated with 'excess' t-BuLi, via the trianion! The acetate ester was also lost during the acid quench.
The next thing the group did was investigate the lactonisation of the thioester and its derivatives, however all attempts were initially met with failure. Undeterred, they expanded their search for a suitable precursor to include more heavily modified variants. Eventually it was found that macrolactonisation could be carried out in good yield, providing that 1. the C-9 stereocentre was inverted to the S configuration and 2. cyclic protecting groups were present on the C-3/C-5 and C-9/C-11 diols. The inversion of the C-9 alcohol was not a major inconvenience as this was ultimately present in the natural product as a carbonyl group, so its stereochemistry at this stage was not important. However, rather than just inverting the centre to the epimeric alcohol the group exercised considerable foresight at this point and decided to also convert it to the corresponding primary amine (which was duly protected as the cyclic carbamate). The reason for this was that the molecule contains a large number of secondary alcohols and it was anticipated that selectively oxidising only the desired one at C-9 would prove problematic, whereas it was expected that the amine could be oxidised chemoselectively over the other alcohols without requiring further protecting group juggling. Returning to the macrolactonisation, the group found that the most successful method was the Corey-Nicolaou procedure that had served Corey so well in his own erythronolide studies, and thus the ester was converted to the necessary thioester.
Heating this thioester in refluxing toluene gave the desired macrocycle in an excellent 70% yield. From this point only deprotection, attachment of the sugars, and oxidation of the amine to the ketone were required for the completion of the synthesis. First of all, the two cyclic protecting groups were removed, but in order to stop the amine messing up the glycosylation it was protected with the somewhat uncommon p-phenylbenzamide group. After a lot of model studies and optimisation it was determined that there was a marked difference in the reactivities of the three secondary hydroxyl groups now exposed, and this could be exploited to allow attachment of the sugar motifs without the use of any more protecting group chemistry. Although the yields are a bit poor here, I think the directness of this approach is appealing and elegant. Finally, the amide was cleaved under reductive conditions with sodium amalgam. As expected, the amine could then easily be oxidised to the imine in the presence of the three secondary alcohols and this was hydrolysed to give the natural product!
For its time I think this synthesis was amazing accomplishment; only 25 years earlier did Woodward himself describe the erythromycin problem as 'quite hopelessly complex'. Unfortunately, Woodward didn't live to see his synthetic plan come to fruition due to his untimely death some two years before the completion of the project. In his stead Kishi oversaw the completion of the target as it acknowledged as such:
"We are indebted to Professor Yoshito Kishi for his help and encouragement and, in particular, for his acceptance of the role of principal investigator upon Professor Woodward's death."
It's a great shame that Woodward died at the age of only 62, although as pointed out by Frank Westheimer, if one accounts for the fact that he only slept around 4 hours each night, making use of parts of the day the rest of us waste, his actual age was more like 73. Still, I'd never get bored of reading his work and wonder what he might have tackled next. My PhD supervisor once told me one of his few regrets as a chemist was never having the chance to attend one of Woodward's legendary lectures. However, while a young postdoc at the ETH he was surprised to answer the phone one day and find himself speaking to none other than R. B. Woodward. I'm a little jealous.
1. Despite appearing online back in July it's still in the ASAPs (DOI: 10.1039/C1CC13323C). This methodology has proved surprisingly popular and contributed to the synthesis of good number of natural products in the decades following the untimely demise of RBW. The review was written by a former Woodward postdoc who worked on the erythromycin project, Dale E. Ward.
2. I think that a lot of people, myself included, think of RaNi as quite an exciting and unselective reagent, but actually desulfurisation can be done even in the presence of numerous double bonds. A neat application of this can be found in Stotter's formal synthesis of juvenile hormone I, a popular target of the late 60s and 70s, via Corey's intermediate. The way the group uses the same starting material to make two building blocks which are then reunited is cool, and conceptually quite similar to what Woodward did for erythromycin:
3. Corey's synthesis of erythronolide B (which, incidentally, is totally awesome) also relied on cyclic stereocontrol using a rigid tetrahydrochromanone ring system to help set many of the stereocentres. The methodology for asymmetric aldol reactions and allylations that is relied on for these targets today didn't exist back then.
Compare Woodward's and Corey's syntheses, with their cyclic precursors, resolutions and huge lists of coauthors to, for example, the recent White synthesis of 6-deoxyerythronolide B, accomplished entirely using acyclic stereocontrol by a single coworker. I wonder what we'll be doing in another 30 years...
4. Due to the amount of information to present here, I'll largely not be putting temperatures on the schemes except where they seem particularly relevant or interesting. Also, courtesy of the Hoye group at the University of Minnesota you can see the man himself talk about this work.