B.R.S.M. To strive, to seek, to find and not to give a yield


(-)-strychnine, (+)-aspidospermidine + 4 others

Collective synthesis of natural products by means of organocascade catalysis

or, "The shortest reported routes so far to (-)-strychnine, (+)-aspidospermidine, (+)-vincadifformine, (-)-kopsanone, (-)-kopsinine and (-)-akuammicine."

MacMillan et al., Nature, 2011, 475, 183–188; [PDF] [SI] [GROUP]

DOI: 10.1038/nature10232

A couple of weeks ago the MacMillan group published a remarkable paper reporting the asymmetric synthesis of six natural products, from three different alkaloid families. I only noticed total syntheses coming out of the MacMillan lab two or three years ago, but now they're something I always find myself getting excited over, and this is no exception.

MacMillan contends at the start of the paper, and he's not the first, that given enough effort, money, and time it's now possible to make at least a few milligrams of almost any known natural product. Just look at Kishi's palytoxin synthesis, for example, or the ongoing Nicolaou efforts towards maitotoxin sporadically appearing in JACS. We all know that it's when you need to make large amounts of any complex molecule (even grams), that your troubles really begin. Another thing that takes a huge amount of effort using conventional chemistry is the preparation of libraries of natural products (and their analogues) for biological testing.

Obviously, Nature is much more better at the construction of complex molecules, thanks largely to its peerless enzyme catalysts and amazing use of cascade reactions. Another thing that makes Nature's approach very efficient is that it tends to produce a number of natural products from a single intermediate, giving rise to families of compounds. The broadest example I can think of is the use of IPP and DMAPP to produce the terpenoids, to which entire series of books have been dedicated, but there are also many smaller groups of natural products thought to share a common biosynthetic precursor. Chemists, on the other hand, traditionally just choose one or two targets, although we are getting a bit better at this, and 'general methods' for the synthesis of families of natural products are now not uncommon. While many different definitions of 'the ideal synthesis' and 'efficiency' have been offered, T. Hud. stated a few years back what he thought was the paragon of modern synthesis:

"The highest possible level of craft requires the synthesis of an entire class of natural products by a unified approach resembling, in principle, their biogenesis. All members of a specific class (as well as all of their possible diastereomers) should be attained in an enantiodivergent fashion as single entities"[1]

This latest work from MacMillan group is one of the few papers I can think of which comes close to this exacting standard. Here, perhaps inspired by Nature, the group accessed a number of quite different alkaloids (spanning three families), in what they call 'collective total synthesis'.

The methodology used was originally developed for a total synthesis of strychnine, which has been a popular benchmark for synthetic chemists for many years. The common core structure the group chose for their collective synthesis is the tetracyclic aldehyde shown below, available as either enantiomer with equal ease using the group's imidazolidinone catalysts.

As in the group's recent syntheses of diazonamide a and minfiensine,[2] they again use propynal as the iminium ion precursor, which I imagine is very little fun to work with.[3] The first step here, as in the minfiensine paper, is the Diels-Alder reaction between the iminium ion dieneophile and the vinylindole. This is immediately followed by elimination of methylselenol and cyclisation of the amine side chain onto the indole 2-position. This is in contrast to the cascade used in the group's minfiensine synthesis, where the same starting material containing sulfur in place of selenium was used, and this group was not eliminated (due to its poorer leaving group ability), causing the reaction to take a very different course.

The aminal resulting from this cyclisation could be isolated by slightly modifying the reaction conditions, providing evidence that the route went via this intermediate and not via immediate 1,4-addition to the a,b-unsaturated iminium ion. Whether the rearrangement of this aminal to the final product occurs under iminium ion or simple Brønsted acid catalysis isn't known for sure, but in any case the key tetracycle was isolated in 82% yield with a very impressive e.e. of 97%. As shown above, use of the other enantiomer of the catalyst allowed the group access to the enantiomeric product required for vincadifformine or aspidospermine in essentially the same yield and e.e.

From here, the group first set their sights on (-)-strychnine and (-)-akuammicine. The first step was a decarbonylation, which unfortunately required using Wilkinson's catalyst stoichiometrically but did at least proceed with good yield.[4] The ester was then installed by reaction of the enamine with phosgene, followed by quenching with methanol. The enamine was then reduced and the Boc group removed, to give the product containing the required cis ring juction but as a mixture of double bond isomers. From here two different routes were followed to access the two different natural products.

For the strychnine synthesis, the secondary amine was alkylated, and the base used for this step also isomerised the double bond into conjugation with the ester to give a single product. DIBAL was used to reduce the methyl ester to the allyl alcohol, and also cleaved off the acetate protecting group on the fragment introduced in the previous alkylation. An intramolecular Heck reaction then formed the fifth and sixth rings of the natural product simultaneously, and cleavage of the PMB group, with thiophenol as a cation scavenger, gave the famous Wieland–Gumlich aldehyde. As well as protecting the indole nitrogen, the PMB was also essential for directing the b-hydride elimination step of the Heck reaction away from the indoline methine and towards the towards the CH2OH of the allylic alcohol, to give the enol that was able to cyclise to the hemiacetal.

Conversion of the Wieland–Gumlich aldehyde to (-)-strychnine using known conditions proceeded smoothly and thus the natural product was completed in just 12 steps (6.4% yield) from commercial materials. This sets a new record for the shortest enantioselective synthesis of this historic natural product, which I think is likely to stand for a while![5]

Alternatively, to access (-)-akuammicine, the same mixture of double bond isomers was first subjected to PMB deprotection using the same conditions as in the (-)-strychnine total synthesis, which also caused isomerisation to the a,b-unsaturated ester. A similar alkylation-Heck sequence as before completed the first reported synthesis of the natural product, in just 10 steps (10% overall yield).

Next, the group targeted the kopsia alkaloids using a very similar starting material (differing only in that it bears a benzyl protecting in place of the PMB previously used). The route began with selective removal of the Boc group, followed by conjugate addition of the newly liberated amine to triphenylvinylphosphonium bromide. The resulting anion was quenched with methanol, the excess evaporated, and THF and KOt-Bu added to effect the intramolecular Wittig reaction to install the fifth ring. The same phosgene - methanol combo installed the ester as before and then the double bond just formed by the Wittig reaction in the preceding step was reduced out using H2 and Pd/C with excellent selectivity. The final two carbon atoms in the natural product were installed by a Diels-Alder reaction with phenyl vinyl sulfone in refluxing benzene. Treatment of the Diels-Alder adduct with Raney nickel then removed the phenyl sulfone, reduced out the final olefin and, surprisingly, removed the N-benzyl group to give (-)-kopsinine.

Hydrolysis of the methyl ester under acidic conditions gave kopsinic acid which was duly converted to kopsanone via a biomimetic thermocyclisation (well, 200ºC probably isn't so biomimetic). Thus, the group managed to make (-)-kopsinine and (-)-kopsanone in 9 (14% yield) and 11 steps (10% yield) respectively. Which is even better when you learn that the previous best route to (-)-kopsinine was 19 steps and that no synthesis of (-)-kopsanone has ever been reported.

I won't cover the group's syntheses of vincadifformine or aspidospermine here; not because dull (they're only 7 and 5 steps respectively from the product of the organocatalytic cascade, and there's plenty of good chemistry), but this post is already quite long. Anyone who wants to know more should check out the full paper, which is full of more excellent chemistry.

1. Tomáš Hudlický in The Way of Synthesis, 2007, 192-195; parentheses added by me. It's interesting to think about how expectations for a synthetic route have changed over the last 50 years. In the 1950 and 60's you were doing well to make a single stereoisomer, as a racemate, maybe with a late stage resolution. From the 1980's people started to make single enantiomers. Later still came the idea of enantiodivergence, the aim of making both enantiomers of a target (free essay on this here), and enantiodivergent catalysis, where two different catalysts could each produce one enantiomer of a product from a single starting material (as we see here).

2. The diazonamide A synthesis can be found in Chem. Sci., 2011, 2, 308-311. A worthwhile read. If I had a blog back then, I'd've been all over it. Tot. Syn. blogged the group's synthesis of the popular minfiensine back in 2009.

3. As an aside, does anyone know what this stuff is like? You can't buy it from Aldrich, so I guess they made it. I've used and made 3-butyn-2-one (as it's overpriced), which is fairly unstable and not very nice. For this I reacted bis(TMS) acetylene, acetyl chloride and aluminium trichloride, to give the corresponding TMS compound which could then be carefully desilylated. One of the few preps from a patent I've successfully reproduced. You can buy 3-(trimethylsilyl)-2-propynal, which I suppose this is made in a different way, and I don't know if I'd want to try and desilylate it.

4. I say 'unfortunately' as Rhodium's pretty expensive, and thus the 3.8g they reportedly used to decarbonylate 1.8g of starting material would have cost around $300 dollars from Sigma-Aldrich. Ouch. I don't know much about this reaction, but I know that for the equivalent reaction on acid chlorides it is possible to use catalytic rhodium, providing you heat the crap out of it, as it's pretty hard to get the Rhodium to part with the CO to regenerate the starting catalyst. Having said that, they already do this reaction at 120 ºC so I'm not sure how much hotter you'd want to go...

5. Previous shortest asymmetric synthesis was by Bonjoch, Bosch and coworkers in 15 steps (I notice the MacMillan paper says 16), with the source of chirality being a (S)-1-phenylethylamine auxiliary. Check it out in Angew. Chem. Int. Ed., 1999, 38, 395-397. Don't forget that the recent Vanderwal synthesis (Tot. Syn. here) was only racemic. And the yield of the last step in that work was about the same as the overall yield of MacMillan's entire 12 step sequence.

Comments (9) Trackbacks (0)
  1. There’s an extra CO2Me in kopsanone. (I may be about to temporarily lose internet – I’ll be back.)

    • Ah yes, so there is.

      • Now that I’ve had a chance to read the full entry (I’d already read the paper) here’s some more:
        In the 2nd figure the yield of the 2nd & 3rd reactions were 93 & 66% for Bn (& the 3rd for PMB included a little of the Z isomer). The propynal addition (both) included PhMe. Kopsanone was 2 steps from kopsinine (11 vs. 9 steps as you later state).
        In the 3rd figure it looks like MeSe- rather than MeSeH would be eliminated & a H+ would be lost when the amine ring closes and added when the ring is opened.
        In the 4th figure the PhMe was added at the start of the 2nd reaction & the DCM at the start of the 3rd, which had a yield of 81% (61% for the 1st 3 together).
        For the 5th figure that’s triphenylvinylphosphonium bromide in the text and DCM was added with the KOt-Bu & THF; the PhMe went in with the COCl2. The kopsinine structure doesn’t show the stereo of the CO2Me and the paper says it was treated with 1N HCl.
        As for the comment in the introduction, what natural (or unnatural) products can’t be made with current technology?

        • “As for the comment in the introduction, what natural (or unnatural) products can’t be made with current technology?”

          tetra-t-butylethylene 🙂

          • I would try to make that C=C of tetra-tBu ethylene by interrupted Wittig, from beta hydroxy phosphineoxide tBu2C(OH)-CtBu2P(O)Me2 and NaH, the 4-membered ring in phosphor-oxethane should keep the t-Bu groups reasonably far apart in the transition state

        • Thanks for the yields in the first scheme; I went through a lot of different versions as I couldn’t get a layout I liked, and in the end they just got missed off for R = Bn, as did the solvent for the organocatalytic steps. Thanks also for correcting the yields and steps. I’m not sure how I managed to miss out the kopsinine stereocentre in both places it appears.

          In the catalytic cycle, I still think the ultimate product of that reaction step will be MeSeH, as after the product has rearranged in the second cycle there’s a free proton. One could argue that it exits the cycle at that point as the anion, but I think I’ll leave it be.

          As for the phosgene reactions and the DIBAL reduction, I’m just following my standard pattern of putting reagents first and solvent last. It’s pretty hard to convey what’s going on for the fairly complicated Wittig step in the scheme – I hope that taken with the text it all makes sense. Does the fact the reaction is in 5:1 THF:CH2Cl2 rather than THF matter to most people who read this?
          The SIs are free and people who want to know all the details should always check them as I will miss stuff off sometimes. Surely the times and temperatures I’m not giving are more of a big deal?

          As for the 1N HCl, for monoprotic acids I’ll always write M in place of N as I much prefer the descriptive term ‘molarity’ to the odd sounding ‘normality’. That may be a British thing, or maybe it’s just me or the teachers I had. I hope that people still understand.

          Finally, regarding the intro. I didn’t mean to imply that I disagreed with MacMillan – I’m not aware of any natural products that I’d consider unmakable with current technology and enough time and effort. As for other unmakable compounds, I don’t really know enough chemistry to say. I’d imagine there are a few compounds strained or reactive enough to pose a challenge on those grounds, but I’m not sure that’s really a flaw with us as synthetic chemists.

  2. so you would start from commerciall hexamethylacetone+ (t-Bu)2CHP(O)(OMe)2. I was thinking that hexamethylacetone should be a very poor electrophile due to sterics but a quick search shows that a lot of nucleophiles can easily attack that carbonyl…

    • Idle speculation by a total amateur: What would happen if MeLi were added to dimethyldi-t-butylallene and then reacted with tBuX (X=OAc, OTf, whatever)?

  3. It doesn’t seem to be answered, so for the propynal thing, a co-worker needed it in an ongoing total synthesis.
    After a bit of discussion, we agreed on the method: as the product itself is fairly unstable and won’t last long, the technique was to make it from propargyl alcohol with an easy-to-work-up oxidation method. It turned out that stirring the alcohol in the presence of PDC (in DCM if I remember well), then filtration on a pad of silica when TLC indicates completion, followed by cold (RT) concentration provides good enough material to be used straight after.

Leave a comment

No trackbacks yet.