Update 02/09: just realised that Fukuyama's 2009 synthesis of (-)-huperzine was covered over at synthetic nature last year and the compound itself has a wikipedia page. Also, I notice that HMPA is mentioned in the paper. Whoops.
A Robust and Scalable Synthesis Of the Potent Neuroprotective Agent (-)-Huperzine A
There's been a quite a lot of interest in this little natural product already, as it's known to be a potent and selective reversible inhibitor of acetylcholine esterase (AChE), with an impressive Ki of 23 nM. Apparently, recent studies have established that this property makes the compound a possible counter to organophosphate chemical weapons, such as the 'nerve gases' sarin and VX, which work by covalently modifying AChE (I, for one, am so glad I wasn't in that clinical trial). There's also some evidence it may be useful in slowing the progression of neurodegenerative diseases. However, the problem is (as usual) the difficulty of getting useful amounts of the darn thing for further studies - in this case the compound comes from a painfully slow growing chinese herb, with an isolation yield of just 0.011%. If my readership is what I anticipate then I expect you're all thinking, "that sounds like a job for total synthesis!", and you'd be right. The best asymmetric synthesis of (-)-huperzine reported prior to this work was that of Kozikowski and coworkers, published way back in 1991, standing at 16 steps with an overall yield of around 2.8%. This new route by Herzon and coworkers manages a significant improvement on both counts, despite actually using the same chiral building block to introduce asymmetry...
The Herzon synthesis starts from (R)-4-methylcyclohex-2-ene-1-one, a known compound, of which a number of syntheses have been reported. Here, the group opted to use the known preparation from inexpensive (+)-pulegone. Oddly, when referencing this paper, the group claim that the sequence to convert between the two is 4 steps, but as shown below it's clearly 6 steps (4 columns and a distillation, by way of purification). Still, it's fairly practical (at least on a research scale) and 20 grams of (+)-pulegone can be quickly converted into around 10 grams of (R)-4-methylcyclohex-2-ene-1-one without too much trouble. The route began with the diastereoselective 1,2-reduction of the enone using sodium borohydride, followed by ozonolysis of the isopropylidene group to give the α-hydroxyketone. This was then protected as the dioxolane by transketalisation, and the alcohol converted to the corresponding triflate. Elimination by heating in neat DBU, followed by a final deprotection with a little sulfuric acid gave the key building block in 42% overall yield.
Having made the starting material, the group set about its conversion to the natural product. Thus, lithium dimethylphenylsilylcuprate was added 1,4- to the enone, and the resulting enolate was trapped by alkylation with the dibromide shown. Unfortunately, the reaction had to be conducted in presence of a few equivalents of HMPA as a cosolvent, presumably because of trouble with the alkylation step. That notwithstanding, the step works well, giving a single diastereomer by NMR and the group were able to conduct it on up to 4.5 gram batches. The product was then converted to the α-cyanoketone by reaction of its lithium enolate with TsCN and a neat Pd-mediated α-arylation was then performed to close the final ring of the natural product. Highly optimised Wittig conditions were then used to convert the bridging ketone to the required olefin. Interestingly, a strong correlation was found between the concentration of the reaction (bizarrely conducted in ether) and the stereoselectity; at 1M a 1.1:1 mixture of E/Z isomers was obtained but at 0.01M a 5:1 mixture in favour of the desired E isomer was obtained. The group concludes that this result is consistent with a salt effect, and that it suggests that the desired E isomer is the kinetic product. Impressively, these three transformations (α-cyanation, arylation and Wittig olefination) could be telescoped together and therefore only required one column between them to give the product in excellent yield over three steps.
Next the olefination product was treated with triflic acid followed by TBAF and hydrogen peroxide to effect Fleming-Tamao oxidative desilylation. The resulting alcohol was then dehydrated using the Burgess reagent in refluxing toluene, unsurprisingly forming the non-bridgehead alkene. The group then began the final sequence to convert the nitrile group into the amine found in the target compound. This began with hydration to the carboxamide using the rather difficult to draw (but surprisingly easy to make) platinum catalyst shown, a transformation I hadn't seen before. Finally, this was converted via an iodine(III) mediated Hofmann rearrangement to the methyl carbamate, which was cleaved by methanolysis to the amine, completing the natural product. Amazingly, all four transformations were accomplished with only a single purification step (at the end); a great example of an optimised and telescoped sequence in action. Even the sequence final step could be performed on greater than 1 gram scale, and over 3.5 grams of (-)-huperzine had been prepared at the time the paper was written. I'm not quite sure where to count the number of steps (and yield) from, but it's a big improvement on the previous state of the art, and for the first time delivers grams of the target compound for biological testing. Well done guys!
1. An enantioselective synthesis is desirable as (+)-huperzine is almost inactive against AChE, meaning you need twice as much racemic material to exert the same effect (and who knows what else the antipode does?). Kozikowski used Corey's classic (-)-8-phenylmenthol chiral auxilliary to achieve a stereoselective synthesis. Funnily enough, this auxiliary and Herzon's starting material are derived from (+)-pulegone, although the way the chirality is transferred to the rest of the molecule could hardly be more different.
2. HMPA is a bit of a carcinogen. I wonder if they ever tried DMPU, which is almost as good a lot of the time. I was made to distil HMPA as a masters student, and have used it a fair bit since so I'm not too uncomfortable with it. I even know where it comes in a proton NMR, but I still think twice about using it in a reaction. Conversely, I've heard of one organic research group who are banned by their supervisor from handling the stuff at all. We've got a chart of acceptable and unacceptable solvents for industry on the wall of our lab. It's not on there...
3. Just stir 5 parts dimethylphosphine oxide with 1 part tetrakis(triphenylphosphine)platinum(0) in toluene. Precipitate with ether. Filter. Dry. Full prep. and proposed mechanism in Journal of Molecular Catalysis A: Chemical, 2000, 160, 249–261.
4. Looks a bit like the last step in Garg's recent N-methylwelwitindolinone C isonitrile synthesis doesn't it? I don't think this goes via the free nitrene, though. The classic Hofmann is thought not to, anyway.