Chu, S.; Wallace, S.; Smith, M. D. Angew. Chemie Int. Ed. 2014, 53, Early View (Link)
The natural product was first isolated in 1979 from the skin of a Columbian poison dart frog, and a combination of low toxicity, an interesting set of neurological effects, and obvious low abundance, make it an interesting synthetic target. So Gephyrotoxin itself wasn't isolated recently, and it's attractiveness as a target has resulted in several previous syntheses:
Those marked with a * intersect with Kishi's original synthesis, and if these previous publications interest you, then these slides from a seminar at Harvard go into a fantastic amount of detail, and include all the references to the primary literature. As can be seen though, this new route affords (-)-Gephyrotoxin in 5 steps fewer than anyone else, and in a higher overall yield. In particular, I appreciated the high number of well-known reactions, with the short step count a result of clever planning, rather than by reliance on some particularly arcane chemistry.
Getting into the synthesis then, starting from L-pyroglutaminol, both the oxygen and nitrogen were protected with TBS and Boc groups respectively. We've next got addition of a cyclopentene-containing Grignard reagent, followed by reduction of the N-acyliminium cation. The cis-geometry is presumed to result from hydride delivery to the least-hindered face.
Next, ozonolysis opens the cyclopentene ring and generates a ketone and a terminal aldehyde, the latter of which is then used in a Wittig reaction to introduce an α,β-unsaturated ketone, completing the set-up for the key cascade ring-forming step.
Personally, it took me a little longer than I'd care to admit to figure out quite what's happening there, so I've tried to flesh out a little how I think this is happening:
The paper goes into more detail than I'm going to trying to rationalise the stereochemical outcome of this, but I will touch on the final reduction of the tricyclic imine, which I thought was nicely explained. By this point you've got quite a hindered "front" face of the compound, so the likes of sodium cyanoborohydride were taking the easy route and introducing the hydride from behind, giving the undesired trans ring junction. Instead, sodium triacetoxyborohydride can first reaction with the C2 hydroxymethyl group to give an intramolecular reducing agent, which then reduces the imine to give the cis ring junction, with good stereoselectivity and yield.
With the framework formed remarkably quickly, "all" that remains is to sort out the substituents. Swern oxidation forms an aldehyde from that C2 hydroxylmethyl, which can be used in a second Wittig reaction to add in the missing carbon in that side chain. Then a rather neat reaction from Negishi I've not seen before, which forms an alkyne from the methyl ketone via an enol phosphate. The enol phosphate is formed by kinetic deprotonation of the methyl ketone and trapping with diethylphosphoryl chloride, then a second equivalent of base gives elimination to the alkyne.
Switching focus, treatment with aqueous acid hydrolyses the methyl enol ether to the aldehyde, which is reduced with sodium borohydride to the alcohol. To finish off, we've got a hydrometalation/cross coupling of the alkyne with a protected iodoalkyne, shown below, (I'll leave the details of that step to the paper), then desilylation with potassium carbonate in the same pot completes the synthesis, to give (-)-gephyrotoxin.
So that's a 9 step synthesis in 14 % overall yield, which as mentioned above, manages to generally rely on chemistry that should be familiar to most undergrads, with a lovely cascade step to form the core. Beautiful.