Some of you might recall me writing about the tremendous therapeutic success of anticancer agents whose structures were inspired by epoxyketone natural products. The reason this is a compelling package from a synthetic chemist’s point of view is that the mechanism of action appears to involve a neat reaction between proteosome’s N-terminal threonine and the epoxyketone warhead, resulting in a 6-membered ring formation. While this seemed to be a reasonable mechanism of action, which was captured crystallographically, there is a new interpretation and, while the difference might appear minor to some of you, I think it is quite substantial.
Oprozomib is an orally bioavailable epoxyketone derivative currently in clinical trials for the treatment of solid tumors. The high resolution electron density maps obtained by Bourenkov and Chari reveal an electron density for the inhibitor-threonine conjugate that is larger than morpholine. It appears that a seven-membered 1,4-oxazepane ring structure is formed instead.
http://science.sciencemag.org/content/353/6299/594
It is interesting that the Sn2-type 7-endo-tet reaction appears to be the preferred pathway. I am curious to know if this finding has any bearing on Onyx’s patent claims. The composition of matter that is being protected by a patent is sometimes linked to the mode of action. As a corollary to that, what is to prevent someone from designing follow-up molecules with higher propensity to form 7-membered rings? There is probably room to innovate here.
amphoteros 
I would not classify this in the framework of Baldwin rules – the ketal formation is obviously reversible, and the regiochemistry of opening this kind of keto-expoxide with amine is exactly what you would expect for acyclic product. It is quite likely the ketal 7-membered ring closes after the epoxide opening. The reason for threonine being attacked selectively – maybe it is just positioned in the right place next to the epoxy and the NH2 and OH of threonine initially hydrogen bond with the keto-epoxide.
Well, yes, but I was not really making a comment about general feasibity of the reaction. In fact, 7-endo-tet processes are well known for epoxide systems and are not hugely disfavoured on the basis of “Baldwin suggestions”. And I agree about reversibility, but what is more concerning is that the regiochemistry was previously assigned on the basis of available density and – how about this: what compelled the original authors to say that it was a 6- membered ring? That bias must have been based on default assumptions about what to expect in this sort of opening… But the density was misleading, as we now know.
I have seen some pretty odd things in solved X-ray co-crystals of inhibitors bound in the active site of a protein – i.e. pyramidal sp2 carbons and other obviously wrong details. It turns out that there is room for interpretation (if not outright fixing things by hand) when the X-ray resolution is not high, or if there are some fuzzy parts due to superposition of several alternate binding modes that differ in minor details, of if there is a sidechain in rotational flux. The interpretation of X-ray is best done in informal discussions between X-ray spectroscopists and synthetic chemists, but the arrangement in the industry could be more compartmentalized (I used to work in a biotech where the separation was quite strict, and as a bench-level synthetic chemist I could not just walk over to another department and chat with the X-ray specialist about my project without him first checking with his boss, and my boss, to get a go-ahead).
It is too bad that things can be compartmentalized to such a degree in industry…
A very interesting post (as usual!). To add to the discussion if this reaction should or should not be covered by the Baldwin rules, I suggest that we should not use the “endo-tet” classification for the cyclizations that involve a nucleophilic epoxide ring opening. I argued why this is wrong in a recent Chem. Comm. Viewpoint (Finding the right path: Baldwin “Rules for Ring Closure” and stereoelectronic control of cyclizations. Chem. Commun., 2013, 49, 11246. http://pubs.rsc.org/en/content/articlehtml/2013/cc/c3cc43872d) and reiterated it in a more recent review (The Baldwin Rules: Revised and Extended. WIREs: Comput. Mol. Sci. 2016, 6, 487–514. http://wires.wiley.com/WileyCDA/WiresArticle/wisId-WCMS1261.html).
In short, the use of endo-tet and exo-tet for epoxide closures is a misconception as the breaking epoxide C O bond is located outside the newly formed ring in both cases. Each cyclization should be considered an exo-tet process. If both epoxide carbons are considered together as a part of the same functional group (as we do for an alkene), such cyclizations start to resemble exo-trig and endo-trig cyclizations. Tim Jamison made this point even earlier and suggested to refer to such reactions as proceeding via ‘spiro’ or ‘fused’ transition states (Vilotijevic I, Jamison TF. Synthesis of marine polycyclic polyethers via endo-selective epoxide-opening cascades. Mar Drugs 2010, 8:763–809).
The take-home message is that the revised pathway is not stereoelectronically disfavored. Enzymes do obey stereoelectronic guidelines embodied by the extended Baldwin rules.
Thanks. This is very useful, Igor! I think your paper should be taught in advanced synthesis classes. Coming back to my point – I was not really trying to argue with what is allowed or not (although your point about the classification is well taken and I will use this in my classes). I was saying that the product was incorrectly assigned on the basis of data from the largely inadequate (low resolution) density that was observed. Why were the authors swayed into thinking that it was a 6-membered ring? It is an intriguing question…
I wish I could answer that question, Andrei! The answer may lie in the depth of human psychology and the fact that the main thing that 90% of students retain from their organic class is the ability to draw a perfect hexagon.
Yes, this is a distinct possibility!