Saturday, April 25, 2020

(–)-Maximiscin


Greetings fellow social isolationists. If you’re here, it may be because you spotted our recent synthesis of (–)-maximiscin in JACS, thanks for stopping by! If you’re looking for a harrowing tale of anguish, suffering and overcoming insurmountable odds, don’t worry the taxol blog post is coming soon. All jokes aside, this molecule was no pushover and it took some creative solutions to bring it across the finish line. Hopefully I can convince you of this, while providing a brief distraction from whatever you might be bingeing on Netflix.

Are you not entertained?

Before getting into how we made maximiscin, it’s worth considering a little about why we targeted it in the first place. In addition to having a cool name, it has a relatively unique structure, originating from the union of three separate biosynthetic pathways. The molecule actually exists as a mixture of interconverting atropisomers, resulting from the hindered sp2-sp3 bond connecting the pyridone and cyclohexyl fragments (C3–7). Back in 2017, when this project began, the lab was very into decarboxylative cross-coupling. As a result, we wondered if (surprise, surprise) a decarboxylative cross coupling could be applied to construct the hindered C3–7 bond of maximiscin. Not only would this enable a simplifying disconnection, but it would also allow us to push the limits of sp2-sp3 cross coupling, which remains challenging for hindered systems. Additionally, an alkyl carboxylic acid coupling partner would be more stable than its halide counterpart, and would be “spring loaded” to undergo radical oxidative addition, via extrusion of CO2


After devising a route to our carboxylic acid coupling partner, we tried a lot of different conditions to effect the cross-coupling. Nothing worked, until we used a set of ligand-free Ni conditions initially reported by the Molander group. With Ni(dpm)2, our heteroaryl zinc reagent and Zn powder (organozinc too hindered to homocouple and access Ni0) we were miraculously able to form the cross-coupled product in 9% yield. We tried many different things to optimize this reaction, (see SI) but could never dig ourselves out of that 9% hole, and eventually abandoned the cross-coupling approach. Despite this disappointment, I still think it’s remarkable that we were able to form the bond at all. The product is doubly alpha-substituted, doubly ortho-substituted and exists as a mixture of atropisomers. It represents, to my knowledge one of the most hindered sp2-sp3 bonds built using cross coupling. Huge shout-out to Amin Minakar, an amazing visiting student who slogged through the better part of this cross coupling saga with me.


Realizing that our previous disconnection wasn’t going to work out, we re-tooled our retrosynthetic strategy. What we came up with was a rather non-intuitive disconnection, opting to cut the central pyridone right down the middle. Perhaps naively, we envisaged that the ring could be forged at a late stage from an activated diacid, by “clicking in” an appropriately designed nucleophile. In addition to making the synthesis completely convergent, we anticipated this might ease construction of that constrained atropisomeric ring system by pre-establishing the hindered C3-C7 bond. We also liked that this would allow us to avoid a late stage pyridone N-oxidation, which I can tell you from experience is a terrible reaction. For the sake of brevity, I’ll touch on a few interesting steps we developed along the way.

Revised retrosynthesis

One of the key steps of this synthesis is a nifty desymmetrizing C-H activation, which defines 4 stereocenters in a single step from a meso-carboxylic acid precursor.  It took us a while to find the right directing group/set of conditions, but we were eventually able to form product in 58% yield with outstanding selectivity. I’d like to thank Feng-yuan Wang, an incredibly talented vising undergrad for all his help preparing the various functionalized directing groups we screened for this reaction. Unfortunately, it wasn’t all smooth sailing, and I learned an important lesson during this process: never trust an old sealed tube, especially on scale. The consequences of such a decision can be seen here… extracting that oil bath was not fun. An important feature of this chemistry was that we could actually remove the directing group after the reaction (a frequent limitation of C-H activation methods), and could even recycle it to make more C-H activation substrate!

Desymmetrizing C-H activation

Sad day in the lab

With desymmetrized material in hand, we turned to a (wait for it…) decarboxylative homologation to build out the diacid fragment. Our initial tactic used conditions previously developed by the lab. The reaction worked beautifully, but our pesky substrate was perfectly setup to do a 1,5-HAT, generating a more stable alpha-oxy radical. This led to a competing double addition pathway via trapping of a second equivalent of phenyl vinyl sulfone. We realized that if we could oxidize that intermediate alpha-oxy radical, it would provide access to a later aldehyde intermediate directly; what we needed was a reductant and oxidant in the same pot. We initially had success with photoredox (PET) chemistry. Blue LEDs aren’t a common sight in the Baran lab, but the reaction worked so we ran with it. A set of conditions were identified which were reasonably efficient, but had issues with scalability and purification (reactions were messy and aldehyde product silica unstable). Nonetheless, this was an important proof of concept and we continued searching for a more tractable solution. Specifically, we wanted to get away from the activated ester and use the carboxylic acid directly. Oxidative decarboxylation seemed like the obvious solution, and we began investigating the Kolbe electrochemical decarboxylation, motivated by its robust 100+ year history. Unfortunately for us, we could never get much more than 5% yield. 

Other motivating factors were also at play
What ended up saving my (Canadian) bacon were the humble Minisci conditions: Ag+ and persulfate. Although the initial yield was modest (6%), we discovered a pretty cool Fe co-catalyst effect, which led to a remarkable enhancement (71%). This finding was completely serendipitous, we were evaluating other more common radical oxidants (e.g. Cu), and Fe was sort of an afterthought – it almost didn’t get screened. We think it may be acting as a selective radical oxidant; it’s known from the old Fenton literature that Fe will selectively oxidize alpha oxy radicals, but leave other alkyl radicals untouched. The final conditions could be run one-pot, directly from the lactone after in situ hydrolysis, and proved highly efficient on scale (91%). As an added bonus, the crude material was remarkably clean (see crude NMR) and could be telescoped into the next step, which solved our aldehyde purification issues. I think there are two important lessons here: (1) don’t overlook the simple solution – we started out using some very trendy, very powerful methodologies, but ultimately dump and stir chemistry from the 70’s worked the best; (2) don’t talk yourself out of an experiment – it’s not as satisfying, but sometimes blind screening is what gets the job done.


Decarboxylative homologation 
Crude NMR using one-pot Minisci conditions

After a few more steps, it was finally time to roll the dice on our late stage pyridone synthesis. Our first attempts came up snake eyes – when we combined our diacid chloride electrophile with a functionalized oxime-ether nucleophile (X=H), we ended up with some pretty sad looking reaction mixtures. Our diacid electrophile pooped itself, as the reaction color suggested, and we recovered the oxime-ether nearly quantitatively. We had to tweak our initial conditions quite a bit, adding AgOTf to activate our diacid electrophile, and using a silane to spice up the nucleophilicity of our oxime-ether(X=TMS). The latter modification was an interesting development, and led to the creation of an “aza-Sakurai”-type system. An added benefit of this TMS activation effect was that it generated TMSOTf in situ (TMS derivatization observed by LCMS). This led to capping of our pyridone C-4 OH  in non-polar reaction solvents (e.g. PhMe) and prevented product from reacting with activated diacid starting material. Unfortunately, despite giving the best yield, acetonitrile proved too harsh for the TMS group to survive, leading to some derivatization of product with starting material. Fortunately, we found that a simple methanol quench could liberate derivitized product and return the diacid as a mixture of mono- and dimethyl-esters. Our final set of conditions delivered the protected natural product in 41% yield. Luckily for us, the final deprotection went smoothly, to reveal maximiscin and complete our synthetic journey. Special thanks to the isolation chemists Dr. Lin Du and Prof. Robert Cichewicz for sending us authentic material for comparison.



Initial attempts at pyridone synthesis

It’s interesting to consider how this synthesis evolved over time. We started with the goal of developing a hindered cross coupling reaction, but this morphed into a completely different strategy, and resulted in the development of some pretty cool chemistry. There were certainly ups and downs on this project, but overall I had a blast making maximiscin – I even got a cake for my efforts (don’t worry, Sol finally got his cake too).  Thanks for reading!

Let them eat... molecules?
– Kyle M.


Edit: Some people have asked about the mechanism/stereomodel for the C–H activation step. Unfortunately we were never able to isolate a palladacycle to validate our stereomodel. Based on the plastic model, there is obvious steric clash between the t-Bu of the directing group and one of the alpha-methyl substituents. We think this leads the palladium to C–H activate one methyl group selectively via CMD mechanism (J. Am. Chem. Soc. 2008130, 10848) to give the favored Pd(II) cycle. This is oxidized by NaIO4 to Pd(IV) which can undergo reversible X-ligand dissociation to form a cationic complex (J. Am. Chem. Soc. 2014136, 12771). The latter species is trapped by methanol in an SN2 fashion to give product.