Monday, July 30, 2018

What if Diels, Alder, and Suzuki had a Love Child?

Today our most recent work combining decarboxylative cross–couplings with cycloadditions to generate complex C(sp3) rich compounds was published in Nature. With over 90 compounds and 1,000 pages of supporting information, this project was a massive undertaking and you might be wondering how it all started. 

Despite the extensive amount of pericyclic and radical reactions described, this work is actually of “poll”ar origin. More specifically, it was the first twitterionic species from our lab that was born out of a twitter poll.  A little over a year ago, Phil posed a question to the group during group meeting: Which reaction do you think is the most studied, yet LEAST used reaction in industry? Not satisfied with such a small sample size, Phil did what he does best (second only to chemistry), he took to Twitter. As you can see, Diels-Alder appears to be DA answer. 
Evidently, if we could find a way to repurpose the Diels-Alder reaction to make it practical for medicinal chemists, we could have a valuable new methodology on our hands. First things first though, we had to figure out why few in industry have used the Diels-Alder reaction relative to the incredible amount of time we spend in graduate school learning about all its vagaries. To do this, we thought it would be helpful to compare the Diels-Alder reaction with one of the most widely used transformations medicinal chemists use: cross-coupling reactions. To begin, the difference between the two transforms was staggering. In spite of its rich history, venerable pericyclic transformations have occupied a somewhat “peri”pheral role in industry, seeing only a handful of applications every decade. In contrast, cross–coupling reactions such as the Suzuki reaction are second only to amide bond formations in terms of popularity (see Figures taken from ACIE and J. Med. Chem.)
To us the differences between these two transforms was apparent. Not only are cross–coupling reactions some of the most reliable transformations in organic chemistry, but they are also extremely useful in terms of modularity as countless products can be made from the same starting materials. For a medicinal chemist, this is ideal as numerous analogs can rapidly be prepared from simple building blocks. However, this is not the case with the Diels-Alder reaction. Due to both steric and electronic requirements, when making even moderately complex products only specialized building blocks can be used in these reactions to ensure the reaction takes place cleanly. As is often the case, when the dienophile is prepared, the diene is already “dying” in the sun (or in the cold room). 
Coming from the Baran lab where we have a passion for natural product total synthesis, we know the value in cycloaddition reactions as they allow us to rapidly generate complexity in a single step. So, we thought wouldn’t it be great if we could just combine the two transformations? And that’s exactly what we did.

To make a long story short, we found that the best solution would be a somewhat non-obvious one. We envisioned a sequence where we first execute a facile, well known Diels-Alder reactions with maleic anhydride, a commerical building block that is cheap, bench stable and most importantly well matched to undergo many cycloadditions. Even better, by combining these cycloadditions with established desymmetrization reactions with cinchona alkaloids, in just two steps we quickly accessed numerous building blocks that were poised to undergo decarboxylative cross–coupling reactions that our lab has developed over the past few years. 
Though we thought this was a good start, we figured why simply stop with just the Diels-Alder reaction. Why not apply this same strategy to all known cycloadditions with maleic anhydride? After much work from our talented team, we found that [4+2], [3+2], [2+2], and [2+1] cycloadditions all worked remarkably well. Even better, due to the radical nature of these cross–couplings, in almost all cases we observed excellent diastereoselectivity influenced by the neighboring substituents. On top of that, after the absolute configuration is set in the initial desymmetrization, there is almost no erosion of the ee (even after two cross–couplings). This means that overall our strategy is both diastereoselective and enantioselective. In the end, we were able to use this approach to make over 80 compounds arising from simple cycloadditions as well as 6 applications in the total synthesis of natural products and drug molecules. 

Interestingly, this approach is not simply limited to radical cross-coupling disconnections. After our initial submission, we realized that when we combined our approach with a classical chemistry such as the Curtius rearrangement, we could access even more building blocks that up until this point were previously inaccessible using traditional cycloaddition chemistry. To demonstrate this, in a little under a week Tie-Gen synthesized chiral compound 118, an inhibitor of the EED protein disclosed by AbbVie that previously required chiral separation. 
Now as you might imagine, seeing as how we have over 90 final compounds in our SI -- it is a bit long. In fact, we definitely have set a lab record on this one. Some of you, including one reviewer, may question the utility of such a long SI. It is not done to evoke a feeling of "shock and awe" as the reviewer suggested (no good deed goes unpunished!). And its not because we like playing the game of CSI (see SI) throughout the manuscript. Rather, we feel that the more data we report and the more transparent we are, the easier it might be for others to reproduce our results. That is why we always try and include graphical supporting information and a Q&A section when we can. Furthermore, between all the many sections, characterization, and spectra, we were worried when we first set out to compile the SI just how to organize it. We felt the best way was to include a table of contents (and a detailed table of contents for the table of contents), descriptive overall scope for each panel in the manuscript, and individual schemes for each compound to minimize any confusion that could possibly arise surrounding the overall synthesis of any one compound.  

The great thing about a long SI for a paper is that if you couldn't care less about our lab's work (highly likely), you don't need to read it. Same goes for this Blog actually.  We actually can't come up with one good reason why an SI that includes absolutely everything you currently know about a research project (in an organized way) is a bad thing.  There is a non-zero chance, after all, that someone, somewhere  one day many decades from now might actually use this chemistry and find value in having all the details present. Please enlighten us if you can think of a reason.
Last but not least, we find ourselves in a bit of a charitable mood here in the Baran lab this summer. 
As we mention in our paper, we synthesized (±)-epibatidine•2HCl on gram scale in just 5 steps. So just let us know if you (reputable scientist with an authentic email address) are interested and we would be happy to share some to you (send email to tiegen@scripps.edu).
We want to thank our collaborators at Leo Pharma and Eisai Pharmaceuticals for the work that they put into this paper. Let us know if you have any questions or comments regarding the work! Thanks for reading!

Lisa and the Cycloaddition Team

22 comments:

  1. I was wondering if you guys tried to use as substrate for the coupling the adduct formed between furan and maleic anhydride (A3 in the text) but with the double bond. Tks

    ReplyDelete
    Replies
    1. We haven’t tested the cross-couplings using the adduct formed between furan and maleic anhydride (A3 with double bond). However, we observed the rearranged products on norbornene scaffolds (A4 and A5 with double bonds). Thanks!

      Delete
  2. How did you make D1 and D2? (I know that are commercially available, but just curious). Did you try to make other kind of cyclopropane derivatives? I'm not familiar with cyclopropanation methods that work on maleic anhydride, it'd be a good addition.

    Great work anyway, probably one of my favorite papers ever :)

    ReplyDelete
    Replies
    1. We never prepared D1 and D2 ourselves and we bought them from Combi-blocks. We haven't tried the other cyclopropane derivatives. We believe that you could access them through cyclopropanation with fumerate/maleate and reformation of anhydrides.

      Delete
  3. This comment has been removed by the author.

    ReplyDelete
  4. I presume substrate C2 is made from a [2+2] photo-cycloaddition? How was the diastereoselectivity in that reaction?

    ReplyDelete
    Replies
    1. Yes it is made through a [2+2] between maleic anhydride and 3-Sulfolene (the procedure was taken from Zh. Orp. Khim., 377-382 (1972)). In terms of the diastereoselectivity, we only observed a single diastereomer which we confirmed by XRay crystallography

      Delete
  5. I have always been an advocate of detailed SI. What you guys think about including pre-work up TLC and crude NMR and exclude IR and melting point in SI?

    ReplyDelete
    Replies
    1. Thanks for the suggestion, we are always trying to improve our SI and add as many details as we can.

      -Lisa

      Delete
  6. With a published LD50 of 0.4 ug/kg in mice, epibatidine production (and distribution) on gram scale strikes me as a bold move.

    ReplyDelete
    Replies
    1. Thanks for your concern. Aldrich currently sells this chemical and other gram scale syntheses have been reported in the past (see SI of our paper for a summary of other routes). Anyone that wants it, assuming they are reputable scientists requesting it, will sign a disclaimer that will remind them of toxicity.

      Delete
  7. There’s a typo on page 489.

    ReplyDelete
  8. How did this get into Nature? I guess these days one can staple together bits of previously published chemistry (Nature, 2017, 545, 213; Angew. Chem. Int. Ed. 2017, 56, asap; etc.) over multiple steps with non-obvious step count. It is sad that we've reached a point in science that selling the work is more important than the actual innovation. Also, if the corresponding author was not a famous name the perceived impact of this work would be nowhere near as high.

    ReplyDelete
    Replies
    1. LOL appreciate your candid comments! All we can say is that hopefully the work will be useful to someone. If its any consolation, publishing in journals is not a zero-sum game so our lab didn't take someone's else's spot here :)

      Delete
    2. LOL? What a disrespectful answer. This commenter made some valid points about lack of novelty and disingenuous representation of step count and all that is essentially said is 'yeah LOL haha we got a Nature paper out of it. might be useful LOL'

      There is zero novel chemistry here, with the Baran group getting 10+ papers using essentially identical chemistry with different nucleophiles. All this new work is is stapling this same old chemistry with Diels-Alder cyclisations, leading to (potentially) useful products in 6 steps. Many of these steps are hidden in the publication. If you gave yields for the overall six step sequence, rather than just the two cross-coupling steps, the overall routes would be pretty low yielding. Of course most science works in small, iterative steps forward, but most researchers don’t have the gall to send their work into the top few journals every time.

      Whilst it is mostly true that journal publishing is not a zero-sum game, I am not sure this applies to the very top tier of journals (science/nature).

      Papers such as this could have a far more pernicious effect on the community. PSB is a high profile researcher that many people are in awe over (hence the overwhelmingly positive comments here) and generally puts out very impressive, useful chemistry. I would cite him as one of the most inspiring chemists today. However, when people see publications such as this in Nature, i.e. low novelty, dubious representation of results, high on sales patter, they may come to believe that they too can get their work published in Nature on similar merit. Of course this wont work out, because they are not PSB. Not only could this tarnish the reputation of the Baran group, but also the reputation of the journals themselves.

      Sorry for the heretical comments.

      tl;dr: no novel chemistry, maybe useful products in low yield – really Nature worthy?

      Delete
    3. Well no disrespect was meant but we did legitimately LOL when we read those comments. It's easy to throw stones like this when you are anonymous and we encourage it. These negative comments could easily be wiped away but we would never do that. But instead of LOLing we will address your comments which have an angry tone which is completely unnecessary:

      The step counts are not disingenuous - In fact prior versions of the MS were far more detailed but due to the sheer length requirements and referee feedback we had to minimize the presentation in the actual text and place it in the SI. Everything is there in the SI to see. It is clearly laid out in the table footnotes which steps the yield applies to.

      Regarding steps you are completely missing the point (with all due respect whomever you are). Medicinal/agriculture/discovery chemists seek MODULAR routes to compounds like this. Diels-Alder is not a modular reaction in the same sense as canonical cross coupling. The only thing that matters is the number of steps after the scaffold is made. Imagine one of these compounds was a lead series (such as the examples in the final Figure)- a medicinal chemist would have a CRO make 10 grams of an enantiopure acid/ester and then cross coupling/hydrolysis/cross coupling and one could easily make an enormous number of analogs (in enantiopure form) in a fraction of the time that a de novo DA approach would require. On top of that, most of the substituents installed cannot even be accessed using DA chemistry. The inclusion of the final Figure was precisely included to anticipate such criticism. If you want to see how this chemistry can save time you dont have to guess or be intellectually lazy - just read.

      The paper received 3 very positive reviews, we tried our best to demonstrate the utility with an absurd number of substrates and clear applications where step counts could be radically reduced (pun intended).

      Your bias seems to be on reaction novelty rather than strategic novelty. I suppose you and your ilk also despise total synthesis where strategic advances can be as important as methodological ones. In the present case, there is value and something useful for the community to take cycloadditions, which have historically NOT been viewed as modular strategies for generating SAR, and turning them into systems that are amenable to library modification.

      Your assertion that our paper could do harm to the community or that we are somehow only able to get a paper accepted because of PSB is not only outrageous (we get papers rejected routinely) but also disrespectful to the students that busted their ass on this project and the reviewers/editors/industrial collaborators that clearly disagree with you.

      Delete
    4. Hey Anon, Don't hate the player, hate the game...

      Delete
  9. Great work, congratulations. I was a bit surprised by the stereoselectivity of the reaction. The desymmetrization product is cis and the final RCC products are always trans. Is there an inversion of the configuration in the first RCC? This wasn´t quite an obvious result to me, because from previous RCC papers from the group in most substrates the carboxylic acid was attached to a non-stereogenic carbon or the reaction wasn´t very stereoselective (as in the decarboxylative borylation paper). Is there a thermodynamic preference for the trans product in the vicinal di-carboxylic acid derivatives?

    ReplyDelete
    Replies
    1. Thanks for your inquiry! Indeed, the RCC reaction generates a radical and the product is the thermodynamic outcome that avoids the steric crowding of the neighboring ester to deliver a trans product.

      Delete
  10. I think the utility of the synthetic sequence is really nice, in that you can build up complexity fast, but that is a beast of a sequence to get to your SMs (cryogenic 60h rxn time, CCl4 as co-solvent, stoich cinchona). Any efforts to explore catalytic/greener/faster methods to generate the desymmetrized anhydrides?

    ReplyDelete
    Replies
    1. Thanks for the comments. We have mainly followed Li Deng and Carsten Bolm’s protocol due to simplicity. Cinchona alkaloids also can be used in 5mol% loading, full account see: https://pubs.acs.org/doi/pdf/10.1021/ar030048s
      Here are other reported procedures using room temperature, less reaction time with more active catalysts. For one example from Song’s group, see https://onlinelibrary.wiley.com/doi/epdf/10.1002/anie.200801636

      Delete