Tuesday, July 26, 2016

Maoecrystal V, or I how I learned to cope with my mediocrity

   Greetings and salutations ladies and gentlemen! Our total synthesis of maoecrystal V has been published in JACS recently. It is a fascinating molecule, in my opinion, with a beautiful array of densely interlocked rings, myriad of quaternary carbons, and an intriguing bioactivity profile to back it up (or not). It is no wonder, that it received so much attention from synthetic community, especially in the first five years after its publication in Org Lett. But don’t be fooled by that pretty face, this molecule is a demon in disguise! Making it in the lab is like having a crush on that popular girl in high school: no matter what you do or how much you try, the only thing you are going to get is a beat down by her much more handsome and stronger boyfriend. This pretty much sums up my experience working with this molecule.

   On the very first day in the Baran Lab, I decided to go for the baddest molecule in town, perhaps out of shear arrogance, or maybe stupidity, but most likely a combination of both. At that time, at least in my eyes, such molecule was maoecrystal V. This decision took me on a hell of journey, a roller-coaster ride if you will, full of adventure and unexpected outcomes, that changed me as person and a semi-scientist. But that is the beauty of total synthesis, in my opinion, it pushes you to your very limits and that is when you truly realize what you can and can’t do.

   The journey began with making the bicyclic ketone 5a or 5b, which turned out to be far from a trivial task. I am embarrassed to admit that it took me six months to actually make it. I have no excuse for that. The figure below summarizes some of the early failed approaches.

The Journey Begins
   When the route to 5a had finally been developed (described in the paper), the stage was set to explore the key addition/pinacol rearrangement sequence. At this point, Phil probably thought, “Yea... this Cernijenko guy needs all the help in the world and then some”, so I was joined by an awesome post-doc Rune Risgaard. We actually made a really good and productive team because we were quite different and complemented each other really well. United by misery and failure, fueled by unhealthy obsession (both of us really really wanted to make this damn molecule) we became good friends and marched on to the battlefield. We had absolutely no clue what was awaiting us.

   The pinacol rearrangement and hydroxymethylation is pretty well covered in the paper and there is additional information regarding these steps in the Supporting Information. Unfortunately, my therapist has forbidden me to discuss these steps to prevent further damage. Consequently, I will skip straight to the installation of the last carbon of the molecule. The cyanide was perfect for our purposes since it was a nucleophilic source of carbon that had the same oxidation sate as the ester and seemed small enough to react with a hindered C-8 ketone. The Figure below summarizes our attempts to accomplish this task.

A small glimpse of the challenge
   To sum up, the reactivity was not the problem. Most of time time the yields were very good, stereocontrol, on the other hand, was an issue. The cyanide was always coming from the undesired face giving us exclusively undesired diastereoisomer. We tried different CN sources, Lewis and Bronsted acids, solvents, and additives. The outcome was the same in almost every case. Interesting example is the reaction mediated by I2 where CN approached from the desired face but attacked the more hindered bis-neopentyl ketone to give 28. I was very surprised to see the X-ray crystal structure and to this date I still have no clue how I2 completely switched the chemoselectivity of this reaction. Other nucleophiles such as vinyl lithium, furanyl lithium, lithium 1,3-dithianes, corresponding Grignard and organocerium reagents did not solve this problem. We also tried blocking the undesired face by epoxidizing the C-15/16 alkene, but CN still approached from the undesired face. However, a very important result for us was the formation of 29 mediated by Zn(OTf)2 because it is the only example where we achieved the desired stereo- and chemoselectivity. The compound itself was synthetically useless but it made us realize that if we form the THF ring first, CN might approach from desired face. Interestingly, that hypothesis lead to the synthesis that was published today. In total synthesis things can go from terrible to great in just one reaction.

   Lastly, I would like to discuss the very last reaction in synthesis. It is quite strange to eliminate iodide with Oxone. Initially, we tried to engage iodo ketone 18 (formed as ~15:1 mixture of iodo epimers 18a (major) and 18b (minor)) in classic base promoted E2 type elimination. But not even a trace of maoecrystal V was formed despite myriad of conditions examined. We believe that there was simply no antiperiplanar hydrogen available for the desired elimination to take place. While the minor 18b iodo ketone, in principle, should have underwent E2 elimination, we think that the chair conformation with axial iodide is highly disfavored due 1,3-diaxial strain. As a result, 18b' is probably more accurate representation of its conformation, where iodide is pseudoequatorial on a boat.

The Iodide Elimination
   This is where we got quite lucky. After oxidation of the iodohydrin with one of the commercial bottles of DMP, iodoketone was formed in excellent yield along with 2-4% of maoecrystal V as a by-product. The relevant portion of crude NMR is attached below.

MCV hiding in the weeds
   How did it form? Freshly made DMP or other bottles of commercial DMP did not give any traces of maoecrystal V. It was just that one bottle that consistently yielded small amounts of maoecrystal V. What was in it? Long story short, we suspected that maybe it was contaminated with Oxone (which is used to make DMP) since oxidative elimination of iodide with m-CPBA was known in the literature (this reaction was very messy and low yielding in my crooked hands). Indeed, simply adding buffered aq. solution of Oxone to the reaction mixture resulted in very clean elimination of iodide to finally give maoecrystal V. That was a very good day in the lab.

   We were able to make decent amount of this natural product using this route, as a result, we had the chance to explore its bioactivity profile with different collaborators. Unexpectedly, it turned out that synthetic maoecrystal V was not active in any of the cancer cell lines (even HeLa) tested. Our dreams of curing cervical cancer were shattered, but we received a lot of valuable lessons from this synthetic campaign. That was a hell of a ride! R.I.P MCV, I hope we will never meet again.

-Art

16 comments:

  1. Seems like whenever you weren't telescoping, you got x-rays! Fantastic work.

    ReplyDelete
    Replies
    1. Thanks! We actually got X-rays for majority of intermediates that were telescoped. They don't appear in the paper but are included in SI.

      Delete
  2. I think there is an alternative explanation for the formation of product 28 with TMSCN + I2: It is not that CN suddenly attacks the more hindered carbonyl, but rather that initial cyanohydrine formation is inhibited in the presence of iodine (scavenging free CN- as cyanogen iodide). I also think in situ formed TMS-iodide as a powerful silylating agent activates the less hindered carbonyl by carbonyl silylation, which then (in the absence of free CN-) accepts the more hindered ketone carbonyl as a nucleophile. The resulting cyclized oxo carbenium cation should be fairly stable but eventually ends up abstracting CN from TMSCN. The facial selectvity in this case is probably due to thermodynamic control and the strain of the cyclic system (the other two conceivable stereoisomers of 28 would be more strained).

    Also, a very impressive problem-solving and optimization work, and with just two chemists.

    ReplyDelete
    Replies
    1. Hello milkshake, that is pretty interesting observation. I haven't done any mechanistic studies for this reaction but we did try TMSI with TMSCN and it gave the same product (attack of C-8 ketone from undesired face) as ZnI2, TMSOTf and majority of other Lewis Acids. Consequently, I don't think in situ formation of TMSI is the explanation.

      Delete
    2. it could be I am wrong, but I think it would be a combined effect of removing free cyanide and generating TMS-iodide. For this mechanism to be operational, you also need to inhibit initial cyanohydrine formation by scavenging free cyanide. Since TMSCN is a decent silylating agent on its own right and undergoes alcoholysis with a primary alcohol almost instantly, and since you do have one unprotected hydroxymethyl group in the substrate that gets silylated, you generate one extra equivalent of HCN in the reaction mix right at the beginning - an equvalent that needs to be mopped up or evaporated before this alternative cyclization can take place.

      I guess the easy way to test it would be to expose the material to excess of TMSCN without any catalyst, evaporated the O-silylated product to remove HCN and then add fresh TMSCN and TMSI. Or treat it with TMSI alone, to see if you can isolated any cyclized product with a skeleton that is the same as 28 but without the cyano group

      Delete
  3. I was looking at the very first step (prep of 7). It seems like everything is crucial here! Questions: Just how profound is the solvent system, in the experimental you used a 1:1 mixture of toluene and MeTHF. What happens if it is 1:2 or 2:1, etc?
    Also you add the Grignard over 4 hours, what happens to the quality and yield of 7 if you do it faster or slower? Also just how critical is the temperature here, adding such a large volume over 4 hours did you not observe deviation from -78°C? I also noted that this reaction seems to be very dilute, any idea if it works just as well much more concentrated?
    Thanks, congrats of a great piece of work.
    Quintus

    ReplyDelete
  4. Hello Quintus! Thank you for your kind words (big fan of your work on Discodermolide btw).


    1) What happens if it is 1:2 or 2:1, etc
    The ratio of PhMe and MeTHF needs to be at least 1:1. It can be 2:1 or 3:1 without any noticeable drop in yield or ee. However, if MeTHF is major component (1:2) you start observing some 1,2 addition and yields lower by 10–20%. The ee drop slightly as well.

    2) Also you add the Grignard over 4 hours, what happens to the quality and yield of 7 if you do it faster or slower?
    If you do it slower - no problem. I did Grignard reagent addition over 8 hours and got almost identical yield and ee. If you do it faster (over 2 h) you get lower yield (due to 1,2 addition) and lower ee.

    3) Also just how critical is the temperature here, adding such a large volume over 4 hours did you not observe deviation from -78°C?
    The temperature is quite important but it doesn't have to be exactly –78 ºC. The reaction works at –40 ºC, you do get lower yield (60-70%) and lower ee (approx. 90% ee). I tried to optimize it for a more practical temperature range, but I failed. I actually did monitor the internal temperature once (although on a smaller scale ~5 g) and it did increase by about 3-4 ºC. It did not seem detrimental though, since I still got good yield and excellent ee. I believe as as long as temperature is approx. –70 ºC, it will give good yield and ee.

    4) I also noted that this reaction seems to be very dilute, any idea if it works just as well much more concentrated?
    The reaction itself did work just as good when more concentrated. However, sometimes the Grignard reagent would precipitate in the syringe and clog it. As a result, I decided to go with less practical (more dilute) but much more reliable conditions.

    Hope I answered your questions. Let me know if you have more.

    ReplyDelete
  5. Thanks for your answers and your comments about discodermolide, more hair falls out when I even think of that one.
    So the preparation of compound 7 does not seem that problematic or at least nothing that could not be optimised.
    The yield of the pinacol, 45% and the side product 22% still leaves missing mass. Was this isolated or was it just "polymer"?
    Now I realise that you needed material for the various reaction screening and so on, but with surely enough on hand why only make 54mg of 1? Surely a couple of grams would have been possible? A 75% yield is not bad at all for such a complex sequence.
    You got an x-ray, which solvent did you use to obtain the single crystal, or did you use Fujita's methodology?

    ReplyDelete
    Replies
    1. The yield of the pinacol, 45% and the side product 22% still leaves missing mass. Was this isolated or was it just "polymer"?

      The remaining mass balance is most likely non-specific decomposition. These 2 products is the only thing I could isolate.

      Now I realise that you needed material for the various reaction screening and so on, but with surely enough on hand why only make 54mg of 1? Surely a couple of grams would have been possible? A 75% yield is not bad at all for such a complex sequence.

      That was the goal initially but since biological data was so disappointing we just decided there is no point in making grams of something that is completely useless and decided to publish as soon as possible instead. Plus I spent so much time on this project, I just wanted to be done with it once and for all.

      You got an x-ray, which solvent did you use to obtain the single crystal, or did you use Fujita's methodology?

      I obtained it by recrystallizing it from acetone. We did try the Fujita's sponge methodology initially under the supervision of one of the gradstudents from Fujita's lab. Unfortunately, MCV did not have the right size to fit in the cavity of the sponge. While it is certainly a great tool with enormous potential, currently it has a lot of restrictions regarding the size of the molecule.

      Delete
  6. Dear Artiom,

    In how many years did you complete this project?

    ReplyDelete
    Replies
    1. It took 3 years and 9 months for us to finish it. Perhaps way too much time for a small molecule like this but the road to it was treacherous and full of surprises.

      Delete
  7. Very helpful suggestions that help in the optimizing website.
    thank for sharing the link
    gclub online
    goldenslot

    ReplyDelete
  8. Were you able to get to a chiral beta keto trifluoromethyl ketone bearing an alkyne (line 2, scheme 1), or were the conia-ene investigations with racemic material?

    ReplyDelete
  9. This comment has been removed by the author.

    ReplyDelete
  10. These studies were done with racemic material.

    ReplyDelete