Taxol

Taxol! The fearsome jungle-gym-like structure of this molecule has attracted worldwide attention since the late 20th century. Our group was of course one of them, and has been working on this molecule for >10 years, which has culminated in this recent JACS article. PLEASE READ the SI because it describes what actually happened and rationalizes why/how we arrived at the final route; looking only at scheme 1 could give a false impression that everything worked as designed from the outset. It was the longest project run in the lab and I happened to be the last person. Here I shall share some personal stories I  experienced through this arduous oxidation-state-pyramid climbing.

My first TOC draft. Phil said no.

In Dec 2015, a super postdoc Shige (currently an assistant professor at Tohoku University) took over the Taxol project. (Phil said) there are "only" two more oxidations to complete the synthesis.  However, these are very specialized and each are one of the most challenging sp3 C–H bonds to be oxidized. Shige recruited me for the project in Jan 2016. For the record, back in June 2015, when I joined the lab, Phil offered three projects that I could work on and one of them was Taxol. I consulted with senior students in the lab and EVERYONE said "DON'T DO Taxol". In spite of this advice, I was naive enough to commit to the Taxol project. It must have sucked for Shige to work with "1st year me," who barely knew how to stain a TLC place. But he patiently taught me a lot of chemistry, experimental techniques and how to strategically carry out a total synthesis. We went through a lot of failures (see the 1st generation route in SI) but finally, it was found that C–2 β-OH was critical for C–1 oxidation, which led us to the 2nd generation route and established access to a handful of C–1 oxidation product.

Around this rough time, we got a letter from another, more junior, "Baran", which made daddy smile and me pale.

We were happy to "complete" one of the two most difficult oxidations. However, the happiness in total synthesis always has a short half-life and we immediately hit a wall called C–2 reduction. The challenge was two-fold: First, this ketone was a brick. The most difficult ketone reduction ever, it sometimes even survived super harsh conditions such as Birch and LiDBB. Second, it always gave us undesired C–2 β-OH in diastereomerically pure form. These were the worst sorts of optimizations because the results did not tell any directions to pursue, we were basically shooting in the dark. It was particularly painful since most substrates took about 20 steps to prepare, which all ended up failing. 

Never saw the physical state of compounds until we started collecting SI, since we never had enough material to see it.

This deceptively benign-looking reaction took us 2 years (the 2nd and 3rd generation routes) to figure out. Long story short, things finally got better when we found the C–4 α-OH β-Me motif, which gave us α-OH upon C–2 reduction on a model substrate. We integrated this moiety and carried out a new oxidase phase. Fortunately, both C–1 oxidation and C–2 reduction worked smoothly. 

For the C–7 oxidation, with the C–5/6 epoxide installed, the oxidation relay idea immediately emerged. C–6 iodoso elimination was more or less the first examined idea and worked beautifully well. The final key C–6/7 epoxide opening also went smoothly with Nugent-RajanBabu reagent in the presence of so many sensitive functional groups (carbonate, tert-OH, OTMS). This completed the two major objectives to achieve the total synthesis of Taxol. And indeed, this is actually our final route. The 5th generation route was already laid out 2 years ago from now.

C–7 oxidation relay.

The day I was confirming the stereo- and regioselectivity of this reaction, Phil was obviously excited. He came to my office asking how the 2D NMR looked every 40 min, even after I told him my NMR wouldn't even start in 4 h. In sum, with all key structures installed, it took only <100 experiments (including scale up) to pave the route to this key intermediate. 

The photo evidence Phil chasing me around.

After the completion of two key oxidations, our substrate and Holton's intermediate looked so similar that we stopped pursuing this route, assuming the downstream reactions should work without problems. Phil Baran is never satisfied when it comes to a synthesis and we started looking into more ambitious strategies (both the 4th generation route and also attempts that were so random that they are not even included in the SI or my thesis). They ended up not being incorporated into the final route but provided some fascinating insights (see SI).

We never really talked about the cyclase-phase in this blog post but we needed to go through the sequence so many times to provide a sufficient quantity of material to do any oxidase-phase chemistry. The cyclase-phase works amazingly well, but our lab is not designed for kg scale reactions and it was just physically tough. I had to run through the entire building and collect all the 5 L flasks from other labs.

#scaleupsucks

It was around this time that Chemveda offered us a collaboration opportunity, and prepared >100 g of the cyclase-phase end product. I have to say the cyclase-phase is not the most straightforward sequence, especially on scale. But they bravely undertook it, overcame issues associated with the scale up, and significantly improved the overall sequence. If there's anyone in Pharma looking for a great company to collaborate with, Chemveda is amazing.

The cyclase-phase end product prepared by Chemveda.

In the following year, we were joined by a talented postdoc Hugh, to finally complete the synthesis. As previously mentioned, our intermediate was almost identical to one of Holton's intermediates, so we optimistically assumed the only thing left was to make more material and push it to the finish line. Starting from the substrate shown below, we could indeed intercept Holton's intermediate (see the dirty NMR).

(top) Holton's and our substrates. (left) Following the tradition, I celebrated the formal synthesis by drawing the structure, which turned out to be too soon. (right) What small-scale life looks like.

However, that was the time we recalled a lesson "There's no such thing in total synthesis as 'almost done,' nor 'the last scale up.' "  The deceptively simple-looking C–4 acetylation (again, challenging steps always look easy on paper) worked in 3% yield. We spent as long as 3 months trying to install this small piece, but could not fix the step. We decided to take a detour which fortunately worked out alright. Although Holton's procedures worked on some of our substrates, their reactivities were often different, which required some modifications to reaction conditions. Overall, the end game ended eventually which completed the two-phase synthesis of Taxol. 

Vanilla with banana cream cake. Yes, I put anchovies on pizza. 

As a reward of 5 years of journey, we got to have a cake thanks to Sol publically shaming Phil. In the end, we submitted from home because of the current situation. This also made me defend from home via Zoom (>100 people came!) but it went remarkably smooth! Maybe I am the only Baran alum who got Ph.D. from home?

Post-defense Zoo party. Thanks everyone for coming!

I should note that this synthesis was achieved by standing on the shoulder(s) of previous practitioners, of course, including those who worked on the two-phase taxane project before me: Yoshi (the first graduate student worked on taxane project), Abraham (a brilliant postdoc now PI at Stockholm University), Nathan, Mine, Changxia and Yehua. I just joined the lab at the end and happened to be the one to finish this. I would like to thank again all those who dedicated their time to this project.

– Yuzuru