Friday, January 9, 2015

Academia–industry collaboration in the route optimization of taxadienone

In early 2012, we reported the gram-scale synthesis of a non-natural taxane, taxadienone, as well as that of a natural taxane, taxadiene. Now, chemists at Albany Molecular Research Inc. (AMRI) report a route optimization of taxadienone in none other than the process chemistry journal, OPRD.

Before we describe how the story of this collaboration came about, we will take you through the route optimization process. The synthetic route itself is identical, as all the intermediates of our synthesis appear in their synthesis as well, but the reactions have been scaled up and the yields have been improved. 

Reaction scale-up: We made ~2 g of taxadienone whereas AMRI made ~10 g.
Reaction yields: Please see the figure below.

Although the contents of AMRI’s paper will not be reiterated, we will show their concluding paragraph here (almost verbatim from their manuscript):

“The reported route to taxadienone was successfully optimized and scaled-up to decagram quantity. Thermal hazards associated with the production of bromodiene were addressed by employing a continuous flow reactor. The crystallization of a cyclized diketone at the penultimate step proved to be a decisive factor for obtaining taxadienone of high quality.”

Now, as for the behind-the-scenes story. This story started out as an interesting “experiment” in academia-industry collaboration. Our laboratory is engaged in many collaborations with industrial groups, including LEO Pharma, Bristol-Myers Squibb, and Sigma-Aldrich, and in most cases, our industrial partner has a project goal toward which we provide expertise and in-house research findings (industry —> academia outsourcing). This Baran–AMRI collaboration has actually been a “reverse collaboration” in which our initial synthetic route was taken up by an industrial group for scale-up (academia —> industry outsourcing). Through many interactions, by email, by phone and in person, AMRI saved our group much time and effort by generating large amounts of enantioenriched taxadienone. With this extra time in hand, we were able to study the front-line chemistry for longer periods of time, resulting, for example, in the synthesis of taxuyunnanine D. AMRI’s work also validated our synthesis by having an independent group reproduce our results, even when some of the reactions can be tricky. This “field-testing” of chemistry further refined our initial work, when some reactions were difficult to scale up (even though our initial synthesis was performed on a decent scale already). 

Although this sort of “reverse” academia-industry collaboration is rare, we learned a lot from this experience! We understand that there is a time and place for this type of collaboration but we believe that in the near future, such collaborative work will be more commonplace. Finally, this is a wonderful advertisement for the impressive capabilities of the AMRI team and we recommend all our industrial friends that are looking to outsource challenging chemistry to give AMRI a try!

Written by Yoshihiro Ishihara
Uploaded by Nathan Wilde


  1. AMRI seems to be indicating that the critical Diels-Alder step doesn't scale very well under the current reaction conditions. They report a much lower isolated yield and, unfortunately, do not comment on the diastereoselectivity that was observed in their hands. Have your groups discussed this discrepancy? Are their recoveries low due to mother liquor loss or are they observing differences with regard to conversion or by-products?

    1. The DA reaction mix is quite rich: lots of isomer + unreacted start material. Seems kind of hard to separate, without running a column, in a good yield.

    2. They don’t mention the dr for the aldol/oxidation sequence, and their aldol looks quite different from ours (no excess EtOH or water, run for 5 days when ours is 24 hours). I would guess the main issue they had was a decrease in diastereoselectivity. I have personally seen the dr waver a bit. Poorer dr gives a poorer yield for the Diels–Alder because we take both diastereomers on. I would guess loss in recrystallization could be a second problem. These reactions are also run at quite a bit larger scale than we reported, so differences in scale might be a third reason for the discrepancy.

      But to answer your question, we haven’t discussed this disparity with AMRI. They were very independent with their findings and with publishing this paper, which I think validates the route.

    3. Thanks, Nathan. It's really great work, nonetheless. As I'm sure you're aware, the strained 8-membered B-ring of taxol has been a problem for chemists for the last 25 years…not easy stuff. Masahisa Nakada's group recently reported a palladium-catalyzed process to close the B-ring in 97% yield, the most efficient to date…probably not conducted on gram-scale though. Reproducibility on scale is really important for a process to be considered viable in an industrial/manufacturing context. I'm sure that you guys will get there. Good luck!

    4. Yes I know all about the B-ring of taxol. The entirety of my nearly 4 years as a grad student has basically been focused on the formation and functionalization of that eight-membered ring. And yes I did see Nakada's synthesis. It is very nice, and focusing on the B-ring problem really highlights the strengths of that synthesis. If you're interested in the different ways to make the B-ring, you might be interested to know that we are not the only, or even the first, to use the Diels-Alder strategy. Williams made taxadiene (JOC 1995, 60, 7215) with pretty much the same conditions but with a slightly different substrate, and they report 31% yield.

  2. It's always preferable to close two rings at once as compared to sequential and long as the yields are acceptable :)

  3. industry —> academia outsourcing, you sir are right !

  4. nice ,i found it a very interesting blog . hay, i m dealing for surgical instruments