The Hitchhiker’s Guide to RAE Cross-Coupling

The Answer to the Ultimate Question of Carboxylic Acid, Heterocycle and Everything

We are excited to announce that our paper “Alkyl-(Hetero)Aryl Bond Formation via Decarboxylative Cross-Coupling: A Systematic Analysis” (in the following referred to as the Guide) has today been published in AngewandteChemie International Edition!

I joined the Baran lab as a visiting student around six months ago. When I had my first meeting with Phil, he had two important messages for me. First of all he introduced me to the project and one of the things he showed me was the sales statistics of N-hydroxytetrachlorophthalimide (TCNHPI). This reagent was commercialized as a typical activating agent used in cross-couplings of redox-active esters. Impressively, all major pharmaceutical companies world-wide could be found in the list of buyers – RAE cross coupling was already being applied broadly and its adoption was proceeding at blistering pace less than a year after the invention of this reaction. At this point four different reaction types had been published, and we were using several different RAEs and various procedures for their generation. We had gathered rich in-house knowledge by that time and after running thousands of reactions my lab mates and I had developed a "feeling" how to get almost any substrate to work. As we did not want it to be lost forever, we were asking ourselves how can we pass this valuable information on to all the chemists out there? The answer was rather simple – we were going to create the Guide.

Owing to my German origin Phil's second advice was rather practical: American highways are not the Autobahn, and if I was speeding I should do it carefully - "there is a science to speeding!"

Starting to work in the lab I was certainly not lost in time and space, but still in need for the friendly advice DON’T PANIC. There was a maze of four publications with more than 900 pages of SI and more unpublished material in front of me. The first challenge was to get an overview of all the published and unpublished procedures, and to arrange them in a simple and clear manner. Our team developed a matrix chart, in which four different catalytic systems are fixed on the x-axis, while four different modes of activation are fixed on the y-axis giving a total of 16 different procedures (Figure 1). Combined with a visual identification of the success of the reactions by a simple color code the Heart of the Guide was born.

Figure 1. a) Activating agents, nucleophiles and catalytic systems in RAE cross-coupling. b) Simple classification combined with visual identification.

Next, Matt and I selected nine different substrates covering a broad chemical space. The Guide offers a selection of variously functionalized primary, secondary and tertiary carboxylic acids including α-amino acids, heterocycles and bioisosteres (Figure 2a). Each of these substrates was subjected to our 16 different reaction conditions and the obtained data was clearly arranged in a one-page table (Figure 2b). Looking at the number of red squares, we were definitely exploring the edges of our methodology here…! However, in each case we found at least one set of conditions to yield the desired product in more than 60% yield! This underlines how being able to fall back on many different conditions combined with a Guide can be a real asset when tackling challenging synthetic problems!

Figure 2. a) Selection of substrates for b) the Guide.   

In the second part of my stay, I tried to expand the scope of the Guide to new areas of RAE cross-coupling. What we were really excited about was applying more heterocycles in our chemistry. At this point we had mainly explored the nature of the alkyl carboxylic acid, as previous investigations suggested that the influence of the aryl nucleophile on the reaction outcome was rather small, except for heteroarenes. After some experimentation we discovered that methylpyrazole magnesium bromide undergoes simple Fe-Kumada coupling in synthetically useful yield! Encouraged by this result we decided to widen the scope of our Kumada, Negishi and Suzuki couplings to medicinally relevant heterocyclic nucleophiles (Figure 3).

Figure 3. Heterocyclic nucleophiles explored in the Guide.

Except for furan and thiophene all of the heteroarenes explored in the Guide contain basic nitrogens (“real heterocycles”). Searching the literature for instances where these heteroaryl nucleophiles were used in cross-coupling reactions revealed what we had already suspected: despite their importance in medicinal chemistry and other chemical industries cross-coupling of N-heterocycles is severly underdeveloped. During our investigations we found that the best choice of method for RAE cross coupling is strongly depending on the electron-density of the metalated carbon atom. While electron rich heteroarenes easily undergo Fe-Kumada coupling, Ni-Suzuki is the best choice for electron deficient heterocycles. Interestingly, we observed 2-pyridyls to be most challenging with our methodology. While the 2-pyridineboronic acids are known to be unstable, Kumada coupling turned out not to be feasible either. We think that the homo-coupling byproduct (bipy) is the culprit and deactivates the Fe catalyst. Fortunately, we could solve these issues by applying a modified Fe-Negishi protocol.

A good guide should provide you with overview and main directions at a glance, but also with detailed information where you need it. That’s why we have put together a Supporting Information which contains clear, exact and easy-to-follow experimental procedures to make the application as straightforward as possible. Besides graphical descriptions of the reactions and detailed characterizations of the products, our Supporting Information includes an overview of all reaction conditions in one scheme (Figure 4). We were delighted that this important part of the Guide was also acknowledged by the referees:

“The supporting information is of high quality […]”

“The Supporting Information, to the best of my ability to determine, is spectacular, and includes pictures detailing each step. Only a full-length motion picture could be better. The products are adequately characterized and look pristine.”

Figure 4. Overview of all 16 different reaction conditions.

Unfortunately, Angewandte did not print the words DON’T PANIC in large and friendly letters on its cover, but I still hope that the Guide will help many organic chemists to find their way through RAE cross-coupling – in our lab it already finds regular application.

Fred and the team of the Guide

We apologize to all those who expected 42 to be the answer to everything – we could at least confirm it in three cases.

11-Step Synthesis of (–)-Thapsigargin

Our synthesis of thapsigargin is out today in ACS Central Sci. Since we’ve prepared a detailed supporting information, we will forego discussion of the failed approaches & optimizations in this blog post (feel free to email us or comment if you’ve got any questions though). We will instead use this post as a means to openly discuss a defining aspect of not only this synthesis, but also of all of our 2016 syntheses: step-count.  Thus, this blog is intended to stimulate an alternative type of peer reviewing–we present our reasoning behind a simple definition of a step and seek opinions from all readers.

Similar to other recent total syntheses (pallambins, maoecrystal V, and araiosamines) emerging this year from our lab, our manuscript was originally titled “11-Step Synthesis of (–)-Thapsigargin” (kinda gave away the punchline to the tweet here). In doing so, we abstained from using subjective and assertive descriptors such as “concise” and “short”. Here is the direct feedback we got from peer review:

"Their claim of 11 steps in a title is unhelpful, I would accept a “Short synthesis of…”, but not a step count. This sets a bad precedence in that we no longer claim “a first synthesis” as being valid. The same is true of step count since as the authors are well aware, step counting is just one of many criteria to judge a synthesis. If we allow a step count number to become normal practice, all total natural product synthesis will have to begin with this count of the synthesis steps. As we know this says nothing about efficiencies or the elegance or costs of an approach."

And:

"The work includes the step count as primary consideration, something that has become a trademark of the author, as noted in the title. There are serious problems with this metric as practiced by the author: (1) the count is conducted in a manner not entirely transparent or logical, as such it does a disservice to the community and pioneering efforts of others that have previously worked on the targets; (2) it leads to erroneous conclusions. The principal investigator has repeatedly been engaged in such miscalculations; and prior ways of counting should not justify continued obfuscation."

This is really valuable feedback as it is likely that if these referees believe this then certain members of the community must also feel the same way. Although we are perplexed by the reviewer’s assertion that somehow the exact step-count included a title makes the first synthesis invalid, the other points raised by the referees deserve comment. To us, inclusion of the step-count in the title solely provides an immediate, objective, and unopinionated depiction of the synthesis – it informs the readers that the synthetic sequence disclosed comprises 11 steps.  And we're certainly not alone as there have been dozens of high-profile total syntheses by other groups published with step counts in the title. By no means is this a Baran lab "trademark".  

The more serious accusation, however, is that our definition of step-count is somehow not logical or transparent. Or even worse that we are engaging in outright deception. Open-Flask was started years ago specifically to bring more transparency to our research. We feel that the most transparent and fair way to discuss this is out in the open rather than in the comfortable anonymous basement of peer-review.

IUPAC defines a step as a process that "proceeds through a single transition state" which largely pertains to physical chemistry phenomena. When organic chemists refer to a step of a synthesis they are usually referring to what goes on in a flask (reagent additions, etc.) followed by some sort of work-up or purification (which signifies the end of a step).  On several occasions we’ve explicitly defined that a single reaction step is one in which a substrate is converted to a product in a single reaction flask (irrespective of the number of transformations) without intermediate workup or purification. This definition seems to be the most pragmatic and encompasses what most organic chemists think of when speaking of a step (things go into a flask followed by a work-up which signifies the end of a step).

Although we can't rule out hacking from an outside group, or other sorts of rigging, the Twitter community seems to mostly agree as well based on this super-scientific poll we did the other day:

Yes, some people will argue that amide-bond formation should be counted as 2 steps...

Some of the discussion on Twitter is really revealing. Many people have different opinions as to what constitutes a step or not with some arguing that the whole concept of a step is outdated and should instead be replaced with other metrics, such as number of operations or even person-hours. 

In our most recent syntheses as well as this one, we have strictly conducted our step-count according to the definition outlined above. While not perfect, it is a simple definition that accounts for many types of reactions. For example, a number of classic transformations (e.g. Swern, Ugi, Passerini, Strecker) involving multiple distinct intermediates (the last 3 isolable) and all fit into this definition. So do cascade or tandem-reactions. How many steps is Corey's legendary aspidophytine total synthesis if we break up the key step into individual components (multiple reagent additions and at least eight elementary intermediates, some of which are isolable)? How about Noyori's classic prostaglandin synthesis that introduced the world to vicinal difunctionalization (multiple reagents added and 2 new C–C bonds generated)?  Let's take the Swern oxidation as a glaring example of this discontinuity in step count – this venerable reaction comprises three distinct transformations as shown here. 

How many steps is this step?

However, since all of them take place in a single flask, it has always been considered as one step. So, three different transformations actually happen, but the net result is that an alcohol is oxidized to a ketone.  Finally, what about the most used reaction in all of organic chemistry: Amide bond formation. There, one adds an activating agent like DCC to form an activated ester (sometimes isolable) followed by addition of an amine. Is that two-steps or one (a few people apparently think 2, see above poll)?   By now you might be rolling your eyes and thats the point. This is common sense. Most people will agree that the sequential addition of reagents or solvents to the same flask does not constitute a new step.  Filtering over silica or Celite, workup of any kind, adding scavenger resins – all of those things signal the end of a step with further operations on crude or semi-crude material representing a "telescope". For example, in Wender's synthesis of Phorbol (summarized graphically here), the following procedure is characterized as a single step in the overall step count (reported as 36 total) as presented the manuscript but I think most people would agree that the SI clearly describes a two step process (ketone to enol ether to alpha-bromoketone):

Flash Chromatography, rapid or slow, signifies the end of a step.

Now lets take a look at one of the steps in our synthesis that triggered the referee comments above. It accomplishes two transformations in a single flask (TBS installation and allylic oxidation).  By the logic outlined above, this should also be counted as one step. There is no deception here.  We are uncertain how alternative definitions of a step can be rewritten – after all, each single transformation within the “step” can be further divided into combinations of elementary steps.  

How many steps here?

Here's the procedure. Still looks like one step to us.

We've had this discussion in the lab and in group meeting to try to derive the contrary view. And it basically boils down to this:  If folks can get away with this then why not just take 40-step syntheses and turn them into 5 step syntheses by doing everything in one pot? But this is a vast oversimplification of how synthesis works and the strategic thinking that underpins the design of a multi-step synthesis in which orthogonal transformations can be incorporated into a single reaction vessel. Simply stated, not all transformations can get along together in the same flask. 

Indeed, one-pot multi-transformation steps are embedded into the strategy of a synthesis, meaning that when steps in which one pot reactions can be engineered (e.g. without sacrificing yields significantly & saving solvents, purifications, and manual labor), we went ahead and performed them. As a result, there’s a very clear logic behind conducting these one pot sequences. Although orchestrating such one-pot sequences can conceivably improve the overall efficiency of a concise synthesis, the improvements will be proportionally less substantial for longer syntheses. In other words, attempting to shorten a 40-step synthesis with just this tactic alone will be fruitless just like speeding up a conveyor belt does not always lead to a higher production rate.

Thus, we believe the discussion on step count is entirely perched on the overall strategy of the synthesis. 

Although step-count is an important metric for measuring the efficiency of a synthesis, we are not claiming that as long as it’s short, one can go ahead and ignore all the other criteria for a good synthesis. The best example we can think of where a longer synthesis can be better is in the commercial synthesis of Halavan where the heroic Eisai team, led by process legend Frank Fang, favored a longer route due to the identification of crystalline intermediates that facilitated purification. In the case of thapsigargin, we actually report an alternative longer sequence (14-steps) that offers a distinct advantage for the production of certain analogs (in this case a higher yielding photo-rearrangement and the ability to incorporate different esters via simple acylations) over the shorter sequence. Details of this other route are included in the manuscript & the SI.

Some might argue that a better way to present step-count in synthesis would be to report actual isolated intermediates instead of actual steps. At the end of the day, someone (referees, students in a group meeting somewhere, readers, your parents?) will be counting the actual steps (which begin and end with isolated intermediates) so we're not sure it makes any difference. Also, we are still on the fence regarding the removal of a solvent as representing a new step (there was some fruitful debate on Twitter about this - the classic Arndt-Eistert reaction for example usually involves a solvent swap but no workup) and lean towards a definition where PURIFICATION of any type signifies the end of a step as the evaporation of a solvent is no different than refluxing a solution (without the cap).

In any event, the short or concise or 11-step or 14-step (your preference) scalable synthesis reported here facilitated LEO Pharma's interest in a medicinal chemistry campaign based on this chemotype. They filed a provisional patent on the route, have successfully outsourced it on 100-gram scale, and we are currently working on interesting analogs with them using advanced intermediates they have provided. 

However one counts it, the synthesis of Thapsigargin is now scalable.

One of the most interesting publications on the topic of trying to design syntheses that incorporate multiple transformations in a single step (cascades) was written by Tom Kieboom, a process chemist at DSM and adjunct professor at Leiden University. In that insightful review, the challenge of mimicking Nature's ability to conduct multiple orthogonal reactions without intermediate work-up or purification is discussed. As someone mentioned on Twitter, by the current definition, Nature makes most of it's products in one "step".  One thing that is clear is that if chemists could make natural products the same way, it would be a cause for celebration rather than a moment to debate step count. 

For some other recent reviews on the topic of efficiency that incorporate step count into the equation, see these from process grandmasters Martin Eastgate and Chris Senanayake. In addition, the Krische lab's impressive perspective published earlier this year in JACS defines a step in a similar fashion (see SI - "a step is defined as an operation that does not involve any intervening purification/separation, including removal of solvent, commencing with compounds that are over $50/gram.").

We welcome any critique and feedback and look forward to an open dialogue of step-count here at Open Flask or on Twitter.  We realize and respect the fact that not everyone will agree with this simple definition of a step and thus, as the referee above stated, it is important to evaluate syntheses based on more than a single variable. 

Yours truly,

Thapsigargin team 
(thanks to Phil and other Baran Lab members for help with this post)

PCC in Your Stocking

Seasons' greetings! This blog post is about The Portable Chemist's Consultant (PCC) (see the last entry on PCC here). But it's not just another announcement on another update – in line with the Christmas spirit, we are announcing a significant price drop: you can now purchase one equivalent of PCC at only $9.99 (the surprise). [We want to make it a winter of our discount, not discontent.] You may recognize the unit "equivalent" as particularly "tantalizing" – we chose it because, unlike pyridinium chlorochromate (example chosen at random and no pun intended), our PCC is a chemical entity of ever increasing "molecular weight". In fact, in conjunction with this discount, we have launched a substantial update today, featuring about 180 new pages (ca. 21% increase in size).

The page count stands at 1003 – compared to its launch in 2013 (with 439 pages) – PCC has come a long way. This is only possible with the support of those who bought and used the book – our heartfelt thanks goes to you all. The number 9 is a homophone for longevity and eternality in the native culture of one of the authors (in Mandarin Chinese). $9.99 signifies our continued commitment to this project to provide the community (specifically those in drug, agrochemical, and materials science) with the most up-to-date information on heterocycles.  

More specifically, this update features:

•   a complete pyrazole substitution section
• a new chapter on quinolines (both synthesis and substitution)
•   updates to the speed consulting section, highlighting latest developments in the field
•   And a chapter on fused heteroarenes, including Phil's top-secret rules on how to disconnect 5,6-fused systems. 

We would also like to highlight the fact that PCC is gaining even more "portability" and can even be your "pocket consultant". Since multi-touch eBooks for the iPhone is now available through an iBooks Author upgrade, we’ve tested the PCC book on the iPhone and it looks and works great! The font is, of course, quite small but you can pinch and zoom to read the text, use the hyperlinks and watch the videos just like on your iPad or Mac.

Considering that there have been, under some estimates, a total of 1 whopping billion iPhones sold and a total of 225 million iPads sold, along with my non-statistical tallying of iPhone and iPad users around me, it’s safe to say that people are much more likely to have iPhones than iPads or Macs. With regards to the PCC book, this has been a pity because multi-touch eBooks worked for the iPad and even the Mac but not for the iPhone – but this is no longer a problem!

PCC on the iPad, Mac, and an iPhone

We're told that Phil himself uses PCC quite often when consulting for simple things like remembering which esoteric additive to use in an amide bond formation, or recommending just the right ligand in one of Buchwald's reactions, or even just a quick reminder on the best way to disconnect an awkwardly substituted imidazole.

In summary, PCC has no ego, no hourly rate, and gets better for free over time. And now, it will fit in your pocket. It's the perfect consultant.  Happy Holidays everyone!!

-Yoshi and Ming 

P.S. Screenshots of PCC on iPhone: