Update 16-09-12: just found out that this has also been written about over at the greenchemblog. If you fancy an alternative take with more detail and less history, look over there!
Stereocontrolled organocatalytic synthesis of prostaglandin PGF2α in seven steps
I’ve heard it said a few times that if you can’t be the first to synthesise a natural product, then you should try and be the last; that is, you should aim to come up with the most elegant and efficient route you can now that the pressure’s off and there’s less of a rush. I think that some really popular molecules are still waiting for a ‘last’ synthesis (e.g. taxol), as even the best routes devised to date still seem ‘too long’ and are larded with too many oxidation level adjustments and protecting group juggling steps. Obviously synthetic organic chemistry – like any science – is constantly improving so the aim is to devise the best solution possible at the time, and to learn something along the way. A good synthesis is timeless.
A recent example of a well thought out synthesis that I recently enjoyed was Varinder Aggarwal’s synthesis of prostaglandin F2α. The prostaglandins are one of those “mature” families of targets that’s had much synthetic love lavished on it by some of the best synthetic chemists of the past forty years including Woodward, Corey, Mulzer, Stork, Noyori and Danishefsky. Consequently, there’s not much point publishing in field unless you’ve come up with something really clever. But it’s still very much worth trying; although they might seem a little simple by today’s standards the prostaglandins are very important bioactive compounds. Heck, the Nobel Prize for Medicine in 1982 was awarded to three chemists (including Brit Sir John Vane) for figuring out what they did in the body. And if that's not reason enough, there's the fact that no less than three analogues of prostaglandin F2a are currently used as prescription drugs. The most successful of these – Pfizer’s Latanoprost – makes around $1.6 billion a year.
Corey achieved the first synthesis of prostaglandin F2a in 1969 in 17 steps; which wasn't a bad benchmark for the year and the target! He then spent a lot of time improving on his first generation route, and it’s interesting to read his follow up papers for this purpose. In fact, despite a few flaws, Corey’s work actually became the basis of the 20-step commercial route used by Pfizer for commercial production of Latanoprost. I had Corey's original route drawn up for something else, so I've included it below.
R. B. Woodward weighed in four years later with a slightly shorter formal synthesis that relied on very similar tactics, but would probably be much harder to make asymmetric. As I had it drawn out for a Woodward Wednesday that never seemed to happen, it's shown below for comparison. Although the chemistry is quite simple and easy to follow it does have a few less common steps including a Payne epoxidation and a Tiffeneau-Demjanov Rearrangement. I don't think I've ever met anyone who’s done either.
Looking at these two classic routes you can appreciate the common tactic of reaching a ‘Corey lactone’ type intermediate, followed by attachment of the two sidechains, an approach shared by pretty much all syntheses of this family (with the notable exception of Stork's, which you can buy printed on a mug here). Unfortunately, both Corey and Woodward access these key bicycles in rather cumbersome ways, forming one ring at a time, and neither is enantioselective (although Corey did eventually develop a number of asymmetric Diels-Alder reactions for his). Now let’s take a look at Varinder’s solution. Although it doesn’t pass through the Corey lactone itself, a similar bicyclic lactone was chosen as a key intermediate prior to attachment of the sidechains. The really clever part was that this was formed in a single step from simple succinaldehyde using a nifty proline-based organocatalytic cascade that delivered the product in excellent ee. Shown below is the sequence of reactions that had to occur, which seems simple enough on paper. Brandon over at Organic Chemistry Tips and Techniques talked about some of the practical details of this step in more detail when the paper came out a couple of weeks ago and if you’re interested in this aspect of the work you should check out his post.
It turns out that most of the optimisation of reaction conditions (i.e. solvent, catalyst, stoichiometry, time and additives) is buried in the paper’s SI, but the many pages of tables do a good job of telling the story of how the group ended up with the conditions that they did. As shown above, the idea was to carry out two sequential aldol reactions in one pot to end up with a bicyclic intermediate a bit like Corey’s. The first of these was a intermolecular aldol reaction between two molecules of succinaldehyde. This reaction was found to occur quite readily with excellent ee under proline catalysis but the problem was finding conditions that would discourage further intermolecular reactions after the first and would favour cyclisation and dehydration of the dialdehyde to give the desired product. Normally you’d be quite confident about selectively carrying out an intramolecular reaction (to give the desired compound) preferentially over competing intermolecular reactions (leading polymerisation and low yields) but it turned out that proline, which so effectively catalysed the first step, did not effect the intramolecular aldol condensation. This meant that after the first step had occurred a second catalyst had to be added to form the desired bicyclic aldehyde. Of course, this second catalyst could not be present at the same time as any of the succinaldehyde starting material as it would then catalyse unselective intermolecular reaction leading to racemic products and erosion of ee. Sound like a pain to optimise? It was.
The conditions the group eventually found gave the desired compound in just 14% yield (after treatment with methanol to give a more stable product). Next, a suitable cuprate was added to the enal, and the resulting enolate was trapped as the TMS enol ether. Ozonolysis of this enol ether, followed by reductive work-up with sodium borohydride successfully set the compound’s final stereogenic centre, with hydride unsurprisingly being delivered from the less hindered convex top face. All that remained was to introduce the remaining side chain, and this was easily carried out using some well established Wittig chemistry to give the product in just six steps from succinaldehyde (which is made in one step by hydrolysis of 2,5-dimethoxytetrahydrofuran).
Despite all the talk above about ‘last’ syntheses, I’m not saying that this work can’t be improved on or will never be beaten, but it’s a nice leap forward making use of one of the great developments of the past decade – organocatalysis – and it’s raised the bar for this family much higher. A great effort!
1. I am still in awe of some of the ‘classic’ syntheses from 50 years ago. Take, for example, R. B. Woodward’s synthesis of reserpine. It’s been made loads of times since in shorter and (much) higher yielding ways, but I’m still blown away by how clever his solution was, even though it only uses common reagents you’d find in any undergrad textbook.
2. First up was the racemic synthesis shown, then a protocol for resolution of an early intermediate using ephedrine was published. Next came a new route based around an asymmetric Diels-Alder reaction, but using a whole equivalent of (Corey’s own) 8-phenylmenthol chiral auxiliary. Finally a couple of catalytic asymmetric Diels-Alder reactions were developed using the now familiar aluminium and oxazaborolidine based catalysts. The setting of the tricky side chain hydroxyl stereocentre was also an early test piece for the CBS reduction methodology. I’m fairly sure that several of Corey’s routes to this target are covered/compared in some detail in the first Classics in Total Synthesis book, but I've lent my copy to a friend so I’m not entirely sure. Although I think the claim that 'total synthesis is the fountainhead of new methodology' is made a bit too often, it’s hard to argue with it here!
3. By any measure that’s poor, but at least it’s the first step. Along with the last step, that’s one of the best places to have your worst transformation. In the total synthesis I’m working on at the moment my first step proceeds in 15% yield, so I can sympathise a bit with these guys. Although at least they can still make 15 grams at a time, which is way more I can.