Wednesday, December 17, 2014

Functionalized Olefin Cross-Coupling

Figure 1. Functionalized olefin cross-coupling at a glance.

The last paper of the year from the inner depths of our lab has made its way out earlier today. It’s the bigger brother of our reductive olefin coupling paper from January. In a nutshell, reductive olefin coupling allows chemists to break apart molecules into carbon-substituted donor and an acceptor olefins. By using an iron catalyst (available through Sigma-Aldrich if money is no object) and PhSiH₃, one can easily forge a C–C bond between those two components. However, all of the products we made in the last paper could already be accessed using other radical conjugate additions since they proceed through the same intermediates.

The new paper kicks things up a notch by throwing heteroatoms into the mix, expanding the scope beyond carbon substitution. Functionalized olefin cross-coupling allows the periphery to be decorated with various heteroatoms. And by various, I don’t mean just one or two types. I mean NINE different types. We even had a nifty graphic to show this, but it never made it to the final draft of the paper.

Figure 2. Heteroatoms utilized in this paper.

As you can see from Figure 2, you can have the donor olefin substituted with pretty much any element you would want to use as a synthetic organic chemist. Traditionally, heteroatoms throw a wrench in the generation of nucleophilic radicals, as they result in starting materials that are either labor intensive to synthesize, are chemically unfeasible entities, or lead to functional group incompatibilities or other chemoselectivity difficulties. Functionalized olefin cross-coupling easily circumvents all of these problems, allowing one to use the hundreds of thousands of heteroatom-substituted olefins that are already readily accessible as latent radical donors. The neatest part of the work conceptually is that you can basically treat all of these heteroatom-substituted donor olefins the same regardless of the exact identity of the heteroatom involved. This is unlike traditional heteroatom substituted olefins where the identity of the heteroatom typically dictates the reactivity of the molecule (e.g., the traditional reactivity of enol ethers in no way parallels the reactivity of vinyl iodides).

The bulk of this work is embodied in Figure 3 in the actual paper. It has some generic caption like “Adducts synthesized by functionalized olefin cross-coupling,” but we’ve always internally referred to it as the “carpet bomb,” which was a term coined by PSB himself. What started out as 10 entries began to multiply like rabbits. At one point I was joking with Phil saying, “before you know it Phil, we’re going to have 100 substrates!” and he got this kid-in-the-candy-store look in his eyes. I then realized I pretty much shot myself in the foot at that point, but luckily we capped out at around 60 substrates.

As a consequence of the size of our carpet bomb, we naturally found some cases where our reaction didn’t give the best yields. We actually decided to include some of those in the paper, but we were afraid that the reviewers would reject it if we stuffed it full of too many low yielding results. Since the best part of a paper is finding out what didn’t go so smoothly, I’ve decided to compile a table of around half of our low yielding results, along with a few examples that didn’t make the paper for one reason or another. The contents could be useful for those who just need to make their product and don’t necessarily care about the yield.

Table 1. Functionalized olefin cross-coupling B-sides.

A fair amount of the products in Table 1 were synthesized during some feasibility studies that I ran in February and were never later optimized (see compounds 1, 3, 8, 12, and 13). Truth be told if we got extremely low yields on the initial hit, we usually opted to abandon the donor immediately to find one that gave a higher yield. Dichloride 4 was an interesting adduct since the corresponding donor contained a pretty sensitive allyl chloride functionality. Despite the conditions for our reaction being pretty mild, they still competitively reduced the allyl chloride to generate an adduct analogous to 4, but bearing an additional methyl group instead of the chloromethyl group. Additionally it’s worth pointing out that yields with phenyl vinyl sulfides bearing extra substitution around the olefin (911) were pretty low even though the analogous alkyl vinyl sulfides were competent donors (see 51 and 52 in the actual paper). 

After we submitted the paper, we realized that we didn’t have nearly enough substrates with nitrogenous heterocycles for a PSB paper, so that’s the reason for the inclusion of pyridine 2 and the super med chemmy 5 in Table 1. We were also able to show post-submission that a vinyl phosphonate (6) and a vinyl phosphonium salt (7) could be used could be used as donors to generate classes of adducts, albeit in moderate yields, that can be used for subsequent Horner-Wadsworth-Emmons and Witting olefinations.

Figure 3. Running the reaction in unconventional solvents.

On a less serious note, after group meeting one day, Phil told me to go to BevMo and pretend I was shopping for solvents for our reaction. The thought was that since we typically run the reaction in ethanol, it’d be pretty cool to see if we could also use various spirits as a solvents. Successful couplings would show that our reaction could tolerate a sea of random flavor and aroma compounds. Sounded like a good task for a Saturday and it was an even better Saturday when we found out that it worked! After our first round of success with vodka, tequila, gin, and whiskey, we found that Stone IPA, a chardonnay, and a merlot also worked as solvents. The purple/reddish spot at the top of the TLCs in Figure 3 is the desired coupled product.

For my previous project’s blog post, I made a video of myself setting up a reductive olefin coupling. However, having a seasoned reductive olefin coupler setting up the reaction just doesn't do justice to how easy it really is. What’s now pretty routine for me might not be easy for someone who’s never touched Fe(acac)₃ and PhSiH₃ before. I’ve spent a decent amount of time thinking about people who could possibly make a guest appearance in the video this time. A couple of months ago, I told my mom how she could probably set up the reaction in her kitchen. But from a safety standpoint, it’s probably not the best idea for her to be cooking up methyl vinyl ketone outside of a hood. Then I thought about the lab admin. She’s not a chemist at all—sounds perfect! But she did set up that zinc sulfinate (now branded as Baran Diversinates, name courtesy of Sigma-Aldrich) trifluromethylation of caffeine so she did do some chemistry recently. I needed someone significantly rustier than her. Someone who hasn’t set up any sort of reaction in years…

I eventually found the perfect person to set up a functionalized olefin cross-coupling reaction. It’s so easy, even a PI can do it.

Wednesday, November 12, 2014

The Recipe!

“What’s cooking?” was the question that Phil had asked me whenever he wanted to start a conversation back when I was a second-year student. Apparently, at that time, he was into “what’s cooking in chemistry.”  In this post, I would like to talk about how did we come up with the idea of using quinone diazides in "our recipe" (a recent ACIE publication,) but firstly…what is a quinone diazide? it is a "handsome" diazo compound! We classify it as an acceptor/acceptor diazo compound. It was discovered in the 19th century, so it’s a really old species. Diazonaphthoquinone (DNQ), an o-quinone diazide, is commonly used in a positive photoresist called DNQ-Novolac photoresist.

    When I first joined the lab, I was given a retrosynthetic analysis of prednisone, a precursor of cortistatin A.  In Baran lab terminology, my research topic is described as “ a cyclase phase of steroids.”  Breaking the B ring of a steroid probably results in the most convergent synthesis, and according to Corey’s retrosynthesis logic book, it is a topologically strategic disconnection. Interestingly, Denmark has proposed an identical retrosynthesis for the cortisone skeleton, and has successfully made the right side, but has never accomplished the molecule (TL, 1984, 1231).  Personally, at first I wasn’t really into this project because originally I wanted to deal with reactions that are “weird,” reactions that you don't really get what is going on without deeper thought. However, since I didn’t have anything else in mind or other options, I told Phil “yes” after a week of consideration.

   The key step of our retrosynthetic analysis is an enol-phenol oxidative coupling that can go wrong for a hundred reasons: thermodynamically unfavorable (dearomatization), kinetically unfavorable (para is more sterically hindered), and stereoselectivity issues... After trying this key step for 9 months without achieving any C-C bond coupling, one afternoon, Phil called me to a white board and started discussing about other retrosynthetic possibilities.  At one point during the discussion, Phil accidently (!) connected the C9-C19 bond to make a cyclopropane and asked me if we can make it from…the corresponding diazo compound. Looking at the weird diazo compound for a few seconds, Phil shook his head, and asked me to come back for another discussion.  
   When I went back to my office, thinking about the strange species, Will (a senior student at the time) yelled to me “it exists!!!” …Apparently Will had also witnessed our discussion and immediately ran a Scifinder search.  I conducted more detailed searches and brought it to Phil’s attention.  He replied, “this is it, Hai!” and this was how we “re-discovered” quinone diazide chemistry. 
   I quickly prepared a simple substrate to test the key reaction.  Luckily, it worked perfectly after several trials.  I went to discuss with Phil after confirmation of the desired product, and in less than 5 minutes, he gave me a “recipe” (this was around July 2012).  If you have read our paper (accepted in Aug 2014), you could easily recognize that the only thing different was that it’s in ACIE instead of JACS, or the lack of some “flavor” that I couldn’t achieve.
   I continued working on this recipe for the next two years.  It took me less than two weeks (14 days) to make the model substrate (without the D ring), however, it eventually took me more than 14 months to make the desired cyclopropane. The rate-determining steps were: making the right side, conditions for the conjugate 1,4-additions (Turbo Grignard was superior), and finding the magic effect of sub-stoichiometric Li2CO3 as an additive for Eschenmoser’s methylenation, and of course the scale-up.  However, I was lucky with the double deprotection to release the quinone diazide and the key cyclopropanation, which worked at the first try.

    The cleavage of the other two C-C bonds of the cyclopropane intermediate was achieved easily, but the C9-C19 bond (that Phil accidentally connected) was unbreakable, even after half a year of hard work.  This is kind of unfortunate because the product of that C-C bond is a natural product, a really important “flavor” of the recipe.  Another 4 months of optimization in the reaction of quinone diazides with olefins (Table 1 in “recipe tables” or 1 and 2 in the paper), and Phil let me submit, and luckily the recipe was accepted to Angew. Chem.

Wednesday, August 27, 2014

Diterpenoid-Alkaloids are so...Fancy (You Already Know)

The second paper on our efforts toward diterpenes and related diterpenoid-alkaloids is out in JACS now.  This is building off of work previously published in Angewante last year.  The story of this project goes back more than 5 years.  I’ll try to spare most of the painful details.

Summer 2009: My initial project was to make ent-kauranes and ent-atisanes and oxidize them with what people used to call C­–H activation and now seem to call C­–H functionalization.  (Does that make me sound old?)

Several months later (Spring, 2010), I came across David Gin’s synthesis of nominine.  Like so many others, I was awestruck. Being relatively new to complex-molecule-synthesis, it took me a few months to realize that nominine (and other hetisines) were structurally related to ent-atisines.  I didn’t notice this right away because people draw these diterpenoid-alkaloids in strange ways (see below).

I realized that I could target these more complex alkaloids from ent-atisanes with C–H functionalization and as they say, the rest is history.  I drew up a ridiculously ambitious and naïve plan for Phil in May 2010.  The plan was impossible, so of course Phil gave me his blessing to work on it.  It was going to be as easy as 1, 2, 3…whatever that means…

Step 1: Nothing ever goes as planned.  Three long years passed before we finished steviol (step 1 of 3).  Meanwhile, behind the scenes we worked on what was arguably the more interesting part of the project: making steviol into a bunch of complex molecules with complex reactions.  How did we start on steps 2 and 3 when we hadn’t made steviol?  Well, we bought it…sort of.  We bought stevioside—5 kilos of it to be exact.

Now, at this point you are probably wondering, “Why would Emily go through the trouble of making steviol if she already had access to decagrams of it?”  That’s a great question, but one that I don’t have time to answer right now.  Remember, I’m giving you the super short version of this story.

Step 2: Making isosteviol from steviol is known and boring, so I’ll go ahead to the synthesis of methyl atisenoate.  This chemistry is pretty straightforward.  The cute maneuver in here is the Mukaiyama peroxygenation/fragmentation sequence (proposed mechanism shown below.)

Step 3:
Part a: For the atisines, the major challenge was the selective C20 C–H activation.  I had to experiment with many different directing groups, light sources, solvents, reagents, and temperatures to optimize this one.  It took about 5 months to identify the best directing group and get the reaction working well.  Depending on the conditions, we can either get over-oxidation to give the imine or not.  In the case of isoatisine, we wanted that oxidation so we ran with it.  In 9 easy steps, we can get to a substrate containing an aza-ent-atisane skeleton and oxidize it.  In the same pot we hydrolyze the imine to an aldehyde.  Why? because it’s labile and it was easier to just take it off for characterization purposes.  After elimination (Martin’s sulfurane was key for exo-selectivity) and diastereoselective allylic oxidation (a la Gin’s nominine synthesis), the synthesis could be completed by simply adding ethanolamine into the iodo-aldehyde.

Part b: “It would be challenging to exaggerate the difficulty experienced while attempting to forge the C20-C14 bond present in the hexacyclic hetidine skeleton.” (I wanted to put that line in the paper, but some people felt it was too sensational.)  No joke folks, it took me YEARS to find a good way around this problem.  Years.  I tried obvious ideas, not-so-obvious ideas, good ideas, plenty of bad ideas, simple ideas, complicated ideas, and every idea in between.  I talked to Phil about this over and over again.  I talked to my poor lab-mates about this over and over again.   I even spoke to random professors visiting Scripps if I had a chance to meet with them and discuss my chemistry.  The final idea is very simple, but please don’t equate simple with easy:   

Yup, all I did was take a very similar iodo imine to the one I had made previously and heat it up with some allyl amine in methanol.  To all the haters out there who look at this and say “well, duh,” I say, “Where were you for the past two years when I needed a good idea!?” 

Part c: When the literature let’s you down:  After deprotecting the hetidine core, we were ready to go after the hetisine skeleton.  I really thought I would just take some old lit procedures performed on pretty much identical compounds and that would be that.  There are only so many ways to magically net-dehydrogenate something with a secondary amine as your functional group handle.  

After trying those lit procedures and having them fail for us over and over again, we did what we always do: we tried to come up with something so crazy it just might work.  We thought of nitrenes, nitreniums, trans-annular hydride shifts, every variation of an HLF reaction we could fathom.  Maybe if I had more time, something would have panned out, but nothing we tried worked before it was time for me to move on to greener pastures.  So yeah, my last step failed and I didn’t make the hetisine skeleton.  Bummer.  This will probably haunt my dreams for the rest of my existence.  (Actually, I graduated about a month ago and I’m over it.)

To end on a positive note: I did a lot of cool C–H activation chemistry.

And I grew this awesome crystal! Check out that sweet N-Cl bond!
Like so many students working on complex natural product total synthesis, I worked many years to try a final, supremely amazing, and well-precedented key step only to have it fail.  Luckily, when your failures are good enough, you get to publish those too. 

At my thesis defense, a first year student asked me if I had any advice to give to the younger students just starting out.  I said something totally expected like, “Don’t give up.”  Now that I have actually had a chance to think about that, I’d like to change my answer.  I would say that when you are going through difficult points in your chemistry during your graduate school, and (if you’re doing it right) you inevitably will, remember that this is SCHOOL.  You are here to LEARN.  Failing at your chemistry is not the same as failing at graduate school, as long as you learn something and become a better chemist as a result.  I really believe that.