It’s hard to imagine a functional group that’s been more heavily studied than amides. They are, of course, the literal backbone of protein chemistry, and they’re hugely important in all sorts of organic synthesis. The number of drug candidates with an amide bond formation in their synthesis could not be counted. So you’d think that we know pretty much what there is to know about them by now.
One of those things that we do know is that much of an amide group’s character is due to resonance. The double-bond oxygen is more properly thought of as less than a double bond, and the single bond to the nitrogen group is actually more than a single bond. That’s a big factor in the slow rotation around that bond, which can often be observed in NMR spectra through separate peaks for the two rotamers. You only see the two because that bond resonance tends to keep the whole group planar, so it’s sort of a click-stop from one to the other (the cis-amide to the trans and back).
But what if it’s not? It’s been known for some time now that many cyclic amides (lactams) and bicyclic systems don’t have the classic planar geometry, and this definitely affects their reactivity. In recent years, people have started taking deliberate advantage of this, and this new paper (from the Szostak group at Rutgers and collaborators at Yangzhou) is a good example of that. If you functionalize the NH of a secondary amide via a Boc or tosyl group (and this can be done transiently as part of a one-pot reaction), you get a nonplanar “non-amide” that’s suddenly much more reactive. You can, as the paper shows, react this with another different amine and do a transamidation at room temperature, which is something that you’d be waiting around for years to happen otherwise. There are other ways to do this, but it’s hard to come up with something else this mild.
Amide formation from an acid and an amine, via some sort of dehydrating reagent, is of course one of the most classic organic chemistry reactions of all. It’s long been a standing joke among medicinal chemists that if you can make amides and run palladium-catalyzed couplings, then you’ve pretty much got the whole toolkit to make a success of yourself in the lab. It’s funny because it’s sometimes uncomfortably close to the truth – few indeed are the medicinal chemists who have not set up arrays of either reaction to crank out a list of new analogs off a given scaffold.
Now, if you can’t get amides to form, there’s either a big problem with your compound or a big problem with you, because most of the time that’s a slam-dunk. Metal-catalyzed couplings are another story. You can usually get something to happen with standard conditions, but it may be ugly, and every substrate in a series can act a bit differently. On the flip side, it’s a truism that every single Pd-catalyzed coupling can be optimized to high yield if you’re just willing to spend enough of your life messing around with the conditions (process chemists will be nodding their heads wearily).
Here’s another new paper that illustrates the point. The authors (from Eli Lilly and collaborators at University College London and Penn) are looking at conditions for the C-N Buchwald-Hartwig coupling, which is very widely used, and whose yields are very widely known to be subject to a lot of different factors. Taking a cue from the “robustness test” proposals for new reactions, they take a model coupling reaction and try running it in the presence of a whole list of other small molecules containing different functional groups, to see how these might interfere. Screening these across a set of standard-ish coupling conditions gave over three thousand reactions, which were evaluated in 96-well plates under high-throughput conditions (which is really the only way to get through something like this).
Many of the functional additives had no real effect on the reaction, but there were several classes that did knock the yields down. (It should go without saying that none of them raised any of the yields; organic chemists know that that just doesn’t happen!) The team then went back and tried optimizing these problematic ones, varying the solvents, temperature, catalysts, etc. in order to find new conditions that would bring things back up. These could indeed be found, but high-throughput techniques came in very useful there, too, because the variations are many and subtle. To give you the idea, the model reaction (2-bromonapthalene coupling with morpholine), when run in the presence of benzenesulfonamide, was very sensitive to the base and solvent conditions. Potassium t-butoxide in dimethoxyethane gave 5% yield, but the same reaction in dioxane (which is merely the cyclized form of dimethoxyethane, when you get down to it) was quantitative. That’s just the sort of thing that makes you hold your head with this kind of chemistry.
Plowing through these variables was easier when some sort of mechanistic insight pointed the way, but that’s not always the case (and some of those mechanistic insights turn out to be wrong, anyway). The idea is to do this sort of mechanized gruntwork up front so that libraries set up with valuable intermediates will go on to have a higher success rate, which is a worthy cause. What’s good here is that the authors have shared the fruits of their labors with the rest of the community, because there are a lot of Buchwald-Hartwig reactions run out there, and a lot of them could use an improvement.