Word has been spreading rapidly about this preprint on Chemrxiv.org, from a joint UCLA/Caltech team. It details the use of the cryo-electron microscopy technique called micro-electron diffraction (MicroED) for the structure determination of small molecules, and it’s absolutely startling. I read it last night, with many exclamations along the way, not all of them fit for print (and that’s in a good way). And it turns out that basically simultaneously, a Swiss/German collaboration has published on the same technique, one day before.
Some quick background on what’s going on here, then we get to the hair-raising stuff. As the chemists in the audience well appreciate, X-ray crystallography is the gold standard for structure determination. If you can grow a good crystal of some substance, you can diffract X-rays through that lattice and work backwards to see what the unit cell of the crystal had to be. The hardware and software to do this have advanced hugely over the years (with worthy Nobel prizes handed out along the way), and although there can still be complications, for most small-molecule substances the main limitation is now contained in that phrase “grow a good crystal”.
That’s the killer for proteins, too, which is why cryo-electron microscopy (cryo-EM) has been getting huge amounts of attention in recent years. The hardware and software techniques in that field (blasting beams of electrons through the sample rather than beams of X-rays) have shown extraordinary improvements, with several no-joke breakthroughs coming right on top of each other. It’s most famous, justifiably so, for being able to deliver structures of huge proteins and protein complexes that can’t even be crystallized, by sampling a huge number of individual protein particles in different orientations and reconstructing the structure (a really impressive computational feat).
There’s another powerful cryo-EM technique, though, that was first reported in 2013: MicroED. That one uses crystals of protein, but ones that are too small to be of any possible use in X-ray diffraction. That happens quite a bit; going from a microcrystalline powder to a suitable single crystal large enough to use has long been one the slogs that protein crystallographers have to go through. Here’s more on MicroED from its discoverers, but the idea is very similar to X-ray work: you rotate the crystals, collecting diffraction data in various positions, and in fact you basically use X-ray structure software to work things up. It’s just that with electron diffraction you can do it on far, far smaller crystals (which are far, far easier to get ahold of!)
Well, what about doing this on small molecules? Would that work? Why the heck not? It’s taken five years for someone to say that, and a lot of people are probably slapping their heads this morning, but what this latest report shows is that it works even better than one might have hoped. If you take microcrystals of some small molecule and floomph them out over a cryo-EM grid (not sure that’s the technical term, but that’s how I’ve always thought of it) you can zoom in on individual particles, fire the electron beam at them, and rotate them (with modern equipment) to get the different angles you need.
The US group first tried it with some progesterone powder right out of the damn bottle off the shelf. Three minutes of data collection from a single nanocrystal (picked at random from the thousands scattered across the grid) gave a data set that produced a one-Ångstrom resolution structure: total powder-to-structure time, less than a half hour. Holy cow. The Swiss/German team reports very similar results on commercial compounds, and also seem to have been taken aback at how straightforward the procedure was. The US team went on to ransack the shelves, and out of eleven small molecule samples, none of them recrystallized for this experiment in any way, they got structures from all eleven (ten using direct methods for the data, one by molecular replacement). This includes complex natural products like brucine, shown at right, and check that resolution. That’s what small-molecule diffraction data can do for you – turn an aryl ring into ping-pong balls and toothpicks like something your kid built on a rainy afternoon. It’s worth noting that electron diffraction tends to give you positions of hydrogen atoms themselves on structures, too (as opposed to X-ray, where it’s definitely less common), since the electrons are strongly affected by those proton nuclei. (Edit: note that reference – others have actually been doing ED on small molecules, although this application to nanocrystals is likely to make it take off even more thoroughly).
Brucine’s pretty rigid, though. How about this next one, the antibiotic thiostrepton? As you can see, the electron density is a bit more raggy, but jeez, this is powder right out of the vial. That brings up another point: a lot of things that get called “amorphous powders” in organic chemistry are no such thing. As a formulations specialist will tell you, getting a truly (and reproducibly) amorphous powder can be a real pain, because a lot of the time what you have is some sort of microcrystalline powder. Tiny little crystals, of just the sort that MicroED can apparently eat for breakfast. The Swiss/German team demonstrates the structure of a large methylene blue derivative that only forms thin needle crystals that are next to useless for X-ray (a common problem!) but are just fine for MicroED. The crystals used in this paper are thousands of times smaller (!) than the ones typically used for X-ray diffraction, and at that level it’s very much worth putting things under the electron microscope and seeing if you might not have some crystals in there somewhere.
That’s what the US group went on to do: they took solids right off the rota-vap after they’d come off silica gel columns, scraped out some solid and took a look. Two of the four samples of that group gave structures, and honestly, I’d try rota-vapping down the other two from a couple of different solvents if I really wanted their structures, because you never know. How about mixtures of compounds? The different crystals therein show up as different shapes on the EM grid, and you can pick them off one by one, allowing you to get several structure determinations out of a complex mixture if the constituents have managed to form microcrystals of their own. No, really, this is great.
There’s a phrase that leads off a paragraph in the UCLA/Caltech manuscript: “Astounded by the ease with which such high quality data was obtained. . .” and I think “astounded” will go for everyone who reads it. Another line, not many sentences later, is “Based on our findings, we anticipate that MicroED will be enthusiastically received by many types of small molecule chemists“, and buddy, they’ve got that right. The Ang. Chem. paper says that this is “the technique of choice for all unsolved cases in which submicron sized crystals were the limiting factor”, and that’s the truth, too.
Just looking over the paper, I can think of plenty of great experiments, and I’m just one guy. Can you let solutions evaporate right onto an EM grid and use those? Is there an analog to anomalous dispersion, as in X-ray work, to get absolute configurations? What happens to solvates under MicroED conditions? Well, we’re going to find all of these out and more. Congratulations to everyone involved for pushing small-molecule organic structure determination into a new era!
Addendum 1: The authors will probably want to correct that line above to “data were” in the final version and fix several typos and misspelled words, but if I were writing up results like these, God knows I’d have plenty of typos, too, because my hands would be shaking!
Addendum 2: here’s Wavefunction on this paper. He seems to be just as blown away by it as I was.