When I first wrote about small-molecule structures obtained by microED (electron diffraction), I wondered if there were some way to get absolute stereochemistry out of the data (as you can with X-ray diffraction under the right conditions). Several groups have been working on just that problem, and this new paper now shows that it can be done (commentary here at Science).
As usual, I Am Not a Crystallographer, and anyone who is already can dive right into the details of how this is done. I will no doubt make caveman mistakes in my explanations. But in the X-ray world, as even us synthetic-organic, med-chem, and chem-bio types already know, it depends on anomalous diffraction. You look for diffraction spots that are pairs related by an inversion symmetry (Bijvoet pairs). If your crystal has a center of symmetry, these will be identical, but if it’s in a non-centrosymmetric (chiral) space group, they can be subtly different. Unfortunately, these differences are determined partly by the size of the atoms involved in the diffraction, and the carbons, hydrogens, oxygens, and nitrogens we’re used to aren’t up to much. That’s why people are always trying to drop in a heavy (or at least heavier) atom to help solve such problems, to make the analysis of the Bijvoet pair differences more solid. (The anomalous diffraction effect is also dependent on wavelength, and adjusting your X-ray wavelength to the sort of heavy atoms you’re targeting, which you can do with a synchrotron source, gives you an edge as well. That’s a little X-ray humor in that last line.)
Anyway, that sort of breaks down in electron diffraction, at least if you treat it as a sort of souped-up X-ray diffraction. Doing it that way has jump-started the field, since you could use the (large) suite of software tools available to process the data, although I did hear one of its practitioners mention that the first time they tried it, the software crashed because the wavelength they entered was way out of what the program would accept as an X-ray value (!) Treating the data under simple X-ray rules means that you’re assuming that only one scattering event occurs (the “kinematic” assumption), and for X-rays you can simplify your life in that manner a lot of the time (although it was realized very early on that you could see multiple events as the X-rays made their way through the crystal). That’s called dynamical scattering.
Electrons interact with the atoms in a crystal far more strongly, which is what makes the microED technique able to work on such ridiculously small crystal samples in the first place. That makes dynamic scattering much more apparent; you get much more quickly out of the range where the kinematic assumption works. In fact, all the dynamic scattering was a nuisance earlier on in the electron-diffraction field, especially because crystals of inorganic compounds were being used where it was more of a factor than ever. Finding that the kinematic assumption held when you did ED on rotating protein crystals was actually a relief, from what I can see. (In fact, trying to measure electron diffraction on a too-thick protein crystal is inviting trouble as the kinematic assumption will start to break down on you). But that dynamic scattering can give you a basis for Bijvoet pair analysis and chirality determination, since the multiple scattering events are sensitive to violation of inversion symmetry, and that’s been done for inorganic crystals.
This new paper, though, charges right into dynamic electron scattering as a way to determine absolute structure for light-atom small molecules, and it’s the first time that’s ever been done. The group had a very challenging situation: a cocrystal of sofosbuvir and L-proline, which is a very pharmaceutical crystal indeed. It had several disadvantages: the ribbon-like crystals were bent and twisted, and were very sensitive to radiation damage. The authors ended up scanning the electron beam along individual crystals, collecting diffraction patterns as they went, which put them into the same sort of regime as “serial electron crystallography“, where you hit a large number of individual nanocrystals and assemble a data set from those. The PEDT (precession electron diffraction tomography) technique (refined by some of the same authors earlier) proved to help out a lot in dealing with this situation. Rotating the sample/electron beam geometry gave more diffraction spots per shot and a better framework for matching them up in the data analysis.
So now that we know it’s possible with our new rotating electron diffractometers, we can expect even more use of microED. You’d expect it to be even less taxing on many crystals than it was here – long bent ribbon-shaped crystals are, frankly, one of the last things you’d want to deal with, given a choice. If we can (1) get both protein and small-molecule crystal structures by electron diffraction, (2) do it on microcrystals that previously would have been considered uselessly small, and (3) directly determine absolute configurations on the small molecule samples at the same time, what’s not to like?