There was apparently a very impressive talk from Sriram Subramaniam on cryo-electron microscopy (cryo-EM) at the Computer-Assisted Drug Design Gordon Conference, and I can well believe it. That field has grown tremendously in capabilities in recent years, and is producing some startling results – and those results are coming faster all the time. Just recently, we’ve had a structure of the vascular protein sorting 4 complex, the Hrd1 channel, tau filaments from Alzheimer’s tissue, an entire 70S ribosome from a mycobacterium, Type I CRISPR enzymes in action, DNA protein kinase, and more. These are at varying resolutions, to be sure, but those resolutions are getting finer and finer, and the best cryo-EM structures are excellent.
One big advantage of this technique over traditional X-ray crystallography is, of course, that you don’t need to crystallize anything. Pure protein, frozen on a surface, is enough. Now there’s a lot of work contained in that last sentence – the various techniques for sample preparation, data collection, and data analysis are nontrivial, particularly for really high-resolution structures, but not very long ago they couldn’t even be described as that (and they’re improving constantly). Structures that were once the domain of the cutting-edge labs are being produced at more and more institutions, and the cutting-edge stuff is moving on to even more impressive results. This has been due to improvements both in electron sources and on the detector end, with direct electron detection being the big change in the latter. New techniques such as the Volta phase plate are in active development to improve contrast and resolution. The ability to handle large amounts of data has also been crucial, since cryo-EM structures are produced from a great number of particles that have landed in all sorts of orientations.
Another big advantage is that electron microscopy doesn’t have the “phase problem” that X-ray does. X-ray detectors can give you the amplitude of the scattered X-ray beam, wherever the spots show up, but they can’t tell you the phase – that’s been lost. The problem is that the phases are very important to the structure determination, so a number of ingenious methods have been worked out to deal with this. It’s really not a problem for small molecules, in general, and hasn’t been for a long time thanks to “direct methods“, but it’s definitely a pain for large proteins. That’s why (for example) you see X-ray crystallographers working heavy atoms into their proteins – it’s a particularly direct way (anomalous dispersion) to get a handle on reconstructing the phase information, if you pick the right X-ray wavelengths for the heavy atom you’re using. (Fortunately, synchrotron X-ray sources allow you to do just that). This lets you determine the position of the heavy atoms using the sorts of direct methods that can be used on small molecule crystals, and that often allows the rest of the protein structure to start falling into place rapidly, with modern software. The electron beam information, though, has the amplitudes with the phases still available, an advantage first realized by electron crystallography pioneer Aaron Klug. Those amplitudes, in general, still can’t be determined as precisely as they can in X-ray work, but having the phase information available up front more than makes up for that. It allows for that processing of all the different orientations, mentioned above – without phase information that would be a nightmare in all directions.
All this makes a person wonder if eventually this technique could take over from macromolecular X-ray crystallography in general – and if so, what definition we’re using for “eventually”. I know that many X-ray specialists in that field have, in recent years, been vigorously brushing up on their electron microscopy skills, with just that thought in mind. The problem that X-ray has with proteins is what’s it’s always been: growing crystals. It’s a black art, if ever there was such a thing, and a glance through the catalogs that supply workers in the field will confirm that. You can buy collection after collection of buffers and additives to try to convince recalcitrant proteins to grow crystals, and a look into any lab working in the area will show you stacks and stacks of 96-well plates trying these combinations out one after the other. Miniaturization and automation have certainly helped that process, but they’re still in the service of making the trial-and-error process run faster. Getting rid of the trial and error entirely has been beyond human capability, so far.
Cryo-EM has its own trial-and-error thing going, but it already seems to be in better shape in that department that crystal growing is, and it’s improving much more rapidly (from what I can see). Add in the number of proteins that have just never yielded to crystallization at all, and the opportunities for multiprotein complexes, etc., and the future looks pretty bright on the electron side of things. Is my outsider’s view of things accurate – are the electrons overtaking the X-rays? Or are there factors I haven’t considered? Comments welcome. . .
Update: added a bit more on the phase problem, etc.