Skip to Content

Protein Structures Without Crystallization

Here’s a report from a team at the University of Toronto of a new way to get protein structures – always a welcome development. (There a summary at Technology Review here). It’s another application of electron cryomicroscopy, which was the subject of a blog post here about a year and half back.
As with many cryo-EM applications, this one is done on a thin film. What you get when you hit individual molecules/particles in this film with a beam of electrons is a series of pictures formed by the electron density of the target – but you don’t know how all those target molecules are oriented. So if you want to assemble the three-dimensional structure that gave you all these snapshots, you’re faced with a massive computational problem, one that’s full of unproductive local minima ready to lead you astray, and one that requires some pretty big assumptions just to get started.
What this new paper offers is what looks like a far more efficient way to get a handle on that problem. They have a new way to increase the signal/noise of the data, which is welcome, because you tend to have to use low-intensity electron beams for fear of demolishing your samples. But the biggest innovation seems to be the realization that protein molecules tend, when frozen into thin films, to lie on along their long axis. Getting rid of the other possible looking-down-the-long-axis poses speeds things up considerably, and avoids some potholes as well. Another key advantage to their method is that it doesn’t require you (by luck or talent) to have picked a good initialization.
My own take on this is that it’s going to be most useful when combined with some modeling (and whatever X-ray data might exist, of course). The hope would be that you can fit the protein sequence to the three-dimensional shape that you obtain from the cryo-EM technique. The shapes that they’re pulling out look like they’ll be pretty useful for this purpose, and I look forward to seeing how general the technique turns out to be.

3 comments on “Protein Structures Without Crystallization”

  1. Anonymous BMS Researcher says:

    Some great stuff has been done with cryo-em. Also some wildly incorrect models. Powerful tool when used wisely.

  2. Matthew K says:

    A learned friend who works in NMR and protein structure tells me that conventional cryo-EM is not all it’s cracked up to be – for one thing the high energy of the bombarding electrons imparts thermal motion to the proteins, blurring the outlines over time. Direct electron detection and new software methods have allowed much more rapid imaging to work around this, and the result is near-atomic resolution. I’ve linked an eLife overview in the comment header for anyone interested.

  3. Marcus says:

    Just wanted to clarify that the MIT Tech Review had an error which was reproduced here. The method does not rely on particles lying on their sides; it can and does work for particles with uniform viewing distributions.
    Cryo-EM used to be restricted to low-res structures. Hardware and software advances have dramatically improved the resolution and it’s still getting better.
    Finally, yes, there were some incorrect structures published based off of Cryo-EM experiments due to poor initial model selection. However this is one of the major advantages of the approach described in the linked article. There is no initial model required, hence there is almost no chance of determining an incorrect structure.

Comments are closed.