Electron cryomicroscopy
reaches landmark molecular resolution

by Eric Drexler on 2010/10/17

Electron microscopes can image biological macromolecules in cryogenic ice, but it shows them as low-contrast features in a grainy image (see below). Using enough electrons to reduce the graininess would first destroy the specimen.

The trick to getting enough information without frying the molecules is to image many specimens that are known to be identical, and to somehow find, align, and combine data from their images (keep in mind that these are grainy, 2D shadows of randomly oriented 3D objects). The quality of the resulting reconstruction depends heavily on the quality of the methods, and the methods are not simple.

A new report in Nature describes improved methods and a result (below) that sharpens resolution from the previous 1.3 nm to 0.66 nm. This provides about 8 times as many voxels, and gave the authors enough information to infer protein secondary structures and build a reliable atomic model of F-actin:


Sample of raw electron cryomicroscopy data, F-actin

Raw electron cryomicroscopy images,
F-actin helical assembly.


CryoEM reconstruction of F-actin

Reconstruction at 0.66 nm resolution.
(Stereo view of density isosurface with fitted C? ribbon diagram.)


“Direct visualization of secondary structures of F-actin by electron cryomicroscopy”,
Takashi Fujii, et al., Nature 467:724–728 (2010).

The authors’ method is apparently quite practical. From the article:

We have demonstrated that our cryoEM technologies now allow us to visualize the secondary structures of such thin objects as F-actin in a few days of work, including image data collection and processing and 3D image reconstruction.

In closing, they remark that

There is also room for further improvement, to reach atomic resolution. The present work offers a new opportunity to look into cellular mechanisms essential for the activities of life.

And in building nonbiological molecular machinery, of course, it’s useful to be able to see the result.


See also:

{ 7 comments… read them below or add one }

Alexander October 19, 2010 at 12:37 pm UTC

Sounds like a significant advance forward in atomic resolution imaging of molecules. Does the new method still require imaging of multiple specimens or is it able to handle just one sample?

I don’t know too much about imaging techniques but are electrons the only available way to do atomic imaging?


Hi Alexander —
For fragile biological materials, the radiation-damage limit on electron exposure achieving atomic resolution difficult, but current methods already require integration of data from many molecules of the same structure. This could do the job, in principle, provided that data from the fuzzy individual images can be aligned well enough to sum up the high-spatial-frequency diffraction signal.

It’s important to keep in mind, however, that the step from resolving 3 nm features to resolving atoms means resolving ~1000 times as many features per unit volume. The minimum necessary x-ray energy is also higher.

I should also note that “aligned well enough” is somewhat misleading shorthand: The 3D alignment of each object must be inferred before the image data can be combined, and this is a computational problem of a different order from merely aligning and superimposing 2D images.

Attaching specimen molecules to (for example) atomically precise, electron-dense nanoparticles, metal clusters, or even functional groups, if done in an atomically precise way, could potentially be a substantial aid in refining alignment.

BTW, the leading method for atomic-resolution characterization of molecular structure is based on diffraction of x-rays from molecular crystals. It’s growing the crystals that’s the problem.
— Eric

GK October 20, 2010 at 2:56 am UTC

Eric:

Do you think that magnetic resonance force microscopy MRFM holds promise for atomic scale resolution? J. Siddles and the University of Washington Quantum System Engineering group has done interesting research in this area: http://courses.washington.edu/goodall/MRFM/

I’ve followed this area for over a decade and they seem to me to have been making slow but gradual progress.


Researchers in the field seem cautious about claiming that atomic resolution can be achieved, but they’re working toward that goal. On the whole, I’m optimistic about it.
— Eric

Eniac October 21, 2010 at 2:15 am UTC

@Alexander: This method by its nature needs to accumulate data over many copies of the molecule, similar to crystallography.

Like GK, I have been fascinated by MRFM for a decade, and think that it has the unique promise to reach atomic resolution in individual molecules, rather than aggregates. See also:

http://www.pnas.org/content/106/5/1313.full

GK October 21, 2010 at 4:54 pm UTC

@Alexander: X-ray wavelength photons (X ray crystallography) and nuclear magnetic resonance imaging are the most common non-electron imaging techniques. NMR has problems with molecules above a certain size. MRFM is a sort of NMR in reverse, without this size limitation, but cannot presently achieve individual atom resolution. Crystallography requires forming molecules into crystal.

@Eniac (cool name BTW)
I agree that this seems similar to crystallography in that multiple molecules are required. But if I’m reading it right, it does not require that the molecules in the same sample crystallize so that x rays scatter and diffract in a manner so that we can determine the structure of the protein. If I’m reading it right, the electron beam determines some of the structure of a single molecule, before destroying it, then moves on to the next molecule. So long as we know that all molecules in the sample are the same, we can eventually determine the entire structure. The protein does not have to able to form a crystal.


Hi GK —
Re. MRFM: For readers starting at the top of this thread, I should note that “MRFM” is magnetic resonance force microscopy, yet another application of microscale cantilevers to nanoscale measurement and imaging. There’s a recent review here [pdf].

Re. electron cryomicroscopy: Yes, there’s no need for crystallization, and this is a critical advantage, since many macromolecules have resisted crystallization. Each of a series of image fields covers a substantial area (and multiple molecules), as in a conventional light microscope. The image gives information about coherent electron scattering, and the “exposure time” ends when radiation damage becomes excessive. I’d expect that low resolution image data (suitable for determining molecular orientation) could be collected over a somewhat longer exposure time, because the diffraction signal at lower spatial frequencies degrades more slowly. I don’t know whether this is part of current practice, however.
— Eric

Eniac October 25, 2010 at 5:53 pm UTC

Yes, what is cool about electron cryomicroscopy is that you don’t need a 3-d crystal. Instead of physically arraying the molecules, you image them individually, and then “array” them in-silico. But you do need an array of some sort, because single molecules cannot yield enough data. That’s why I think MRFM is so much cooler, provided it works out to angstrom resolution in practice.

Incidentally, electron cryomicroscopy got started with 2d-arrays (notably the solution of the structure of bacteriorhodopsin as arrayed in purple membrane), and the application to randomly oriented molecules was developed later. I suspect, but don’t know, that the present work on fibers uses the knowledge of fiber structure, so that you really do have a physical 1-d array here, but the method can be applied to randomly oriented molecules, too.


Yes, imaging structures that are organized as fibers can help greatly with the crucial alignment step.

Researchers in this area might want to pursue ways to bind other molecules to fibers. Using in-vitro directed evolution to develop proteins that bind other proteins is routine, and using these methods to bind proteins to fiber-forming proteins should be quite practical. For the protein being evolved, the main requirement is high stability, since their frameworks must tolerate extensive mutagenesis directed to a surface patch. Note that there are ways to use protein engineering and directed evolution to increase stability, too.

A range of phage and cytoskeletal structures offer potential starting points.

— Eric

GK October 26, 2010 at 4:00 am UTC

It seems to me that MRFM could be one of those technologies subject to positive feed back loops. Better nanotechnology (or microfabrication) leads to a smaller more precise cantilever, leads to better nanotechnology, leads to a smaller more precise cantilever…….

Of course we realistically can’t predict bottlenecks or rates of progress.


Yes, I’d describe it as part of a large, multi-component feedback loop that runs through this and a range of other technologies for modeling, design, sensing, and fabrication. Some of the feedback runs through direct scientific and technological enablement; some runs through the recruitment of people and funding, and the infrastructure of laboratories and organizations. This adds up. (Or should that read “multiplies up”?)

Thomas October 29, 2010 at 12:18 pm UTC

I thought STMs/AFMs were able give atomic resolution molecules?


In some instances, but not with big, soft, 3D macromolecules. — Eric

{ 1 trackback }

Leave a Comment

Previous post:

Next post: