The Technology That Came in from the Cold, Part 2 February 11, 2000

Last week I gave some background on several technologies dealing with protein structure determination. I introduced a new technique, single particle analysis using cryo-EM. This week, I will discuss some of the uses this technology has been put to recently.

Viruses have some properties that make them especially amenable to cryo-EM. They can be huge, making them easy to see by EM but pretty hard to examine by any other approach. In addition, many viruses have highly symmetrical capsids. This means that the computer can use the internal symmetry to enhance the image. A single image can thus provide much more information than a non-symetrical protein macromolecule.

The herpes simplex virus is an extremely important infectious agent with a diameter over 1000 Å (see Zhou et al., 1999). It has 4 compartments: a core, containing the double stranded DNA genome of 152,000 bases; the capsid, a protein structures surrounding the DNA and forming the icosahedron commonly seen; the tegument, defined as proteins found in the virus after it has exited the cell but not found in virions isolated from inside the cell; and the envelope, a lipid bilayer surrounding the entire virus. Cryo-EM allowed the visualization of all four compartments. The raw data are ahown in (A) and the white bar shows just how large these particles are. The model (B) was generated by merging the data of 146 particles. At the low resolution shown, you can not only see the icosahedral shape of the capsid, but you can also make out the envelope in purple.

The tegument proteins (C) can be easily examined by comparing the cryo-EM models for the viral particles recovered from outside the cell (A) to those recovered from inside (B). Incredibly, the cryo-EM approach works much like other tomographic techniques, allowing a cross-section of the virus to be reconstructed. The DNA inside the capsid could be seen. The similarity between this and the packaging seen in small bacteriophages provided insight into how the DNA enters the preformed capsid. This approach does not simply tell you what the outside of a structure looks like. It can let you examine the inside of the structure also.

Cryo-EM can also be used to examine asymmetrical structures. I have talked a little bit about the structure of the ribosome before. This structure, made up of 3 large rRNA molecules and over 70 proteins, is the most studied large macromolecular structure. It is easy to obtain in large amounts and in present in every cell. Its large size has precluded its easy examination by NMR or X-ray crystallography. Cryo-electron microscopy has been used to examine the structuire of entire 70S macromolecule, as well as the 30S and 50S. The resulting structures have a resolution of at least 7 Å which is good enough to see secondary structure.

Ribosomal proteins that have been crystallized separately can be seen in the models determined by cryo-EM. The overall shape of the rRNAs matches what is known from biochemical approaches. Cryo-EM, though, puts them all together. Now, models of ribosomes can be constructed using other approaches. However, the power of cryo-EM is that it can look at dynamic processes that crystallization has difficulties resolving.

Cryo-EM uses aqueous solutions, so by simply adding other proteins, such as initiation factors, and comparing the resulting images to those derived without the additional proteins, the locations of the new proteins in the overall complex can be determined. In the case of the ribosome, tRNAs, elongation factors, initiation factors, pore forming proteins etc. can all be visualized, even though the proteins themselves are quite small. Cryo-EM demonstrated that the nascent chain does not exit at the side of the ribosome but uses a "tunnel" through the ribosome to exit. The importance of the shape of this tunnel is only now being examined.

Here is a Quicktime movie, giving a low resolution look at a 70S ribosmal subunit with a tRNA present. Cryo-EM demonstrated that there are not only 2 sites on the ribosomes that tRNAs bind to. It demonstrated the presence of a third site, the exit or E site, that a tRNA fills just before it exits the ribosome. An even more compelling Quicktime movie was made demonstrating what happened on the ribosome as the tRNAs move to occupy the 3 sites on the ribosome, as elongation factors are used and the amino acylation reaction occurs. Each of the important steps of the process (i.e. initiation, translation, elongation, exit) could be seen by using the proper combinations of charged tRNAs and protein factors. This simply is not possible using any other technique. Both of these movies came from Frank et al., J. Struct. Biol. 128: 15-18 (1999).

Theoretically, cryo-EM can be used to resolve proteins less than 50 kD in size. While technical problems currently prevent this, we are getting closer. A recent paper in Nature Structural Biology details the structure of black widow venom using cryo-EM. This protein is made up of 4 copies of a 130 kDa protein. In fact, the paper reports determining the structure of the 260 kDa dimer. They looked at 7000 images to produce a resolution of 14Å. This is not high enough to resolve individual amino acid residues, but you can see lots of secondary strutcures (a). Alpha helices look like sausages. The amino terminus end was found using antibodies (b). Yes, you can see where the antibodies bind in the electrom micrographs. If you have multiple epitopes, you can locate specific residues on the model. While it appears to be currently impossible to create models with 1-2 Å resolution, 5-10 Å resolution determined from 5-10,000 images looks to be the practical limit.

But, this could be overcome using the power of homology modeling and threading. These approaches try to use known protein structures and folds to generate the structure of an unknown protein. The belief is that similar sequences will produce similar structures. This seems to work quite well for small stretches of protein sequence. In the case of black widow venom, they were able to perfectly fit the atomic structure of a protein, GABPb, over the region of the venom protein that displayed the greatest homology to GABPb (see part e of this figure). Similar approaches are also being taken with higher resolution viral models generated by cryo-EM. So, by combining protein threading, which is reasonably accurate for modeling short stretches of a protein structure, with cryo-EM, a potentially useful model can be generated. It may not be atomic resolution, but, as seen above, it can still ascertain a tremendous amount of biological information.

This technique will not replace other approaches. But it serves as an incredibly useful adjunct to some of them. This can be easily demonstrated by examination of the recent models for the 50S and 70S ribosomal particles. Both of these have been solved to <9Å resolution by using X-ray crystallography, demonstrating that it IS possible to crystallize almost anything if someone has the determination. However, the crystal structures were solved using the model generated by cyro-EM to solve the phase problem. Since they knew what the structure should look like at low resolution from the cryo-EM, they could fudge a phase determination to generate an accurate structure using their crystal data. This can really speed up and simplify structure determination, not only for X-ray crystallography but for model construction using threading or even ab initio computational approaches. And in many cases, the need for atomic resolution will not be necessary. Simply knowing the general area a subunit interacts with another macromolecule will tell us a great deal.

The use of cryo-EM has tremendous potential to allow us to examine biological systems that are simply not observable by any other approach. In addition, it can work with other approaches to increase our understanding of biological systems. So, as the world heats up, there is a real possibility that many of the higher ordered structures of the cell will be determined in the cold.