CERN Accelerating science

Electron Cryo-Microscopy

What is the structure of large molecules? How do they function?

A primary goal in structural biology is to elucidate the structure of large biological macro-molecules to atomic resolution; this is expected to increase our understanding of how they function. The traditional route to such structure determination is to make good three dimensional crystals and then use X-ray crystallography and synchrotron radiation. However, the electron cryo-microscopy (cryo-EM) of isolated individual particles provides an alternative for a large class of molecules, such as membrane proteins, which are difficult or impossible to crystallize. The two techniques, namely cryo-EM and X-ray crystallography, are increasingly used in a complementary manner: a ‘modified’ large molecule, originally studied with X-rays, can be more conveniently studied post-modification by cryo-EM. This eliminates the need to re-crystallise the ‘large’ modified molecule followed by a lengthy process of data collection and analysis.

How does electron cryo-microscopy work?

To reduce radiation damage, molecules are imaged at low temperatures, usually in a frozen hydrated state, embedded in vitreous ice. Proteins and nucleic acids have poor contrast and the electron dose is limited by radiation damage. Consequently the images are quite noisy. Due to radiation damage it is not possible to improve the signal-to-noise ratio of individual ‘noisy’ images so extreme averaging is necessary. The use of detectors with the highest possible detection efficiency, capable of counting every electron without adding extra noise (i.e. having a high detective quantum efficiency DQE) allows the structure to be determined using fewer average particle images and on smaller particles. For non-radiation sensitive specimen (in materials science, for example) it is possible to use considerably higher radiation dose to get improved signal-to-noise ratios.

Another reason for obtaining relatively poor resolution in images is blurring due to beam-induced movement of the specimen during imaging causing image blurring. The precise causes of the beam-sample interaction are not well understood but it is clear that even extremely small amounts of radiation are sufficient to cause specimen movement. A very successful approach to alleviate this problem is to divide the exposure time into multiple frames of shorter exposure time, created in effect a movie in which later computer processing can be used to de-blur the final image.

What are the detectors requirements?

Detector requirements for cryo-EM have several unique features, which make them different from other radiation detectors. It was found, after considerable experimentation, that monolithic active pixel sensors (MAPS) were the most suitable detectors for this work.  

Cryo-EM using ~300 keV electrons produces higher quality images with fewer aberrations. A high detection efficiency for imaging fine features in the specimen is most important, i.e. we need detectors with high detective quantum efficiency as a function of spatial frequency. For 300 keV electrons it is essential for the detector to be backthinned to 50 µm or less to avoid deterioration of DQE(ω) and consequently resolution. Since the detectors record the image directly, they accumulate significant amounts of integrated dose over time, hence, for a detector to be useful it must be able to operate over periods of ~2 years without suffering from radiation damage and this would translate into a dose of ~20 Mrads.

The detector specifications can be summarised below:
Size                                         8kx8k pixels
Pixel size                                 ~ 10 µm square
Detector thickness                  ~30 µm
Radiation tolerance     `           20 Mrads (electrons only)
Framing Speed                        1000 Frames/sec
Readout                                   capable of digital and analog modes
High DQE as a function of spatial frequency

What detectors are used in electron cryo-microscopy?

One reason for obtaining relatively poor resolution in images is due to beam-induced movement of the specimen during imaging causing image blurring. A very successful approach to alleviate this problem has been adopted by Bai et al (Bai et al. eLife 2013), who divided a 1 second exposure (of 80S ribosome) into 16 sub-frames. It was assumed, correctly, that the specimen movement would be reduced considerably in the shorter exposure time. The particles from successive frames were aligned before carrying out the averaging procedure. This procedure was very successful and the authors have obtained a 3.5Å resolution map of the 80S ribosome particle using only ~30,000 particles; previous work, where data was recorded on film, needed ~1 million particles of 80S ribosome to attain a resolution of 5.5 Å. Another example showing the advantage of Falcon II’s (made by the FEI Company ) higher DQE(ω) compared to film is the recent structural work on the three dimensional structure of beta-galactosidase (Chen et al. Ultramicroscopy, HR Noise-Substitution, 2013) along with a comparison with atomic model (see figure). The higher DQE(ω) of the backthinned Falcon II allows higher resolution data with fewer single particles. Film acquired data required 49000 particles to attain 11 Å resolution while with Falcon II, only 43000 particles gave 6 Å resolution.

Both fast readout, to record movie mode and high DQE were needed for this approach provided by the newly developed backthinned CMOS detector, Falcon II, for data collection (N Guerrini 2011). The detector is in commercial production by the FEI Company in the Netherlands and is being supplied in their higher end electron microscopes.