Further applications and future developments in cryo-STET
For many of the samples studied in cryo-STET, they can be directly deposited (or grown) and frozen on cryo-EM grids. This allows the cells to be studied using light and fluorescence microscopy before cryo-fixation. This workflow is referred to as the correlative light and electron microscopy method (also known as CLEM). The correlative study can be performed on the whole sample because the cells do not require cutting before observation at the electron microscope as the thickness requirement of cryo-STET allows it. The cryo-fixation method for such samples is easily performed using plunge-freezing, where the grid is rapidly plunged inside a cryogen (usually ethane) cooled-down at about -175°C by liquid nitrogen. It must be noted that such samples are the thickest ones that can be properly frozen using plunge-freezing. Interestingly, the thickness limitation of cryo-STET is equivalent to the thickness limitation of plunge-freezing. For thicker samples, plunge-freezing can no longer be used without the formation of ice crystals. These crystals would be deleterious for imaging as they would diffract electrons which would no longer be detected on the BF detector. This means that thicker samples must be frozen at high-pressure. This cryo-fixation method, far more complex than plunge-freezing, involves an equipment which is more expansive than a simple plunge-freezer and generates a thicker layer of vitrified ice which is at least several tens of µm thick. It is possible to image this slab of frozen sample using a cryo-FIB SEM. But if cryo-STET is required (e.g. for resolution purposes since cryo-STET offers higher resolutions compared to cryo-FIB SEM), then a cryo-lamella could be prepared to thin-down the thick slab of ice down to 1 µm using a cryo-FIB. This process adds even more complexity to the realisation of the experiment.
The examples presented in this review all used BF STEM, yet this is not the only modality available. As previously briefly presented, ADF and HAADF imaging can inform about the presence of atoms heavier than the most abundant organic atoms (H, C and O) such as K, Ca, Fe. Another method, differential phase contrast (DPC) has recently been applied in biology39–41. This method uses a modified ADF detector which is composed of 4 quadrants, each representing a 90° portion of the full detector. It is then possible to split the electrons arriving on the detector in 4 different populations (top, bottom, left and right) and their analysis allows to retrieve the sample phase information, hence the name of the method. As the application of this method to biological samples is extremely recent, its successful application on thick samples still has to be demonstrated.
Current cryo-STET resolution and sample thickness limits are associated to the electron dose used during the experiment. To achieve high resolutions, the sampling (pixel size) has to be small, however a dense sampling of the specimen mechanically increases the electron dose. To achieve a two-fold resolution increase, the pixel size must be two times smaller, leading in a four-fold increase of the electron dose. Since images are already collected with the maximum electron dose the specimen can sustain, experiments aiming at resolution increases are dose-limited and must deal with the collection of low-dose, low signal-to-noise ratio images. To image thick specimens in BF STEM, the electron dose must be high as thick samples scatter electrons more than thin ones. Thus, there is a limit in the maximum achievable sample thickness which is when the electron dose becomes too high and starts generating visual damages to the sample. Sparse imaging has the potential to tackle both limitations at once. Sparse imaging consists in the collection of a limited number of pixels (as opposed to collecting the whole image pixels), which mechanically reduces the electron dose applied to the area of interest. This method is possible in STEM as the point-to-point image generation allows it. The uncollected missing pixels must be somehow reconstructed to generate the image of the area of interest. This can be performed using inpainting algorithms42. The sparse imaging + inpainting reconstruction workflow can be applied to room temperature and cryo-samples in 2D and 3D experiments43,44. By reducing the electron dose, it opens the way to higher resolution or thicker sample studies.