Three-dimensional transmission electron microscopy and its application to mitosis research

Abstract

Transmission electron microscopy produces images that are projections of the original object, with the consequence that features from different depths of the specimen overlap and give a confusing image. This problem is overcome by reconstructing the object in 3D from a series of 2D views using either serial thin section reconstruction or electron tomography. In the serial section approach, the series of 2D views is generated from images of successive serial sections cut thin enough to be effectively 2D slices of the specimen. For electron tomography the series of 2D views is generated by tilting a single, usually thicker, section in the electron beam. Resolution in the depth dimension is limited to twice the section thickness for serial section reconstruction and is determined by the number of tilt views collected (i.e., by the fineness of the angular interval between successive tilt views) for electron tomography. Both methods produce distorted 3D reconstructions because of missing material and alignment difficulties in the case of serial sections and the limited angular tilt range in the case of electron tomography. However, techniques have evolved for minimizing and circumventing these distortions and, as long as the user is aware of the limitations, misinterpretations can be avoided. Since electron tomography provides better resolution (generally 5-20 nm), it is the method of choice for determining detailed structural interactions such as the depth of kinetochore MT penetration into the kinetochore outer plate. On the other hand, serial section reconstruction is more effective for projects that require tracking through a complete object in the specimen, such as counting the number of kinetochore MTs on each kinetochore. If the project requires finding a relatively small object in a large specimen (e.g., finding centrioles in an oocyte), then it is sometimes advantageous to cut thicker plastic sections and analyze them via stereo viewing. The mitotic spindle, however, is generally too complex to be analyzed via stereo viewing. Currently, collapse of plastic sections in the electron beam limits the utility of serial section electron tomography. Once a 3D reconstruction is completed it must be analyzed with the 2D medium of the screen on a computer monitor. The easiest approach is usually to walk through the 3D reconstruction volume slice by slice. However, in order to appreciate 3D interactions, and to communicate the results to others, it is generally necessary to segment key components from the rest of the volume and use modeling and rendering techniques. Rendered surface views can easily be color coded and provided with a number of depth cues to simulate the surface viewing encountered in everyday life. In some instances, it is useful to look through a smaller portion of the reconstruction volume with "X-ray vision." This can accomplished by using volume rendering to create a series of semitransparent views from different tilt angles.