3D alignment and averaging of quasi equivalent volumes extracted from tomograms

Electron tomographic studies more and more require the averaging of volumes extracted from one or more reconstructions. In case of cryo-tomography, the structures extracted are generally equivalent from both topological and geometrical point of view. In case of room-temperature tomography of native specimen, however, shrinkage and stain in-homogeneities make so that a perfect superimposition is often not possible, even for topologically equivalent complexes. These latter are difficult to align and even more to correctly averaging, often also because of their size. We have recently developed a method to reconstruct portions of tomograms rotated with respect to the original orientation, starting from patches of the original micrographs and proper orientation parameters. Using this approach and after a coarse alignment of the volumes, the average coming from all the patches of all reconstructions is performed like in “single particle” strategy. The preliminary model is then refined by use of correlation techniques. If the model is obtained by averaging the volumes in direct space, the missing wedges often combine in such a way that a missing volume in Fourier space still remains. When the reconstruction is performed via Radon transform, however, the Fourier space may be almost filled even if only a few volumes are considered. We applied our strategy to study Intra-Fagellar Transport (IFT) particles in axonemes of Chlamydomonas. Many tomograms were reconstructed in dual-axes geometry, from which 20 quasi-equivalent “short trains” of IFT particles were extracted. They were aligned by use of different strategy: 1) use of gross features, like mass center and inertia axes; 2) 3D correlation; 3) alignment of doublets by means of their coordinates extracted by segmentation. Often a combination of these approaches was used. After a few cycles of refinement, the average of the trains appeared more and more defined and some periodicities ascribable to the trains were identified. Furthermore, some differences between the trains showed a possible subdivision in two families, not previously observed.