Visualising proteins

Proteins are the work horses of the living cell. Either alone or in complex with other proteins and/or other biological macromolecules, proteins perform the majority of the functions that are needed to keep the cell alive. Understanding how proteins work is a strategic goal in modern molecular biology, but studying them is often challenging. Knowing the three-dimensional structure of a protein, i.e. determining the three-dimensional arrangement of all its atoms, is very useful in figuring out how it works. This is the objective of Structural Biology.

To determine a protein structure, one first needs to isolate that protein from all the other cellular components using biochemical purification techniques. Fragile inter-molecular interactions make it typically difficult to purify intact protein complexes, while different functional states may be hard to separate biochemically. Consequently, purified samples of protein complexes often suffer from various extents of non-stoichiometric complex formation and/or conformational variability. The occurrence of multiple different structures, also called structural heterogeneity, poses problems for many tools in Structural Biology. It often interferes with crystallisation and decrease the effectiveness of biophysical techniques that study assemblies in bulk solution.

Electron cryo-microscopy (cryo-EM) allows visualisation of individual protein complexes down to atomic-level details. Rapidly frozen in a thin layer of ice, these complexes are free to adopt any of their functional states. Provided that images of distinct 3D structures can be separated in the computer, cryo-EM poses less stringent requirements on sample homogeneity than alternative techniques. In principle, from a single cryo-EM sample one may therefore obtain structural information about a range of "snapshots" along the functional cycle of these protein complexes. Combining multiple snapshots into a three-dimensional movie of a functioning protein is then likely to improve our insights into its mechanism.

Our approach

Part of our research focuses on the development of image processing methods to determine protein structures from cryo-EM data. We introduced various statistical image processing algorithms to the field. We showed that a 3D maximum-likelihood classification approach can separate 2D projection images of distinct 3D conformations without the need for a priori knowledge about the structural heterogeneity in the data. We also introduced an empirical Bayesian view on the single-particle reconstruction problem that provides a convenient statistical framework, in which many aspects that were previously regarded as separate steps all come together in optimising a single, regularised likelihood function. An important advantage of this approach is that most of the corresponding parameters are estimated from the data themselves, so that user expertise is no longer required in obtaining cutting-edge reconstructions. Its implementation in the RELION software package is nowadays the most used computer program worldwide for cryo-EM structure determination.

Our methods developments are often driven by challenging structure determination projects in our own group or in collaboration with other Structural Biology groups. Through such projects, we have solved many different protein structures. One early example that is related to Alzheimer's disease is the human gamma-secretase complex, which cuts amyloid-precursor protein to form amyloid-beta peptides. Yigong Shi's group from Tsinghua University sent us purified gamma-secretase samples, and we then solved its structure. We also worked extensively with Venki Ramakrishnan on ribosomes and with Kiyoshi Nagai on spliceosomes. See our publications for other examples.

Amyloid structures in disease

Most efforts in Structural Biology are focussed on the structure of proteins in their functioning state. There are however several proteins in the humane genome that can adopt a second, malicious type of structure: amyloids. How such proteins initially switch from their healthy structure to this disease-related structure is unknown, but once a few copies of the protein have come together to form amyloid filaments, other copies of the proteins can adopt this structure much more easily too. This process is called "seeding", and it is thought to underlie the spreading of pathology in many neurodegenerative diseases.

In 2016, we teamed up with the group of Michel Goedert at the Neurobiology division of the MRC-LMB to determine the structure of amyloid filaments that are formed by the Tau protein in Alzheimer's disease. Michel had previously identified Tau as the protein that forms amyloid filaments inside neuronal inclusions in Alzheimer's disease. Together with Tony Crowther, also at MRC-LMB, they had also visualised these filaments at low-resolution using negative-stain electron microscopy. Thanks to the adaptation of RELION for helical procesing, we could use cryo-EM imaging to solve the atomic structures of Tau filaments that were extracted from the brain of an individual with Alzheimer's disease.

However, Alzheimer's disease is just one of many so-called Tauopathies, a family of neurodegenerative diseases which are characterised by the presence of Tau filaments. Other examples are, Pick's disease, chronic traumatic encephalopathy (CTE), posterior supranuclear palsy (PSP), and cortico-basal degeneration (CBD). Other neurodegenerative diseases are characterised by the formation of filaments of alpha-synuclein (e.g. Parksinson's disease) or TDP-43 (e.g. amyotrophic lateral sclerosis). Together with Michel, we have since solved atomic structures of Tau and alpha-synuclein filaments from most of these diseases. In the mean time, Benjamin Ryskeldi-Falcon, who is also at LMB, has used our methods to solve structures of TDP-43 filaments. Surprisingly, we have found that specific amyloid structures of tau, alpha-synuclein and TDP-43 define different diseases, while different patients with the same disease always seem to have the same structures. Currently, we are trying to understand what are the factors of this remarkable specificity of amyloid structure in each disease. Hopefully, in the longer term, this will lead to improved understanding of filament formation and seeding in neurodegenerative disease, which may then lead to insights into the design of medical interventions.