Cryo-electron microscopy using DNA-templated protein arrays

Molecular machines, signalling molecules and structural components made from protein and RNA are some of the most important fundamental building blocks of biology. One way to study their function is to determine their structure - this can often be done by X-ray crystallography with sufficient accuracy to reveal the positions of individual atoms. For some proteins, however, it is difficult to obtain the required crystals - either because it is difficult to produce a sufficient quantity of the protein, or because the molecule is inherently difficult to crystallize. Important examples are the large class of membrane proteins which represent about 30% of all proteins and 60% of drug targets but only 1% of known structures. We are developing new techniques to facilitate structure determination based on an alternative method - cryo-electron microscopy - which requires only very small quantities of protein and no three-dimensional crystals. Molecular structures can be determined from electron microscope images of thousands of molecules embedded in a thin layer of ice. These images are averaged to improve their quality, and the three-dimensional structure of the molecule is obtained by comparing images of the molecule in different orientations - either using a sample of randomly oriented molecules or obtained by tilting the sample. The results depend critically on the quality of the sample - how flat is it? how even is the ice thickness? how many molecules can be identified? do images of neighbouring molecules interfere with each other? We are developing a new way to prepare protein samples for cryo-electron microscopy which has the potential to increase its usefulness as a tool for determining the structures of hard-to-crystallize molecules. Our method is based on the use of self-assembled lattices made from synthetic DNA that provide anchorages to bind the target molecules in a regular pattern with a spacing of around 10 nanometres. This creates a dense, flat array of non-overlapping molecules which is ideal for structure determination.We have carried out pilot studies on a membrane protein with previously unknown structure, and have shown that our techniques can achieve excellent results. We will develop this technique while studying other important proteins whose structures are unknown. Our goal is to develop a generally applicable technique that can be adopted by other groups to improve the quality of structural information available from cryo-electron microscopy and the throughput of the technique.

We have achieved 6.9A resolution (0.5 FSC) for a 43kDa asymmetric membrane protein, the neurotensin receptor NTS-1, by cryo-EM using a DNA lattice to template a dense protein array. Proteins were attached to the lattice either through the neurotensin ligand or via a NTA - His tag linkage. This is a significant achievement: NTS-1 is an order of magnitude smaller than the current state of the art for similar resolution by cryo-EM. We will optimize and develop the use of self-assembled DNA templates to facilitate structure determination for hard-to-crystallize biomolecules, including protein complexes. a) We will use the NTS-1 GPCR to optimize factors influencing resolution. DNA lattices will be used to create spatially periodic, orientationally disordered protein arrays (~10nm spacing) suitable for single-particle reconstruction. We will investigate array structure and lattice constant, the position and nature of the DNA-protein link and grid preparation protocols. We can already identify alpha helices in the reconstruction: we aim to discriminate between protein conformations with and without bound ligand and to identify the ligand binding site. We will also study ligand-induced G-protein binding to elucidate the mechanism of transmembrane signalling. b) We will apply the same techniques to the membrane-associated protein ApoL1. To obtain the first structure would be a fundamentally important result and would demonstrate that our techniques can be applied generally. c) We will use DNA-binding protein SCL to study the use of DNA templates to produce orientational as well as spatial order. Binding of SCL to its recognition sequence in the DNA array will allow the application of Fourier space reconstruction techniques. Use of a 2D crystal to obtain a 3D structure requires sample tilt, which usually has the undesirable effect of introducing variable defocus: we will investigate the design of DNA templates to introduce controllable tilt in an essentially flat array.