MAPPING THE BRAIN: Sub-100nm resolution, large volume X-ray connectomics with near-field multislice ptychography

Mapping the thousands of connections between individual neurons in the brain, a field called connectomics, is critical to our understanding of the mechanisms behind neurodegenerative conditions, such as autism and schizophrenia, and the brain's complex responses to stimuli, such as images, smells and sounds.

For many decades, electron microscopy (EM) has been the dominant imaging technique in connectomics, and recent advances in EM methods now enable 3D imaging of regions of the brain up to several hundred micrometres cubed in volume. This is sufficient to capture the entire nervous systems of invertebrates or small vertebrate animals, such as larvae. Unfortunately, these EM techniques work by shaving very thin slices from the sample to be imaged one by one. This means they require months to years of data acquisition and processing times, whilst the slicing process itself is destructive and highly error-prone. Thus, while providing the highest resolution, using EM alone it is difficult to obtain wider contextual information, such as the identity of neurons that are linked together by the synapses visible in the EM 3D images. This project aims to develop a bridging method that can provide correlative, non-destructive imaging of brain tissue at sub-100nm resolution, to target and contextualise EM connectomics.

Advanced forms of synchrotron X-ray microscopy already go some way toward providing this contextual information, and advantageously, the penetrative power of X-rays means these methods can image large sample volumes quickly and without destructive slicing; the problem is that sample volume and resolution must trade off against one another - larger, thicker samples scatter the X-ray beam leading to a rapid falloff in image resolution. Current state-of-the-art X-ray microscopy can achieve a resolution of approximately 100nm over a 200 micron thick sample; this project will develop a new 3D X-ray tool to image brain tissue at sub-100nm resolution over a cubic millimetre volume.

Until recently, the ideas we explore in this project would have been impossible given the computing resources required. Today however, phenomenal advances in computer hardware, especially parallel computing on Graphic Processing Units (GPUs), mean processes that required many hours to run a decade ago are now possible in close to real time. This is transforming the way Researchers think about the role and potential of computing in microscopy. Our work in this project is based on one such transformative technique called ptychography, which uses iterative algorithms to reconstruct an image of an object from diffraction data captured by a very simple, lens-free optical system. Essentially ptychography replaces the lenses in an X-ray microscope with code.

The field of ptychography has grown exponentially over the past decade and dedicated ptychography beamlines are now coming online at most synchrotrons around the world. The UK is at the forefront of this research, with a strong track record in algorithm development and novel experimental approaches. Our project will complement these on-going efforts and ensure ptychography remains an active, competitive topic within the UK, and that the UK remains a world-leader in this exciting and rapidly growing field.

Our Programme brings together Sheffield University, the Diamond Light Source and the Crick Institute in a new and exciting collaboration. The Investigative team holds expertise at every step of the technique development journey, from optical bench proof of principle, through implementation at the synchrotron to cutting edge, high impact application studies in collaboration with the brain specialists at the Crick Institute.

For many decades, electron microscopy (EM) has been the dominant technique giving access to ultrastructural information in tissues and cells. In neuroscience in particular, EM is at present the only technique to densely identify synapses, and recent advances in volume electron microscopy now enable volumetric imaging of up to several hundred micrometres in three dimensions. However, these approaches are in general still limited to small parts of brain regions, remain hugely time consuming, rarely exceed volumes of 0.005 mm^3, and involve physical sectioning steps that are destructive and highly error-prone.

Thus, while providing the highest resolution ultrastructural information, it is difficult to obtain context information for EM, such as the identity of neurons linked by synapses: bridging methods are required to provide correlative, non-destructive imaging at the <100nm level, to target and contextualise EM analysis.

Advanced forms of synchrotron X-ray microscopy can image large sample volumes at the required resolution both quickly and without destructive slicing. The problem is that sample volume and resolution must trade off against one another - larger, thicker samples scatter the X-ray beam leading to a rapid falloff in image resolution. This project aims to develop a new 3D X-ray technique that avoids this trade off, to image brain tissue at <100nm resolution over a cubic millimetre volume. The new method will combine for the first time near-field ptychography and multislice ptychography in order to handle multiple scattering within thick samples, producing high contrast phase images which will be used to target subsequent correlative EM analysis.