Ultrastructural visualisation of synaptic function in brains of behaving mice

The basic function of the brain is to process information to trigger action: receiving sensory input and integrating it with prior experience to generate appropriate responses. The information is encoded in the form of small electrical activity signals that are passed between specialized cells (neurons) wired up together to form circuits. To understand how neurons are able to compute their responses - the essence of a working brain - we need to know two key things: (1) how neurons are connected together (the 'wiring pattern'), and (2) what signals are occurring at the specialized neuronal connection points, called synapses, as an animal carries out behaviours. It is essential to understand both elements: identify "who is talking to who", and also identify which synapses are active and how strong they are - in order to understand how the brain works.

When examining the large and richly interconnected networks in mammalian brains, solving these problems is a major challenge. This has severely limited our understanding of brain operation. Recently, researchers have found a way to address the problem of identifying synaptic connections between neurons using a special type of electron microscope which allows a target brain region to be reconstructed in three-dimensional detail down to nanometre resolution, an approach called 3D-EM. When combined with powerful computational analysis approaches, it then becomes possible to map out the neuronal connections in a circuit and therefore reveal the wiring diagram. What is still missing, however is the functional information at the synapses - their strength and pattern of activation - that is essential for a full understanding of circuit operation.

The aim of this project is to address problems (1) and (2) in parallel by developing state-of-the-art approaches to provide us with a revolutionary new way to read out synaptic activity and strength using 3D-EM. We will apply our methods in the planned work to generate, for the first time, functional maps of synaptic activity overlaid onto the wiring diagram of the same circuit as an animal processes sensory (visual) inputs and performs complex behaviours. In our pilot experiments we have already shown that our technique can be used to reliably identify synapses and estimate their strength. We will optimise this strategy and combine it with powerful new machine-learning technologies for computer-based image analysis which will be developed as part of the research program. This will permit an automated analysis of tens of thousands of structures in a 3D brain tissue volume that is both much faster and yields better accuracy and reproducibility than is achievable by expert humans.

These ground-breaking new methodologies should give us fundamental insights into the relationship between the function of individual synapses and behaviour - a holy grail in the field of systems neuroscience. In the future, our unique methodology may also be used to examine the disorders in information signalling that occur in neurodegenerative diseases, offering potential targets for therapeutics. The findings and the topic are very well-aligned with current BBSRC initiatives including the Research and Innovation Priority, 'Advancing the frontiers of bioscience discovery', which includes 'Understanding the rules of life' and 'Transformative technologies' as two of its principal aims. Our core objectives are also directly relevant to the BBSRC responsive mode priorities 'Data-driven biology', 'Technology development for the biosciences' and 'Systems approaches to the biosciences'.

Synapses are key nodes for information flow and memory storage in the mammalian brain. Detecting patterns of activation and strength of individual synapses and understanding how these parameters change with experience is essential to fully comprehend the function of neural circuits. Although high-throughput electron microscopy now allows us to identify connectivity patterns in networks, the resulting wiring diagrams still lack crucial information: the strength of individual synapses, and the pattern of synaptic activation. We will address these challenges by developing and harnessing a new strategy for ultrastructural readout of synaptic function in intact mammalian brain. This involves labelling activated synapses in vivo in a behaving animal followed by focused ion beam scanning electron microscopy (FIBSEM) to yield 3-dimensional reconstructions of activated synapses at nanoscale resolution. This approach will be benchmarked and validated using "all-optical" experiments where neurons are selectively activated in precise spatial and temporal patterns using 2-photon optogenetics in vivo. To provide a robust quantitative framework, we will harness a machine learning-based approach to identify synapses and vesicles. This will allow us to automatically assign synapse type (excitatory/inhibitory), quantify details of synaptic structure, and predict release probability with superhuman performance. We will harness this strategy to attack three fundamental problems in systems neuroscience: (1) the functional balance of excitatory and inhibitory synapses on individual dendritic branches; (2) the synaptic mechanisms underlying detection of sparse activity in sensory cortex; (3) the synaptic engram underlying memory-guided spatial navigation in hippocampus. Our strategy will revolutionize connectomics, providing a readout of synaptic activity and synaptic strength that offers fundamental new insights into how synaptic function drives behaviour in the mammalian brain.