Understanding brain function requires knowledge about how information flows across the different parts of the brain. Since the early days of neuronal reconstructions by Ramón y Cajal, observations about the structure of neurons and their axonal arborizations or collateral maps have driven hypotheses regarding the function of neural circuits. Thus, examination of the nervous system as a collection of “cell-types” has fueled neuroscience for more than a century. Furthermore, this approach has gained considerable prominence in recent years, as it has become apparent that neuronal diversity is even greater than previously appreciated. Classes of neurons that had previously been treated as monolithic, can now be subdivided on the basis of their developmental history and/or gene expression profiles, physiology or morphology. Importantly, even neurons that appear similar in many ways can differ greatly in their local and long-range axonal projections. To examine the relationship of axon collaterals among neurons that share other properties, it is important to be able to follow the complete collateral maps of multiple neurons within the same brain.
For instance in a sparse group of neurons labeled based on their projections to a particular brain area, what is the relationship between collateral maps within this group? How diverse are the collateral maps of a subset of neurons in one area of the brain, all activated during a particular behavior? To answer questions like these, it is necessary to label multiple neurons in a brain and subsequently examine the axon collaterals of individual neurons in the group. To do so, it is desirable to image the entire brain at a resolution sufficient to resolve and trace individual axons.
Examining axon collateral maps in this way could reveal details of connectivity in the mouse brain much more rapidly and inexpensively than other methods, such as EM connectomics.
Our project takes advantage of a new imaging platform developed at Janelia. The MouseLight microscope is based on serial two photon tomography (S2PT; Ragan et al., 2012). It is capable of fast volumetric imaging of entire mouse brains at sufficiently high resolution needed to reconstruct axons of individual neurons projecting across long distances. We have developed strategies to brightly label sparse subsets of neurons and efficiently clear brain tissue for high signal-to-noise imaging. We have also implemented a computational pipeline for stitching, assembling, visualizing and analyzing these large datasets. Multiple brains have been imaged and complete reconstructions of small populations of neurons in the motor cortex have been achieved. A major part of our current effort is to increase throughput of data mining by automation of the annotation process. This is a significant bottleneck that needs to be overcome in order to achieve our goal to reconstruct thousands of neurons in the mouse brain.