In the 1660s, with the help of a simple, homemade light microscope that magnified samples more than 250 times, a Dutch fabric merchant named Antoine van Leeuwenhoek became the first person to document a close-up view of bacteria, red blood cells, sperm cells, and many other scientific sights. Since then, light microscopy has solidified its place as a bedrock technique in our quest to understand living organisms. Today, it is nearly ubiquitous in life science laboratories, enabling biologists to identify and characterize cells, organs and tissues and to diagnose many diseases.
One field that light microscopy has not managed to penetrate, however, is connectomics — an area of neuroscience in which Google has made fundamental contributions over the past decade. Efforts to comprehensively map all the neurons in a region — including our previous connectomics work — have instead relied on a technique called electron microscopy, which can capture an extremely close-up view of structural information within a cell. Electron microscopy has a major limitation, however: it requires expensive, highly specialized equipment that is not readily accessible to most neuroscience labs.
Today, in collaboration with colleagues at the Institute of Science and Technology Austria (ISTA), we published in the journal Nature, “Light-microscopy based connectomic reconstruction of mammalian brain tissue”, in which we report the first-ever method for using light microscopy to comprehensively map all the neurons and their connections in a block of mouse brain tissue. We achieved this by customizing several well-established and validated techniques and combining them into a single workflow that we call LICONN (light microscopy-based connectomics). Our colleagues at ISTA led the project’s key innovation — a protocol that physically expands brain tissue while preserving structural integrity, and at the same time chemically labels all proteins in order to provide the image contrast necessary for tracing neurons and detecting other cellular structures such as synapses.
We iterated with ISTA on the details of the protocol, applying our suite of image analysis and machine learning (ML) tools for connectomics, and ultimately validating LICONN at scale by providing an automated reconstruction of a nearly one-million cubic micron volume of mouse cortex. We then comprehensively verified the traceability of all ~0.5 meters of neurites packed within a smaller volume of mouse hippocampus tissue, demonstrating that LICONN works comparably well to electron microscope–based connectomics. We also showed that LICONN unlocks the ability to simultaneously measure structural and molecular information in a tissue sample, which will enable fundamental new opportunities to understand the workings of the brain.
