Over the first two years of the project we have advanced towards all three main research objectives. Firstly, we have studied insect brains at the level of entire brains. To that aim, we have obtained samples of over 50 insect species from across the phylogenetic tree and have used X-ray based imaging to obtain 3D image stacks of the brains of a subset of those. We used these data to optimize our sample preparation and scanning procedures and have started to quantitatively analyze the different regions comprising their brains, illuminating evolutionary trends at the level of entire brains, covering over 450 million years of evolution. Second, we have pioneered a novel approach in comparative connectomics to advance our understanding of how neural circuits evolve. For that purpose we have established an imaging regime to obtain multi resolution 3D electron microscopical image volumes. Five large insect species have already been imaged in this way, representing the first such data outside of the fruit fly Drosophila. To analyze the many terabytes of image data resulting from these scans, we have developed a new image processing and analysis pipeline. This pipeline combines complex image alignment protocols with subsequent extraction of neural morphologies. These are either manually traced as neural skeletons (in low resolution overview data stacks) or via machine learning based image segmentation (in synaptic resolution data). Together with automatically detected synaptic contacts, we combine all neural data into synaptic level connectivity graphs, i.e. connectomes. This has been completed for the neural circuits that encode head direction in bees and revealed surprising levels of conservation across hundreds of million of years of evolution. Importantly, we have also identified highly evolvable circuit components - evolutionary hotspots - as well as completely unique circuits that correlate with bee’s complex navigational abilities. Both now serve as foundation for the third line of enquiry, namely attaching functional significance to neural circuit data. This is achieved in three ways: Firstly by translating the circuit data directly into computational models of the brain. Second, by probing the identified neurons with electrophysiological methods, and third by directly observing and quantifying the behavioral abilities of our insect species. In the first two years of this project we have generated an overarching model of the brain region of interest (the central complex), that is now used as foundation to generate species specific models in which to directly generate predictions for links between circuits and behavior. Behaviorally we have developed novel approaches to quantitatively dissect insect behavior using walking arenas of different sizes in a controlled laboratory based setting. Filming their behavior and using machine learning to extract the fine structure of movements and their movements strategies we can now compare diverse species with reproducible methods. This way we have identified complex navigation abilities in walking bumblebees that perfectly reflect specialization identified at the level of connectomes. Those are first examples of how specializations in neural machinery could generate novel behavioral traits and showcases that the strategy pursued by this project can indeed deliver groundbreaking results.
