Towards large-scale neural circuit mapping and analysis using electron microscopy

One of the great scientific challenges of our time is the reverse-engineering of the algorithmic principles operating in nervous systems. Progress in our understanding of these principles requires the elucidation of the structure of neural circuits. The structure of neural circuits can be obtained by mapping the morphology of neurons and annotating their synaptic connections using electron microscopy (EM) methods. Large EM image volumes of neural tissue can be generated routinely at nanometer resolution with automated acquisition methods. However, the size of these volumes and the complexity of neuronal arbors renders the extraction and analysis of neural circuits slow and tedious. This represents a major bottleneck for the analysis of neural circuits. To tackle this bottleneck, we developed a web-based open-source software for the mapping, analysis, and visualization of neural circuits (www.catmaid.org). This software enables fast, accurate, and collaborative mapping of neural circuits of interest by globally distributed groups of researchers. The software implements a novel iterative, non-redundant circuit mapping approach. This approach was validated by mapping a subset of neurons in a proprio-motor circuit of the Drosophila melanogaster larval ventral nerve cord. We compared the reconstruction speed and accuracy from our approach to state-of-the-art, redundant methods. Results yielded similar levels of accuracy at a faster reconstruction speed for our approach. Detailed analyses suggest that cellular neuroanatomy of connectivity of Drosophila neurons are decisive for the achieved accuracy. These properties generalize to Drosophila neurons at different life stages and cell types, and enable robust and efficient mapping of neural circuits in Drosophila. The toolkit is currently applied to map circuits across the phylogenetic tree including mammals. I applied this novel mapping approach to investigate how synaptic circuits and their properties change between developmental stages using the Drosophila nociceptive system as a model. Previous studies suggest that noxious stimulation causes behavioral phenotypes at the late stages of larval development that are absent at early stages. The question thus arises how the underlying synaptic circuits for nociception change across development. Volumetric EM datasets were obtained from several individuals using large-scale, serialsection transmission electron microscopy. Our novel circuit mapping method was applied to reconstruct the class IV multi-dendritic nociceptors and their postsynaptic circuitry at both early and late developmental stages. Changes in synaptic connectivity patterns and morphological properties of neurons were investigated between early and late developmental stages. Results revealed that the general organization and synaptic connectivity of all nociceptive postsynaptic interneurons were preserved from the early to the late stages. Moreover, across developmental stages, interneuron arbors grew considerably and synapse numbers increased 3-4-fold. However, the proportion of the nociceptor inputs relative to the total number of dendritic synaptic inputs remained similar and was cell type-specific. Furthermore, different types of local interneurons receive inputs from different subsets of somatotopically-organized nociceptors, suggesting parallel, specialized pathways for noxious signal processing. The newly identified interneuron types and their genetic driver lines provide a basis for future research to dissect this tractable model system for nociception. In summary, this work contributes tools and methods towards mapping, analyzing and visualizing large-scale neural circuits derived from volumetric EM and demonstrates their practical applicability. This work hints at the tremendous potential of circuit mapping studies to elucidate the relationship of synaptic circuit maps to neural function and animal behavior.