Experimental Progress Zebrafish visual behaviors: We are in the process of combining the two parallel experiments to characterize the functional response properties of neurons in the zebrafish tectum and to reconstruct the underlying synaptic circuitry. We expect to collect a dataset that combines functional imaging with EM circuit reconstruction within the next year. Neurogenesis in the olfactory bulb: We are continuing our efforts to measure the connectivity between adult born neurons and projection neurons in the mouse olfactory bulb using a combination of optogenetic and electrophysiological approaches. We have successfully demonstrated that we can optically stimulate individual neurons in olfactory bulb slices and are now optically imaging connected neurons. In addition, we have developed a method to allow penetration of antibodies into thick tissue sections with minimal disruption to tissue ultrastructure. We are using this method to fluorescently identify neurons containing specific proteins prior to the acquisition of electron microscopy data. EM analysis of retinal circuitry: A collaboration with Dr. Gautam Awatramani led to an article published in Neuron in 2015. We are also in the process of submitting a manuscript following up on our previous work on direction-selective circuitry in collaboration with Drs. Rob Smith and Jeff Diamond. Technical Progress Development of custom serial-block face microtome: We continue to make improvements to our serial block-face microtome including the incorporation of less expensive motors. The patent application for the microtome is under review. 3D EM data analysis: We have pursued two avenues to aid the automated analysis of 3D EM data. On the experimental side, we conducted a study examining the benefits of preserving the extracellular space during chemical tissue fixation that is normally lost. We have confirmed that the preservation of extracellular space improves the automated segmentation of 2D and 3D EM volumes. A manuscript describing this method is under revision at Elife. On the data analysis side, we have continued to improve the performance of neural networks we have implemented to automate reconstruction. Over the last year we have incorporated unsupervised methods to train networks on large image stacks in which labeled data is limited. Web based Proofreading Tool: Our web based EM segmentation proofreading tool is online and externally accessible. We added a tutorial to demonstrate how to use the site. We are using the tool to accumulate more training data for the neural networks.
