Medulla Connectome Obtained Through Serial Section Electron Microscopy
Fly Optic Lobe
The Drosophila melanogaster optic lobe comprises more than 50% of the total volume of its brain, emphasizing the ethological importance of visual input to this flying insect. In our recent work, we have completed the reconstruction of the synaptic connections between neurons within a section of the largest neuropil within the optic lobe – the optic medulla.
Neuron Reconstructions
Our reconstruction was constructed on a substrate of transmission electron microscopy (TEM) images of serially sectioned 40 nm thick slices of tissue through the medulla. The process of reconstruction is detailed in the following video, and involves computer-aided segmentation of the images into membrane bounded cells, that were then visually inspected and corrected by multiple proofreaders.
This process allowed us to reconstruct 379 neurons. These reconstructions are shown below, with all the neurons being shown sequentially in the video.
Download Reconstructions of 379 neurons
The 379 neurons are available to download, here (in a single .zip file). The files are in .swc format, a standard format for reconstructions which can be accessed using many different viewing programs (we recommend the publicly available Vaa3D). These reconstructions will also be available in the near future in the Neuromorpho public repository, under the Chklovskii Archive.
Synaptic Contacts
In addition to neuron reconstruction, the EM images allow us to identify pre and postsynaptic elements within the region. In our reconstruction, we annotated a subset of these synapses.
The medulla is organized into repeating units called medulla columns, each receiving inputs from a single point in visual space. In our reconstruction, we annotated all synapses within a single medulla column, identified the neurons to which they belonged, and then attempted to identify the postsynaptic targets of all these synapses. In this process, we were able to identify 8637 synaptic contacts from the neurons within our central, reference column. The number of synaptic contacts between any two of our reconstructed neurons (supplemented with additional contacts proofread during our analysis of the implementation of motion detection) can be found through our search engine, here.
Connectome Module of a Medulla Column
In our analysis, we were able to classify 290 of our 379 reconstructed neurons into 56 cell types (with the remaining reconstructions too sparse to classify). 27 of these types could be confidently identified as occurring within every medulla column. Utilizing the synaptic contacts between this subset of neurons, we constructed a connectivity matrix of the repeating circuit module within each medulla column.
The following image shows the connections in this circuit module, with the thickness of each connection proportional to the number of synaptic contacts between the two cell types, and with more strongly connected neuron types placed more closely together.
The Implementation of a Neural Computation: Motion Detection
Utilizing our connectome module, and insight from electrophysiological and behavioral experiments into fly motion vision, we were able to identify two cell types, Mi1 and Tm3, likely to implement components of the motion detection circuitry in flies. By additional sparse proofreading of synapses from input L1s to Tm3s, L1s to Mi1s, and both Tm3s and Mi1s to an additional cell type, T4, we obtained evidence that Mi1 and Tm3 mediated receptive field components might constitute spatially displaced inputs to the motion detection circuitry, necessary to implement the computation. This evidence is detailed, for one T4, in the following video.
This approach demonstrates the key insight that connectomics can provide in deducing the computational roles of neurons.