Perisomatic Features Enable Efficient and Dataset Wide Cell-Type Classifications Across Large-Scale Electron Microscopy Volumes

Abstract

Mammalian neocortex contains a highly diverse set of cell types. These types have been mapped systematically using a variety of molecular, electrophysiological and morphological approaches. Each modality offers new perspectives on the variation of biological processes underlying cell type specialization. While many morphological surveys focus on branching patterns of individual cells, fewer have been devoted to sub-cellular structure of cells. Electron microscopy (EM) provides dense ultrastructural examination and an unbiased perspective into the subcellular organization of brain cells, including their synaptic connectivity and nanometer scale morphology. Here we present the first systematic survey of the somatic region of nearly 100,000 cortical cells, using quantitative features obtained from EM. This analysis demonstrates a surprising sufficiency of the perisomatic region to recapitulate many known aspects of cortical organization, while also revealing novel relationships. Parameters of cell size, nuclear infolding and somatic synaptic innervation co-vary with distinct patterns across depth and between types. Further, we describe how these subcellular features can be used to create highly accurate predictions of cell-types across large scale EM datasets. More generally, our results suggest that the shifts in cellular physiology and molecular programming seen across cell types accompany profound differences in the fine-scale structure of cells.

An unsupervised map of excitatory neurons’ dendritic morphology in the mouse visual cortex

“…We refer to these cells as displaced L4 cells. The presence of these cells suggests there are no precise laminar boundaries based on morphological features of neurons, but instead different layers blend into one another, a finding also observed by other authors [15,4].…”

Multi-Layered Maps of Neuropil with Segmentation-Guided Contrastive Learning

“…Comparison of our grouping of layer-5 pyramidal cells to previous work suggests that, based on its distinct morphology and relative rarity, our cluster 2 group (Fig. 4b, inset) corresponds to the type described in the literature as near-projecting (NP) or cortico-cortical non-striatal (CC-NS) 42–44 cells. Consistent with NP basal dendrites being sparse and more morphologically similar to apical dendrites than in typical pyramidal cells, the NP basal cluster 2 was also closer to the apical cluster 1 in SegCLR UMAP space.…”

“…Our cluster 3 “tract” group (Fig. 4d) likely corresponds to the type described in the literature as extra-telencephalic (ET), cortico-subcortical (CS), thick-tufted (TT) or pyramidal tract (PT), which provide the primary cortical output onto subcortical brain areas 42–44 . The cluster 3 “no-tract” cells (Fig.…”

Cell-type-specific inhibitory circuitry from a connectomic census of mouse visual cortex

“…This assignment network was run for each detected cleft, and coordinates of both the presynaptic and postsynaptic partner predictions were logged along with each cleft prediction. For nucleus detection 51,96 a convolutional network was trained to predict whether a voxel participated in a cell nucleus. Following the methods described in Wu et al 104 , a nucleus prediction map was produced on the entire dataset at 64 × 64 × 40 nm 3 .…”

“…Using these boundaries and previously computed nucleus centroids 51 , we identified all cells inside the columnar volume. Coarse cell classes (excitatory, inhibitory, and non-neuronal) were assigned based on brief manual examination and rechecked by subsequent proofreading and automated cell typing 96 . To facilitate concurrent analysis and proofreading, we split all false merges connecting any column neurons to other cells (as defined by detected nuclei) before continuing with other work.…”