Connectomic analysis reveals an interneuron with an integral role in the retinal circuit for night vision

Night vision in mammals depends fundamentally on rod photoreceptors and the well-studied rod bipolar (RB) cell pathway. The central neuron in this pathway, the AII amacrine cell (AC), exhibits a spatially tuned receptive field, composed of an excitatory center and an inhibitory surround, that propagates to ganglion cells, the retina's projection neurons. The circuitry underlying the surround of the AII, however, remains unresolved. Here, we combined structural, functional and optogenetic analyses of the mouse retina to discover that surround inhibition of the AII depends primarily on a single interneuron type, the NOS-1 AC: a multistratified, axon-bearing GABAergic cell, with dendrites in both ON and OFF synaptic layers, but with a pure ON (depolarizing) response to light. Our study demonstrates generally that novel neural circuits can be identified from targeted connectomic analyses and specifically that the NOS-1 AC mediates long-range inhibition during night vision and is a major element of the RB pathway.

Knowing What You Know in Brain Segmentation Using Bayesian Deep Neural Networks

In this paper, we describe a Bayesian deep neural network (DNN) for predicting FreeSurfer segmentations of structural MRI volumes, in minutes rather than hours. The network was trained and evaluated on a large dataset (n = 11,480), obtained by combining data from more than a hundred different sites, and also evaluated on another completely held-out dataset (n = 418). The network was trained using a novel spike-and-slab dropout-based variational inference approach. We show that, on these datasets, the proposed Bayesian DNN outperforms previously proposed methods, in terms of the similarity between the segmentation predictions and the FreeSurfer labels, and the usefulness of the estimate uncertainty of these predictions. In particular, we demonstrated that the prediction uncertainty of this network at each voxel is a good indicator of whether the network has made an error and that the uncertainty across the whole brain can predict the manual quality control ratings of a scan. The proposed Bayesian DNN method should be applicable to any new network architecture for addressing the segmentation problem.

Synaptic Transfer between Rod and Cone Pathways Mediated by AII Amacrine Cells in the Mouse Retina

To understand computation in a neural circuit requires a complete synaptic connectivity map and a thorough grasp of the information-processing tasks performed by the circuit. Here, we dissect a microcircuit in the mouse retina in which scotopic visual information (i.e., single photon events, luminance, contrast) is encoded by rod bipolar cells (RBCs) and distributed to parallel ON and OFF cone bipolar cell (CBC) circuits via the AII amacrine cell, an inhibitory interneuron. Serial block-face electron microscopy (SBEM) reconstructions indicate that AIIs preferentially connect to one OFF CBC subtype (CBC2); paired whole-cell patch-clamp recordings demonstrate that, depending on the level of network activation, AIIs transmit distinct components of synaptic input from single RBCs to downstream ON and OFF CBCs. These findings highlight specific synaptic and circuit-level features that allow intermediate neurons (e.g., AIIs) within a microcircuit to filter and propagate information to downstream neurons.