49 — Structural and functional diversity of a dense sample of retinal ganglion cells

This paper represents an enormous effort to map retinal tissue from a mouse eye. Retina, which is part of the central nervous system, is highly organized: Each “pixel” of visual field is represented by relatively similar cellular machinery, so mapping a large sample of retina, like this, is highly informative of how all mouse retina will look.

In this study, over 0.3mm² of retina (full-depth) are reconstructed in 3D, using manual tracing aided by deep-learning, through the EyeWire web application. This work aimed to find out if the somewhat arbitrary segmentation of sub-layers of inner-plexiform-layer (IPL) of retina, first put forth by Cajal in the 1980s, holds significance in light of modern advances, or if we can do a better job segmenting the tissue today.

Prior works have improved this IPL sublaminar demarcation: Wassle et al demonstrated three staining-boundaries between four of Cajal’s five layers — S1 and S2, S2 and S3, and finally, S3 and S4/5, using differences in calbindin/calretinin staining. Famiglietti & Kolb demonstrated a coarser, two-layer system, differentiated only be functional response to ON or OFF stimuli.

This paper uses bipolar-cell arbor segregation as an anatomical IPL sublaminar marker: The system put forth is Outer Marginal, Outer Central, Inner Central, Inner Marginal. The boundary between the two inners and outers is the same as Famiglietti & Kolb’s ON/OFF boundary, and the boundaries between marginals and centrals are delineated by the presence of starburst amacrine cells (SACs).

The authors then use the full retinal reconstructions to construct clusters of cells based on the localization of their arbors along the depth of retina. The paper shows the presence of 47 total clusters, six of which appear to be novel, as-of-yet uncharacterized ganglion cells.

This work is fascinating because it paves the way for groundbreaking connectomic analysis of large-scale circuitry in the central nervous system. The Seung lab is also planning to release all data and code used for this study, which will represent an enormous contribution to the field of neuroscience. I’m looking forward to checking it out!