In vertebrates, visual information is detected and conveyed from the eye to the brain by multiple retinal ganglion cell (RGC) types, which can be subdivided based on anatomical, physiological and molecular characteristics. Intrinsic features as well as selective connectivity with other neuronal cell types render each RGC type selectively sensitive to a distinct visual feature, and characterizing these RGC types is essential to further understand the complex transmission of visual information. Recent studies have used high-throughput single cell RNA sequencing (scRNA-seq) to generate comprehensive molecular atlas of RGC types in several vertebrate species. It is challenging, however, to associate molecularly defined RGC types with the behaviors their activation elicits. To address this, the zebrafish is a particularly attractive model system as the larvae are transparent, allowing for detailed
in vivo imaging of structure and imaging. Additionally, they develop rapidly and display a diverse range of visual behaviors as early as 5 days post fertilization. Here, we used scRNA-seq to classify larval and adult zebrafish RGCs at the transcriptional level, identifying 33 putative RGC types in the adult stage and 29 types in larva. These types possessed unique molecular signatures and most could be uniquely labeled based on selective expression of a single gene. This molecular atlas allowed us to further explore the expression of transcription factors, neuropeptides, and cell surface molecules across RGC types and their role in establishing unique cell identities. To investigate the extent to which RGC diversification is complete in the larval stage, we were able to assign adult identities to larval RGCs using a supervised classifier trained on larval clusters. The results indicated a high degree of transcriptional correspondence between adult and larval clusters and provide insight into the diversification of RGC types.
Using CRISPR genome engineering and intersectional genetics, reporter lines were generated that labeled RGCs based on their expression of select transcription factors. These lines allowed us to characterize the anatomical and physiological characteristics of molecularly defined RGC types. Fluorescent imaging revealed that different molecular types project to different areas of the brain. Using in vivo calcium imaging, we showed that RGCs respond in stereotyped patterns to visual stimuli that included bright and dark flashes, moving gratings, and prey-like stimuli. We functionally characterized a small set of RGC types defined by selective expression of the transcription factor eomesa to find that they exhibit response profiles that encode ambient luminance levels. Finally, we performed loss-of-function experiments to show that eomesa+ RGCs, in relation with their physiology, specifically regulate light-seeking navigation. Taken together, this molecular taxonomy of RGC types is comprehensive and allows us to examine RGC types from molecular composition to anatomical and physiological properties to behavior.
