Abstract
Purpose :
The mouse retina contains 45 molecularly defined types of retinal ganglion cells (RGCs), each extracting and relaying distinct aspects of a visual scene to the brain. However, how cells of each RGC type are distributed across the retinal surface relative to each other as well as other RGC types is largely unknown. Therefore, we set out to generate a comprehensive spatial map of all 45 RGC types, obtain each type’s distribution across the retina, and calculate homotypic and heterotypic cell-cell interactions based on proximity.
Methods :
We used Multiplexed Error Correcting Fluorescent in situ Hybridization (MERFISH) to classify mouse RGC types. The MERFISH panel contained 140 markers, including genes that allowed us to distinguish RGCs from non-RGCs in the ganglion cell layer (GCL) and within each of the 45 RGC types. The flat-mount preparation was optimized to image thin sections (12 microns) while maintaining RNA integrity. We stitched consecutive flat-mount sections to reconstruct the GCL. By optimizing cell segmentation and registration methods, we assigned RGCs to each respective RGC type.
Results :
We profiled 2 adult mouse retinae at P56 (C57BL/6J). Among several areas we reconstructed, 2.7 x 107 micron2 of flat-mount retinae contained ~50000 RGCs. RGCs were segmented and classified into 45 RGC types using an XGBoost model trained on our previous single-cell RNA sequencing atlas of adult RGCs. As an independent test, we applied the same MERFISH panel onto vGlut2-Cre; Ai14 mice retinae and found that genetically marked excitatory neurons in the GCL were annotated as RGCs. Additionally, we integrated with MERFISH immunostaining of Spp1, which selectively marks 4 RGC types known collectively as αRGCs, and showed that MERFISH-annotated αRGCs were Spp1-positive. A deeper analysis revealed the relative frequency and spatial distribution of each RGC type. Preliminary analysis identified interesting homotypic and heterotypic spatial interactions among RGC types.
Conclusions :
We demonstrate a spatial transcriptomics approach for RGC-type mapping on whole retinal flat-mounts. The current technology platform comprehensively maps all 45 RGC types simultaneously in tissue and bypasses mouse genetic labeling to visualize individual RGC and RGC types, enabling high-throughput studies surrounding mouse RGC development and neurodegeneration.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.