As the light sensing part of the visual system, the human retina is composed of five classes of neuron, including photoreceptors, horizontal cells, amacrine, bipolar, and retinal ganglion cells. Each class of neuron can be further classified into subgroups with the abundance varying three orders of magnitude. Therefore, to capture all cell types in the retina and generate a complete single cell reference atlas, it is essential to scale up from currently published single cell profiling studies to improve the sensitivity. In addition, to gain a better understanding of gene regulation at single cell level, it is important to include sufficient scATAC-seq data in the reference. To fill the gap, we performed snRNA-seq and snATAC-seq for the retina from healthy donors. To further increase the size of the dataset, we then collected and incorporated publicly available datasets. All data underwent a unified preprocessing pipeline and data integration. Multiple integration methods were benchmarked by scIB, and scVI was chosen. To harness the power of multiomics, snATAC-seq datasets were also preprocessed, and scGlue was used to generate co-embeddings between snRNA-seq and snATAC-seq cells. To facilitate the public use of references, we employ CELLxGENE and UCSC Cell Browser for visualization. By combining previously published and newly generated datasets, a single cell atlas of the human retina that is composed of 2.5 million single cells from 48 donors has been generated. As a result, over 90 distinct cell types are identified based on the transcriptomics profile with the rarest cell type accounting for about 0.01% of the cell population. In addition, open chromatin profiling has been generated for over 400K nuclei via single nuclei ATAC-seq, allowing systematic characterization of cis-regulatory elements for individual cell type. Integrative analysis reveals intriguing differences in the transcriptome, chromatin landscape, and gene regulatory network among cell class, subgroup, and type. In addition, changes in cell proportion, gene expression and chromatin openness have been observed between different gender and over age. Accessible through interactive browsers, this study represents the most comprehensive reference cell atlas of the human retina to date. As part of the human cell atlas project, this resource lays the foundation for further research in understanding retina biology and diseases.
This atlas is an integrated collection of over 260K cells, derived from single-cell RNA-seq data of the human retina. These retinal cells have been categorized into 10 major classes.
Over 72K bipolar cells derived from scRNA-seq datasets were included in the current atlas, which can be divided into 14 cell types based marker genes.
Over 691K bipolar nuclei were included in the current atlas, which can be divided into 14 cell types based marker genes.
A total of 11 retinal ganglion cell clusters were identified from over 399K retinal ganglion cell nuclei in the current atlas. Utilizing previously characterized markers from macaque, 5 clusters can be annotated.
Over 89 amacrine cell types were identified among over 571K amacrine nuclei in the current atlas. These 89 cell types can be categorized into GABAergic amacrine cells, Glycinergic amacrine cells, and Both type using marker genes, specifically the GABA-synthetic enzymes GAD1 and GAD2, and the glycine transporter SLC6A9. Based on previously characterized markers, 14 of the 89 amacrine cell clusters could be annotated as known amacrine types.
This atlas is an integrated collection of over 3.1 million nuclei, derived from single-nucleus RNA-seq data of the human retina. These retinal nuclei have been categorized into 10 major classes and more than 123 cell types and clusters.
