Single cell ATAC-seq of the developing human retina and stem cell-derived retinal organoids shows changing epigenomic landscapes during cell fate acquisition.

Single cell genomic analyses provide a new perspective on the development of complex tissues, like the vertebrate neural retina. We previously used single cell transcriptomic analysis to characterize human fetal retinal development and assessed the degree to which retinal organoids recapitulate normal fetal development. In this study we extend the transcriptomic analyses of fetal human retina to incorporate single nuclear scATAC-seq, a powerful new method to characterize potential gene regulatory networks through the changes in accessible chromatin that accompany cell state changes. The combination of scATAC-seq and sc-RNA-seq can be used to study the process of neurogenesis in the developing fetal human retina at unprecedented resolution. We identify the key transcription factors relevant to specific fates and the order of the transcription factor cascades that define each of the major retinal cell types. These transcriptomic and epigenomic features are largely recapitulated in retinal organoids; however, there are differences in Notch signaling and amacrine cell gene regulation, which are not as robust in organoids. The datasets we generated constitute an excellent resource for the continued improvement of retinal organoid technology, and have the potential to inform and accelerate regenerative medicine approaches to retinal disease. Overall design: Initially, we performed scATAC-seq on freshly dissociated human fetal retina and stem cell-derived retinal organoids. We collected 12 fetal retinal samples as follows: two eyes from fetal week 8 (FW8), three eyes from FW11 and five eyes from FW13 samples. To see the developmental changes within a sample, we separated the central and peripheral retina at FW13. We also processed 5 samples of retinal organoids at equivalent ages of the fetal retina (FW8 and FW11). We sequenced a total of 82,648 cells, with 6,309 average median fragments per cell. After sequencing, we used CellRanger ATAC1.2.0 to align the reads and carry out initial stages of data processing. Peaks were called with MACS2 and further analyzed using Signac/Seurat (Hao et al., 2021), Chromvar (Schep et al., 2017) and Monocle (Cao et al., 2019; Trapnell et al., 2014).