Single-nuclei multiomics meta-analysis of human organoids and fetal retina reveals developmental transcriptional and epigenetic differences

Abstract

Purpose : Retinal organoids represent a potentially excellent model system to study developmental processes in vivo, however it is uncertain to what extent retinal organoids can model the development of the human retina. We address this issue through a systematic comparison of gene expression profiles, transcriptional trajectories, and chromatin accessibility between retinal organoids and the developing human retina.

Methods : Human fetal retinal samples between 10- and 23-weeks gestational age and retinal organoid samples from 5- and 45-weeks were collected and subjected to single nuclei multiome profiling using the 10X Genomics platform to generate transcriptomic and open chromatin data from the same nuclei. Additionally, this data was integrated with retinal organoid single-nuclei transcriptomic data from previous publications spanning 5- to 28-weeks. After co-embedding using SCVI and Scanpy, RNA velocity and pseudotime trajectory were calculated with Monocle and MultiVelo. ArchR was used to analyze changes in chromatin openness. The gene regulatory networks were uncovered using Pando.

Results : Single nuclei multiome profiling of more than 450,000 nuclei has been generated from 13 retinal organoid samples and 24 fetal retinal samples (including 100,000 nuclei from previous publications). All major cell classes and more than 30 cell types were identified in the organoid. Furthermore, based on pseudotime analysis, the organoid trajectory was well matched to the trajectory inferred from fetal samples. By analyzing cells at different developmental stages within the same cell type between organoid and fetal samples, transcriptional and chromatin openness similarities/differences were identified. Additionally, the effects of different organoid culturing protocols were identified.

Conclusions : Through integrative and comparative analysis of these datasets, we identified both shared and unique regulatory mechanisms that control the developmental process, both between fetal and organoid, and between organoids. These mechanisms were revealed through the comparison between the transcriptome and open chromatin profile in the organoid and fetal samples and involve complex and dynamic interactions between the genome and epigenome. Overall, our study broadens our understanding of the extent to which retinal organoids mimic normal human retinal development.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.