The developing retina expresses multiple bHLH transcription factors. Their precise functions and interactions in uncommitted retinal progenitors remain to be fully elucidated. Here, we investigate the roles of bHLH factors ATOH7 and Neurog2 in human ES cell-derived retinal organoids.  Single-cell transcriptome analyses identify three states of proliferating retinal progenitors: pre-neurogenic, neurogenic, and cell cycle-exiting progenitors. Each shows different expression profile of bHLH factors. The cell cycle-exiting progenitors feed into a postmitotic heterozygous neuroblast pool that gives rise to early born neuronal lineages. Elevating ATOH7 or Neurog2 expression accelerates the transition from the pre-neurogenic to the neurogenic state, and expands the exiting progenitor and neuroblast populations. In addition, ATOH7 and Neurog2 significantly, yet differentially, enhance retinal ganglion cell and cone photoreceptor production. Moreover, single-cell transcriptome analyses reveal that ATOH7 and Neurog2 assert positive autoregulation, suppress key bHLH factors associated with the neurogenic progenitors, and elevate bHLH factors expressed by exiting progenitors and differentiating neuroblasts. This study thus provides novel insight regarding how ATOH7 and Neurog2 impact human retinal progenitor behaviors and neuroblast fate choices.

The datasets were collected by performing single-cell RNA-sequencing of human ESC-derived retinal organoids transduced by lentiviral vectors expressing EGFP, ATOH7, or Neurog2. Viral transduced cells were FACS sorted and subjected to 10X Genomics single-cell barcoding and cDNA library construction. Illumina NovaSeq6000 S2 paired-end 2x50bp mode was used to sequence the libraries. Spliced Transcripts Alignment to a Reference (STAR) version 2.5.1b (cellranger count) was used to perform sequence alignments to the reference human genome (GRCh38), barcode counts, and UMI counts to yield summary reports and t-Stochastic Neighboring Embedding (t-SNE) dimensionality reduction. For downstream analyses, cells with a number of unique molecular identifiers (UMI) > 2500 per cell and < 0.1 % mitochondrial gene expression were used. For LV-GFP, LV-AEP, LV-NEP samples, the mean reads per cell ranged from 139,000-195,000, with mean gene per cell ranging from 2935-3079. The resulting total single-cell counts used for analysis were 3004 for LV-EGFP, 2063 for LV-AEP, and 3909 for LV-NEP infected samples.