A simple approach to single cell RNAseq analysis facilitates virtual cell sorting of retinal cell subtypes derived from human pluripotent stem cells

Abstract

Purpose : To probe the authenticity of human pluripotent stem cell (hPSC)-derived photoreceptors (PRs) and other retinal subtypes generated within 3D hPSC-optic vesicle (OV) like structures using single cell RNA-seq (scRNAseq) analysis. While scRNAseq provides a powerful tool to profile individual cell gene expression, the datasets are challenging to interpret, particularly when profiling heterogenous cell populations. Thus, we sought to develop a straightforward strategy to mine scRNAseq data to extract reliable information on the multitude of cell populations in differentiating hPSC-OVs.

Methods : hPSC-OVs were generated from a custom CRX-tdTomato hPSC reporter line using our established protocols. RT-PCR and immunocytochemistry (ICC) were performed at multiple time points to assess the differentiation status of retinal cell types, focusing on PR production and maturation. Thereafter, we performed scRNAseq analysis on dissociated hPSC-OVs at 70 (n=157 cells) and 218 (n=195 cells) days of differentiation using the Fluidigm C1 system and the Illumina HiSeq 2500 sequencing platform.

Results : Molecular and ICC analyses showed robust tdTomato expression in PRs, as well as RPE cells, beginning at early stages of differentiation. Production of all other retinal cell types was also confirmed. Subsequent scRNAseq analyses using commonly employed, unbiased algorithms were unfruitful in clearly identifying cell types. Therefore, we developed a simple and versatile approach to generate transcriptome datasets akin to those produced following high throughput analysis of FAC-sorted gene reporter lines. This method takes into account the rank order of expression of a given target (or “bait”) gene across the entire scRNAseq dataset and searches for genes with the highest correlations in an unbiased fashion. Moreover, using this method, we could input multiple “bait” genes simultaneously to refine datasets that more reliably reflect specific retinal cell subtypes, demonstrating the expression of known and novel retinal cell subtype specific genes. Novel gene expression was then validated by ICC on hPSC-OV and human retinal sections.

Conclusions : Using our novel scRNAseq analysis method, we provide further evidence of the authentic nature of hPSC-derived PR precursors and uncover previously unknown genes that are expressed in differentiating human retinal cell subtypes.

This is an abstract that was submitted for the 2017 ARVO Annual Meeting, held in Baltimore, MD, May 7-11, 2017.