Using Transcriptomic Analysis to Assess Double-Strand Break Repair Activity: Towards Precise in vivo Genome Editing

Abstract

Mutations in more than 200 retina-specific genes have been associated with inherited retinal diseases. Genome editing represents a promising emerging field in the treatment of monogenic disorders, as it aims to correct disease-causing mutations within the genome. Genome editing relies on highly specific endonucleases and the capacity of the cells to repair double-strand breaks (DSBs). As DSB pathways are cell-cycle dependent, their activity in postmitotic retinal neurons, with a focus on photoreceptors, needs to be assessed in order to develop therapeutic in vivo genome editing. Three DSB-repair pathways are found in mammalian cells: Non-homologous end joining (NHEJ); microhomology-mediated end joining (MMEJ); and homology-directed repair (HDR). While NHEJ can be used to knock out mutant alleles in dominant disorders, HDR and MMEJ are better suited for precise genome editing, or for replacing entire mutation hotspots in genomic regions. Here, we analyzed transcriptomic in vivo and in vitro data and revealed that HDR is indeed downregulated in postmitotic neurons, whereas MMEJ and NHEJ are active. Using single-cell RNA sequencing analysis, we characterized the dynamics of DSB repair pathways in the transition from dividing cells to postmitotic retinal cells. Time-course bulk RNA-seq data confirmed DSB repair gene expression in both in vivo and in vitro samples. Transcriptomic DSB repair pathway profiles are very similar in adult human, macaque, and mouse retinas, but not in ground squirrel retinas. Moreover, human-induced pluripotent stem-cell-derived neurons and retinal organoids can serve as well suited in vitro testbeds for developing genomic engineering approaches in photoreceptors. Our study provides additional support for designing precise in vivo genome-editing approaches via MMEJ, which is active in mature photoreceptors.

Keywords: DSB repair pathways; double-strand breaks; hiPSC-derived retinal organoids; inherited retinal dystrophies; photoreceptors; single-cell RNA-seq.

Figure A1

E15 mouse retina scRNA-seq. (a) E15 mouse retinal cells in two-dimensional UMAP showing canonical retinal markers expressed in relation to the annotated clusters. (b) 2D visualization of E15 retinal cells, showing cell-cycle scores across the cells. Retinal progenitors form a circle in the UMAP embedding, resembling the cell-cycle before cell-type commitment. (c) DSB repair pathway scores compared (t-test) across progenitor, neuroblast, and photoreceptor precursor cell-types. (d) Photoreceptor developmental trajectory of E15 in two-dimensional UMAP colored by diffusion pseudotime (dpt pseudotime) order.

Figure A2

Retinal organoids scRNA-seq. (a) Matrix plot of the top five markers, calculated by t-test, for each Leiden cluster of hiPSC-derived retinal organoids. (b) Matrix showing the overlap score between the markers calculated for the Leiden clusters (0–13) and the known markers for retinal cell-types (Table A1). (c) hiPSC-derived retinal organoid cells in two-dimensional UMAP, showing the cell-cycle score computed per cell. (d) Subset of neuronal cell-fate in hiPSC retinal organoids. Cells are plotted in two diffusion dimensions (DC) and UMAP to show the highest variation found in the transition from progenitor cells to retinal neurons.

Figure A3

Mouse and human adult retina scRNA-seq. Three adult retina scRNA-seq dataset are analyzed in this figure: (a,b) human adult retina, (c) mouse adult retina. Each dataset is visualized in two-dimensional UMAP, colored by annotated clusters (cell types) and PR markers (ARR3 and GNAT1). Violin plots show DSB repair pathway scores in rods and cones.

Figure A4

MMEJ repair components viewed across species. Expression level of MMEJ resection factors and PARP1 are plotted across species. Gene expression level are plotted individually for each replicate.

Figure A5

Bioinformatic pipeline for bulk and single-cell RNA-seq. Schematic view of the pipelines to process raw reads from both bulk and scRNA seq datasets.

Figure 1

Large genomic portions can be corrected by in vivo genome editing relying on endogenous DSB repair pathways activity. (a) Coding sequence length of IRD genes listed in RetNet. The dashed line indicates the maximum cargo limit for AAV transfer. (b) Scheme illustrating homology-dependent genome engineering (HDR or MMEJ) to replace an entire mutation hotspot exon. Colored boxes (orange and blue) indicate homology regions. (c) DSB repair pathway diagram illustrating the enzymes listed in Table 1. After a DSB in the genomic DNA, the broken ends can be resectioned or protected. Protection by TP53BP1 or WRN, and Ku complex leads to NHEJ repair (orange). On the other hand, resection forms single-strand DNA overhangs that can reveal homology. Binding of PARP1 and the presence of micro-homology leads to MMEJ repair (blue). If resection is prolonged, repair is by SSA or HDR depending on the cell-cycle stage (green).

Figure 2

Cell-cycle-related gene expression correlates to DSB pathways. (a) E15 mouse retinal cells in two UMAP dimensions showing clusters of known amacrine cells, retinal ganglion cells and PR cells at this developmental stage. (b) Subset of the developmental PR trajectory in two UMAP dimensions. Arrows represent the vector calculated for each cell according to the RNA velocity approach. (c) Pseudo-timed HDR, MMEJ, and NHEJ scores within the photoreceptor developmental trajectory. Cell colors refer to clusters in b). (d) DSB repair pathway activity during cell-cycle exit. Lines represent the mean value, and the error bands the 95% confidence intervals.

Figure 3

hiPSC-derived retinal organoids. (a) hiPSC-derived organoid sequenced cells visualized in two t-sne dimensions, showing different cell-type composition at four different time points. (b) 26,700 quality filtered cells merged together and colored by batch, visualized in two-dimensional UMAPs. (c) Merged retinal organoid dataset visualized in two-dimensional UMAPs and colored by annotated retinal clusters. (d) Immunohistochemistry indicates, after scRNA-seq analysis, the presence of distinct rod (GNAT1) and cone (ARR3) populations in the outer layer of the hiPSC-derived retinal organoids. (e) Pseudo-timed HDR, MMEJ, and NHEJ scores within the neuronal lineage of hiPSC-derived organoids. Lines represent the polynomial fit to the data. (f) DSB repair pathway activity scores of rods and cones in hiPSC-derived retinal organoids.

Figure 4

Cell-cycle-related genes correlate to DSB pathways. (a) Violin plot of gene expression abundance in DSB repair pathways in primates and rodents. Pairwise comparison significance was tested by Wilcoxon signed-rank test. (b,c) PCA plot and Spearman correlation of all sequencing replicates showing the relationship between DSB repair pathway genes in human retinas and developing in vitro neurons.