An integrated transcriptional analysis of the developing human retina

Abstract

The scarcity of embryonic/foetal material as a resource for direct study means that there is still limited understanding of human retina development. Here, we present an integrated transcriptome analysis combined with immunohistochemistry in human eye and retinal samples from 4 to 19 post-conception weeks. This analysis reveals three developmental windows with specific gene expression patterns that informed the sequential emergence of retinal cell types and enabled identification of stage-specific cellular and biological processes, and transcriptional regulators. Each stage is characterised by a specific set of alternatively spliced transcripts that code for proteins involved in the formation of the photoreceptor connecting cilium, pre-mRNA splicing and epigenetic modifiers. Importantly, our data show that the transition from foetal to adult retina is characterised by a large increase in the percentage of mutually exclusive exons that code for proteins involved in photoreceptor maintenance. The circular RNA population is also defined and shown to increase during retinal development. Collectively, these data increase our understanding of human retinal development and the pre-mRNA splicing process, and help to identify new candidate disease genes.

Keywords: Alternative splicing; Circular RNAs; Developing human retina; Immunohistochemistry; Retinal cells; Transcriptome.

Fig. 1.

RNA-seq analysis defines three developmental windows characterised by stage-specific transcriptional expression. (A,B) Kurtosis analysis showing a decrease as development proceeds; the adult retina samples are shown in red (A). (C) ME-based cluster analysis with developmental windows highlighted in red (4.6-7.2 PCW), green (7.7-10 PCW), blue (12-18 PCW) and purple (adult retina); the scores for the 15 different MEs are shown in black (positive) or white (negative), with circle size proportional to the absolute value. (D) Z scores of upregulated genes for each stage of development defined by ME-based cluster analysis (comparisons performed between two sequential stages, e.g.4.6-7.2 PCW versus 7.7-10 PCW). Blue and red colours in the heat map correspond to low and high gene expression, respectively, and bars on the right-hand side indicate the developmental stages (red for 4.6-7.2 PCW, green for 7.7-10 PCW, blue for 12-18 PCW and purple for adult retina). (E) GO analysis of upregulated genes for each stage of development defined by ME-based cluster analysis (comparisons performed between two sequential stages, e.g. 4.6-7.2 PCW versus 7.7-10 PCW).

Fig. 2.

PCA analysis and heatmap data. Comparison of the RNA-seq samples analysed in this study (red) versus these reported by Hoshino et al. (2017) (blue). Both datasets display a consistent pattern of developmental progression in accordance with one another. Note also that the spread of the datasets mirrors the extent of their developmental time span.

Fig. 3.

The expression of retinal marker genes. (A) Expression of retinal marker genes plotted as log2 transformed counts at each developmental stage included in the RNA-seq analysis. The Wilcoxon rank-sum test was performed on the expression differences between developmental stages to identify the earliest stage with a significant and sustained increase in the expression of retinal lineage cell markers. Ad, adult retina. (B) Expression of retinal markers across the three developmental windows (defined by ME-based cluster analysis) and adult retina plotted as log2 transformed counts per million. The Wilcoxon rank-sum test was performed on the expression differences to identify developmental stages with peak expression of these retinal markers. The red arrows indicate a significant and sustained increase in expression (P<0.05). (A,B) The line that divides the box shows the median, while the box indicates the upper and lower quartiles. Whiskers represent the highest and lowest value excluding outliers, while dots show outlier values outside 1.5 times the interquartile range above the upper quartile and below the lower quartile.

Fig. 4.

Immunohistochemical analysis of developing human retina at 12-18 PCW. (A-D) PAX6 and (E-L) the retinal ganglion cell markers HuC/D, islet 1/2 and TUJ1in the human foetal retina at 12-18 PCW. INBZ, inner neuroblastic zone; ONBZ, outer neuroblastic zone; RPE, retinal pigment epithelium. Scale bars: 50 µm.

Fig. 5.

Analysis of photoreceptor marker expression in the human foetal retina at 12-19 PCW. The emergence of developing and subtypes of photoreceptors were determined by detection with photoreceptor precursor marker CRX (A-E), pan photoreceptor marker recoverin (F-J), cone photoreceptor markers (K-O) opsin blue and (P-T) opsin red/green, and (U-Y) the rod photoreceptor precursor marker NRL. Hoechst staining is in blue. Scale bars: 100 µm in B,C,K,L,M,O,Q,T,V,X; 50 µm in A,D,F,G-J,N,P,R,S,W; 20 µm in Y. High-magnification insets at 18 and 19 PCW are added to show photoreceptor morphology.

Fig. 6.

Stage-specific pre-mRNA splicing during human retinal development. (A) rMATS analysis showing the percentage of transcripts containing retained introns (RI), skipped exons (SE), alternative 3′ splice sites (A3SS), alternative 5′ splice sites (A5SS) and mutually exclusive exons (MXE). (B-D) Gene ontology enrichment analysis showing biological (left-hand panels) and cellular (right-hand panels) processes affected by alternative splicing during human retinal development. (B) 7.7-10 versus 4.6-7.2 PCW; (C) 12-18 versus 7.7-10 PCW; (D) adult retina versus 12-18 PCW.

Fig. 7.

Alternatively spliced transcripts include genes associated with inherited retinal disease and ciliogenesis. (A) Cross comparison of alternatively spliced transcripts identified during retinal development with genes associated with retinal disease (retnet) and cilia genes (syscilia). (B) Examples of three genes (PROM1, CEP290 and CC2D2A) regulated via alternative splicing between 12-18 and 7.7-10 PCW. The splicing events are illustrated using IGV sashimi plots (see Table S4 for a full list). Transcript numbers are Ensembl identifiers. Green highlights indicate alternative splicing events. (C) Schematic representation of key processes affected by pre-mRNA splicing during human retinal development from 7.7 PCW to adult.

Fig. 8.

circRNAs abundance increases during human retinal development. (A) Swarm plot showing frequency of reads spanning circRNA (backsplice) junctions and canonical junctions in each sample relative to reads mapping to other RNA biotypes, as a proportion of the total reads mapped to all biotypes. Samples are colour coded according to the developmental windows defined by ME-based cluster analysis. (B) CircRNA enrichment across the developmental windows defined by ME-based cluster analysis. Ratios were derived by dividing total number of backsplice reads with canonical junction reads. Data are mean±s.e.m. The increase in number across all stages is statistically significant (Jonckheere-Terpstra test). (C) Boxplot showing distribution of circRNA sizes (genomic span between donor and acceptor splice sites) across the developmental windows defined by ME-based cluster analysis. The increase in size is statistically significant (Jonckheere-Terpstra test). Boxes define upper and lower quartiles, with the median indicated and outliers shown as solid circles. (D) Changes in abundance of circRNA derived from genes differentially expressed between developmental windows defined by ME-based cluster analysis. Pearson correlation coefficients are shown. Genes with very low circRNA expression levels at both time points being compared (<1 junction read per million reads per sample) were excluded from the analysis.