Rapid and Integrative Discovery of Retina Regulatory Molecules

Abstract

Retinal function relies on precisely organized neurons and synapses and a properly patterned vasculature to support them. Alterations in these features can result in vision loss. However, our understanding of retinal organization pathways remains incomplete because of a lack of methods to rapidly identify neuron and vasculature regulators in mammals. Here we developed a pipeline for the identification of neural and synaptic integrity genes by high-throughput retinal screening (INSiGHT) that analyzes candidate expression, vascular patterning, cellular organization, and synaptic arrangement. Using this system, we examined 102 mutant mouse lines and identified 16 unique retinal regulatory genes. Fifteen of these candidates are identified as novel retina regulators, and many (9 of 16) are associated with human neural diseases. These results expand the genetic landscape involved in retinal circuit organization and provide a road map for continued discovery of mammalian retinal regulators and disease-causing alleles.

Keywords: high-throughput screening; nervous system; neural disease; neuron; retina; synapse; vasculature; vision loss.

Figure 1.. Large-Scale Identification of Retinal Regulatory Genes

(A) INSiGHT pipeline. 102 of the available 276 lines were selected for inclusion and assayed for defects in topographic patterning of blood vessels and retinal cells, cellular integrity, and synaptic organization. Sixteen lines were identified with defects in one or more of these categories. (B) Schematic of the retina. The outer nuclear layer (ONL) contains rod and cone photoreceptors with outer segments (OSs) and inner segments (ISs). These cells synapse onto interneurons in the outer plexiform layer (OPL). The inner nuclear layer (INL) is comprised of horizontal cells (HCs), amacrine cells (ACs), rod bipolars (RBPs), and cone bipolars (CBPs); the latter three make connections with retinal ganglion cells in the inner plexiform layer (IPL). The ganglion cell layer (GCL) contains retinal ganglion cells (RGCs), whose axons project to the brain, as well as displaced amacrine cells (not pictured). Blood vessels integrate into this network and form a superficial plexus (SP), intermediate plexus (IP), and deep vascular plexus (DP) in the GCL, IPL, and OPL, respectively. (C) IMPC allele structure of the LacZ reporter gene and example images of varied expression patterns. Scale bar, 50 μm. (D) Quantification of the LacZ expression pattern in all 102 mutants across distinct sub-regions of the retina. (E) Percentage of retinal regulatory genes identified in this study that demonstrated defects in one or more of the tested categories. See also Data S1, S2, S3, and S4 and Tables S1, S2, S3, and S4.

Figure 2.. Vascular Organization Defects in INSiGHT Mutant Lines

(A) OCT was used to examine mutants for defective patterning of the superficial blood vessels relative to WT controls. Phenotypes included abnormal vessel crossing as in Bbox1^−/− mutants (n = 7), a decreased number of arterioles as exemplified in Rnf10^−/− mutants (n = 13), and tortuosity as exemplified in Adsl^+/− mutants (n = 16). Black arrows denote changes. Scale bar, 250 μm. (B) To examine inner retina vascularization, mutants were stained with an antibody to CD31. Representative images of the deep (green) and merged vascular layers (deep, green; intermediate, red; superficial, blue) of WT animals and five mutants that demonstrated vascular phenotypes are shown. Scale bar, 100 μm. (C) Slc44a1^−/−animals showed marked defects in superficial blood vessel patterning by OCT (scale bar, 250 µm) as well as alterations in each of the vascular layers following staining for CD31 (deep, green; intermediate, red; superficial, blue; scale bar, 100 µm). (D) Abnormalities in mutants were quantified in the deep vascular layer by counting the number of vascular branch points relative to WT controls (n = 6) following staining with CD31. Significant alterations were observed in Adsl^+/− (n = 4, p < 0.0001), Abca13^−/− (n = 4, p = 0.0002), Bbox1^−/− (n = 7, p < 0.0001), Anapc13^+/− (n = 3, p < 0.0001), Rnf10^−/− (n = 5, p < 0.0001), and Slc44a1^−/− (n = 3, p < 0.0001, two-way ANOVA and Dunnett’s multiple comparisons test for significance). Data are represented as the mean ± SEM. (E) The Slc44a1^−/− retina showed a lack of proper restriction to vessel layers relative to WT animals (arrows). Scale bar, 250 μm. ***p < 0.001. See also Figures S1 and S2 and Data S2 and S4.

Figure 3.. Cellular Defects in INSiGHT Mutant Lines

(A) Nuclei were stained using DAPI (gray) to identify mutants with defects in cellular organization. Representative images of WT animals and five mutants that demonstrated nuclear thinning are shown. Scale bar, 50 μm. (B) Cellular abnormalities in affected mutants were quantified by measuring the thickness of the retina relative to WT controls (n = 10) following staining with DAPI. Significant alterations were observed in Rad9a+/− (n = 4, p = 0.0005), Hmga1^+/− (n = 4, p = 0.0046), Dbn1^+/− (n = 4, p = 0.0066), Bbox1^−/− (n = 7, p = 0.0279), and Ap4e1^−/− (n = 4, p = 0.0153, unpaired two-tailed Student’s t test). (C) In Hmga1^+/− animals, the thickness of each cellular and synaptic layer was visualized and quantified using DAPI to assess the width relative to WT controls (n = 4). Significant thinning was noted in the ONL (p < 0.0001) and INL (p = 0.0366, unpaired two-tailed Student’s t test). (D and E) All major neuron types were visualized in WT and Hmga1^+/− animals using antibodies specific for each population (D): rods (transducin, white), cones (MCAR, red), bipolar cells (Chx10, magenta), retinal ganglion cells (Brn3b/c, yellow), amacrine cells (Syntaxin, green), and horizontal cells (calbindin, cyan) (n = 4). Scale bars, 25 μm. The relative numbers of all major neuron types were then quantified (E). Significant reductions were observed in several neuron types (cones, p = 0.0065; retinal ganglion cells, p = 0.0020; amacrine cells, p = 0.0154; rods, p = 0.0269; unpaired two-tailed Student’s t test). Data are represented as the mean ± SEM. ***p < 0.001, **p < 0.01, *p < 0.05. See alsoFigure S2B.

Figure 4.. Synaptic Organization Defects in INSiGHT Mutant Lines

(A) Staining for the synaptic marker PSD95 (cyan) and horizontal cells (calbindin, magenta) was used to identify mutants with defects in OPL patterning. Representative images from WT animals and four affected mutants are shown. Arrows indicate synaptic defects. Scale bar, 15 μm. (B) The relative levels of synaptic alterations were quantified in WT mice and affected mutants using an antibody to PSD95. Significant increases in the number of mislocalized synapses were observed in Aph1a+/− (n = 5, p = 0.0043), Chpt1^+/− (n = 5, p = 0.0043), Erc1^+/− (n = 4, p = 0.0286), and Ap4e1^−/− (n = 7, p = 0.0012) animals (unpaired two-tailed Student’s t test). (C–E) Ap4e1^−/− animals (n = 7) display marked remodeling of the OPL (C, arrows; quantified in D and E) compared to the WT (n = 6), as revealed by staining for horizontal cells and presynaptic terminals (calbindin, magenta; PSD95, cyan). Both the length and the number of ectopic terminals were increased relative to WT controls (p = 0.0031 and p = 0.0012, respectively; unpaired two-tailed Student’s t test). (F–H) Full-field electroretinograms were recorded from WT (n = 6) and Ap4e1^−/− animals (n = 11). Representative traces are shown. (F) A significant reduction was observed in the a-wave amplitude for flash intensities greater than 0.01 cd*s/m2 (G) and all b-wave amplitudes for flash intensities greater than 0.003 cd*s/m2 for dark-adapted mice relative to WT controls (H). p = 0.0428, p = 0.0162, p = 0.0053, p = 0.0091, p = 0.0155, and p = 0.0158 for scotopic a-waves at 0.05, 0.1, 0.5, 1.0, 5.0, and 20.0 cd*s/m2, respectively; p = 0.0485, p = 0.0180, p = 0.0113, p = 0.0104, p = 0.0078, p = 0.0022, and p = 0.0180 for scotopic b-waves at 0.01, 0.05, 0.1, 0.5, 1.0, 5.0, and 20.0 cd*s/m2, respectively; unpaired t test using the Holm-Sidak method to correct for multiple comparisons. Data are represented as the mean ± SEM. **p < 0.01, *p < 0.05.

Figure 5.. Protein Network Interactions and Gene Ontology Analysis of Retinal Candidates and Known Retinal Disease Genes

(A) A STRING interaction map, including known and predicted interactions, was generated for our candidates(16)andmouseorthologs oftheknown human retinal disease-causing genes (113). Proteins are presented as nodes, where orange denotes the INSiGHT candidate genes, and blue denotes known human retinal disease genes. Nodes are connected by gray edges whose thickness represents the combined confidence score of ≥ 0.4, summated from the evidence types ‘‘experiments,’’ ‘‘databases,’’ ‘‘co-expression,’’ ‘‘neighborhood,’’ ‘‘gene fusion,’’ and ‘‘cooccurrence.’’ We observed and curated seven clusters based on shared themes that most differentiate the known human retinal diseasecausing genes (‘‘metabolic process,’’ ‘‘neuron development,’’ ‘‘sensory projection,’’ ‘‘signal transduction,’’ ‘‘cellular component biogenesis,’’ ‘‘cell communication,’’ and ‘‘transport’’). The densely connected clusters (shaded) consist of 104 known retinal disease-causing genes and 2 candidate genes. The majority of INSiGHT candidate genes (14 of 16) are unconnected to the shaded clusters. (B) Over-represented biologic process gene ontology terms (p ≤ 0.05) for the combined list of the retinal regulatory candidates (16) and known human retinal disease-causing genes were identified using g:GOSt. These terms were used to generate a directed tree graph in which we show gene ontology terms that contain one or more candidate genes. The ontological level and degree of significance are indicated within the circles, where higher degrees of significance are depicted as successively darker shades of gray, and lower ontological levels are depicted as successively smaller circle sizes. For a given gene ontology term,theassociated INSiGHTcandidates(orange) and known retinal disease-causing genes (blue) were counted. The inset contains data from levels L8–L16 and illustrates that candidates are depleted at lower ontology levels. See also Table S6.