Genetic Analysis of Retinal Cell Types in Neuropsychiatric Disorders

Abstract

Importance: As an accessible part of the central nervous system, the retina provides a unique window to study pathophysiological mechanisms of brain disorders in humans. Imaging and electrophysiological studies have revealed retinal alterations across several neuropsychiatric and neurological disorders, but it remains largely unclear which specific cell types and biological mechanisms are involved.

Objective: To determine whether specific retinal cell types are affected by genomic risk for neuropsychiatric and neurological disorders and to explore the mechanisms through which genomic risk converges in these cell types.

Design, setting, and participants: This genetic association study combined findings from genome-wide association studies in schizophrenia, bipolar disorder, major depressive disorder, multiple sclerosis, Parkinson disease, Alzheimer disease, and stroke with retinal single-cell transcriptomic datasets from humans, macaques, and mice. To identify susceptible cell types, Multi-Marker Analysis of Genomic Annotation (MAGMA) cell-type enrichment analyses were applied and subsequent pathway analyses performed. The cellular top hits were translated to the structural level using retinal optical coherence tomography (acquired between 2009 and 2010) and genotyping data in the large population-based UK Biobank cohort study. Data analysis was conducted between 2022 and 2024.

Main outcomes and measures: Cell type-specific enrichment of genetic risk loading for neuropsychiatric and neurological disorder traits in the gene expression profiles of retinal cells.

Results: Expression profiles of amacrine cells (interneurons within the retina) were robustly enriched in schizophrenia genetic risk across mammalian species and in different developmental stages. This enrichment was primarily driven by genes involved in synapse biology. Moreover, expression profiles of retinal immune cell populations were enriched in multiple sclerosis genetic risk. No consistent cell-type associations were found for bipolar disorder, major depressive disorder, Parkinson disease, Alzheimer disease, or stroke. On the structural level, higher polygenic risk for schizophrenia was associated with thinning of the ganglion cell inner plexiform layer, which contains dendrites and synaptic connections of amacrine cells (B, -0.09; 95% CI, -0.16 to -0.03; P = .007; n = 36 349; mean [SD] age, 57.50 [8.00] years; 19 859 female [54.63%]). Higher polygenic risk for multiple sclerosis was associated with increased thickness of the retinal nerve fiber layer (B, 0.06; 95% CI, 0.02 to 0.10; P = .007; n = 36 371; mean [SD] age, 57.51 [8.00] years; 19 843 female [54.56%]).

Conclusions and relevance: This study provides novel insights into the cellular underpinnings of retinal alterations in neuropsychiatric and neurological disorders and highlights the retina as a potential proxy to study synaptic pathology in schizophrenia.

Figure 1.. Enrichment of Genetic Risk for Major Neurological and Neuropsychiatric Disorders in Retinal Cells in Gene Set Analysis

A, Single-nucleotide variants from genome-wide association studies (GWASs) for several neurological and neuropsychiatric disorders were assigned to their nearby genes. In addition, the specificity of all genes to retinal cell types was calculated from single-cell RNA sequencing (scRNA-seq) expression data. Then, a cell-type enrichment analysis was performed to reveal whether genes specifically expressed in different cell types were enriched in genetic associations with brain disorders. B, Dotplots were based on human scRNA-seq data for the fovea and the peripheral retina. P values were adjusted for the false discovery rate (FDR) within each trait. Inner black dots indicate statistical significance after additional Bonferroni correction for the 7 investigated traits. The size of the outer dots represents the log-transformed FDR-adjusted Multi-Marker Analysis of Genomic Annotation (MAGMA) P value. AC indicates amacrine cell; AD, Alzheimer disease; BC, bipolar cell; BD, bipolar disorder; HC, horizontal cell; MDD, major depressive disorder; MS, multiple sclerosis; PD, Parkinson disease; RGC, retinal ganglion cell; RPE, retinal pigment epithelium; SCZ, schizophrenia; snRNA-seq, single-nucleus RNA sequencing.

Figure 2.. Associations Between Traits and Retinal Cell Types Across Mammalian Species

The bar plots illustrate the log-transformed Multi-Marker Analysis of Genomic Annotation (MAGMA) P values (false discovery rate [FDR] corrected across each trait) for the association between retinal cells and schizophrenia (SCZ), bipolar disorder (BD), major depressive disorder (MDD), Parkinson disease (PD), Alzheimer disease (AD), multiple sclerosis (MS), and stroke in the nonhuman primate macaca fascicularis and in mice. The dashed lines represent the Bonferroni-corrected significance level of α = .05/7. Vascular cells, retinal pigment epithelium (RPE), pericytes, endothelial cells, and astrocytes were not similarly annotated in the 2 species. AC indicates amacrine cell; BC, bipolar cell; HC, horizontal cell; MG, Müller glia; RGC, retinal ganglion cell.

Figure 3.. Enrichment Analysis Implicating Disrupted Retinal Synapses in Schizophrenia (SCZ)

A-C, Bar plots display the top 10 gene sets from Gene Ontology biological process, cellular component, and molecular function that were overrepresented among those genes that were specifically expressed in amacrine cells (ACs) and significantly associated with schizophrenia (SCZ), as obtained with Enrichr. Specifically, the enrichment analysis was based on the overlap of 226 genes that were in the top specificity decile for both foveal and peripheral ACs and had an unadjusted Multi-Marker Analysis of Genomic Annotation (MAGMA) gene-level P value of less than .05. For additional analyses for each relevant cell type and region separately, see eFigure 4 in Supplement 1. D, Sunburst plot from Synaptic Gene Ontologies and Annotations (SynGO) based on 84 of 226 genes that were significantly associated with SCZ at the gene level, had a cellular component annotation in SynGO, and were specifically expressed in both foveal and peripheral ACs. Colors indicate gene count per term, including child terms. Fields with 10 or more genes mapping to them or their child terms are annotated. E-H, Bar plots illustrate the 10 genes with the highest MAGMA z scores for SCZ among the top 10% cell type–specific genes in each SCZ-relevant cell type. Analyses were based on the human single-cell RNA sequencing data of Cowan et al. HC indicates horizontal cell; RGC, retinal ganglion cell.

Figure 4.. Association Between Ganglion Cell Inner Plexiform Layer (GCIPL)Thickness and Polygenic Risk for Schizophrenia (SCZ) in a Population-Based Cohort

A, Schematic illustration of the retinal layers (retinal nerve fiber layer [RNFL], GCIPL, inner nuclear layer [INL]) that were investigated within the population-based cohort and of the anatomical location of the amacrine cells (ACs). B, Partial robust regression plot between GCIPL thickness (averaged across both eyes) and polygenic risk scores (PRSs) for SCZ in the UK Biobank cohort, adjusted for age, age squared, genetic sex, body mass index, smoking status, hypertension, optical coherence tomography image quality, genotype array, and the first 10 ancestral principal components (n = 36 349; B, −0.09; 95% CI, −0.16 to −0.03; P = .007). The solid line represents the regression line, and the complementary shaded area corresponds to the 95% CI. C, Comparison of GCIPL thickness between individuals in the top (n = 891) and bottom (n = 884) 2.5% of SCZ PRS distribution, using an independent-samples t test (t_1771.1 = −3.22, P = .001). Box-and-whisker plots show the medians (middle lines), first and third quartiles (box ends), and whiskers extending to the lowest and highest values within 1.5× the interquartile range below the first quartile and above the third quartile, respectively. Values outside this range are considered outliers and are not shown. Each dot represents one individual. ^aP < .01.