Retinal ganglion cell circuits and glial interactions in humans and mice

Abstract

Retinal ganglion cells (RGCs) are the brain's gateway for vision, and their degeneration underlies several blinding diseases. RGCs interact with other neuronal cell types, microglia, and astrocytes in the retina and in the brain. Much knowledge has been gained about RGCs and glia from mice and other model organisms, often with the assumption that certain aspects of their biology may be conserved in humans. However, RGCs vary considerably between species, which could affect how they interact with their neuronal and glial partners. This review details which RGC and glial features are conserved between mice, humans, and primates, and which differ. We also discuss experimental approaches for studying human and primate RGCs. These strategies will help to bridge the gap between rodent and human RGC studies and increase study translatability to guide future therapeutic strategies.

Keywords: astrocyte; development; microglia; organoid; retinal disease; vision.

Figure 1.. RGC densities and topography in primate and mouse.

(A) Schematic of a flat mount retina in humans and mouse. The human retina is larger than the mouse retina, shows lower RGC subtype diversity, and has a macula and fovea. The primate macula has a much higher RGCs density than the peripheral retina. The mouse retina lacks a macula and fovea, but displays slightly higher RGC densities in the retina center compared to the periphery. (B) Cross section schematic and the retinal thickness of the human and mouse retina[4]. In both species, the retina has three cellular layers, two synaptic layers, and a retina never fiber layer. Rod and cone photoreceptor nuclei reside in the outermost layer, whereas horizontal, bipolar, most amacrine, and Müller glia nuclei (not shown) reside in the inner nuclear layer. RGC and displaced amacrine nuclei reside in the ganglion cell layer, and RGC axons and astrocytes reside in the retina never fiber layer. Microglia processes are enriched in the synaptic layers. The primate fovea shows a characteristic displacement of inner retina neurons so that light has direct access to the dense cones populating this region. (C) Approximate spatial distributions of RGC densities in primates versus mice. RGCs are most numerous in the primate macular region and absent from the foveola, whereas in mouse, RGC density is only moderately higher in the central retina (modified from [9]). (D) Shared transcription factors contribute to RGC specification in mouse and primates[39]. Figure was constructed using BioRender.

Figure 2.. Timeline of critical events in human and mouse RGC and retinal development.

(A) In mice, eye field specification begins at embryonic day 8 (E8). (B). RGCs are born from retinal progenitor cells in the outer retina beginning at E11, peaking at E14.5, and ending just prior to birth [26]. Microglia are born in the yolk sac and colonize the retina at E10–11. (C) Mouse RGCs begin to extend axons at E12. (D) Mouse astrocytes are born in the optic disc progenitor zone. (E). Mouse RGC axons are myelinated by oligodendrocytes starting from the optic disc between postnatal day 8 (P8) and P16. (F) In humans, eye field development begins at fetal week 3(Fw3). (G) Human RGCs are born between Fw4 and Fw14 and, like mice, are specified from retinal progenitor cells. (H) Starting at Fw6, human optic nerve formation begins, followed by retinal glia colonization. RGC axons cross the optic chiasm at Fw10–11. (I). Human macula and fovea development begins at Fw11. (J). RGC axon myelination begins in the brain at Fw22 and reaches the optic nerve head by Fw36 to 40. (K). Schematic of a timeline depicting the birth order of retinal neuron and glia cell types in mouse and human. The relative order in which retina cell types are born is conserved in human and mice, and all cell types are born in windows that span weeks (mice) to months (humans). In mice, some retinal cell types are born postnatally, while in humans all cells are generated prior to birth. In both mice and humans, astrocytes and microglia develop outside of the retina and migrate into the retina early in development. The overall developmental stages are similar between monkeys and humans. Data adapted from [175]. Figure was constructed using BioRender.

Figure 3.. RGC developmental events that impact retinal connectivity and function.

(A) In mice, a subset of early born, non-apoptotic RGCs are engulfed by microglia, which reduces overall RGC numbers. The complement molecule C1q appears to help drive this selective engulfment. (B). RGCs, as well as most other retina neuron types, undergo normal developmental apoptosis, which leads to the death of approximately 30% all RGCs that are generated from retina progenitor cells. (C). RGCs selectively extend their dendrites to specific retinal synaptic layers, and their neurites are further refined through RCG subtype-specific processes that involve selective pruning, sequential addition, and biased dendritic targeting[176]. In some cases, molecular pathways that help drive laminar selection have been identified in mice, but these pathways are unknown for the majority of RGC types. (D) All RGC survival and dendritic refinement outcomes are complete prior to mouse eye opening. These processes are impacted by a spontaneous form of activity called retinal waves. Different cells drive three types of retinal waves that follow each other in development from E16 to P14. Gap junctions between RGCs are thought be important for phase 1 retinal waves, while acetylcholine release by a type of amacrine cell known as a starburst amacrine is important for phase 2. Glutamate release from bipolar cells contributes to phase 3 retinal waves[72]. Figure was constructed using BioRender.

Figure 4.. Development of eye-brain connectivity in primates and mouse.

(A) Approximate distributions of contralateral and ipsilateral RGC axonal projections from the eye to the brain. (B) Structural schematic of the mouse and primate lateral geniculate nucleus (LGN). In mice, the LGN is composed of a central region innervated by ipsilateral RGC axons and a surrounding region that receives contralateral axons. These areas first overlap but then become segregated through the activity of microglia, retinal waves, and intrinsically photosensitive RGCs (ipRGCs) activity. The primate LGN differs significantly from that of mouse. The developing primate LGN is arranged into 2 layers, the parvocellular and the magnocellular. In adults, the LGN has a highly layered organization with 12 laminae. Among these, layers 2, 3, and 5 receive ipsilateral RGC input while layers 1,4, and 6 receive contralateral input. The koniocellular layers (K) receive sparse and heterogeneous RGC input. LGN patterning emerges after E91 in In fetal rhesus monkeys. (C) Timing and organization of RGC axon projections to the mouse and human LGN. In mouse, RGC axons are initially numerous and complex. As RGCs undergo developmental cell death, the number of RGC axon terminal decreases, and the remaining arbors are developmentally refined. Mouse RGC axons reach adult complexity by P30. In contrast, in primates RGC axonal terminal arbors show a consistent increase in complexity during fetal week 12 to 16. Mouse data were adapted from [95]. (D-E). Timing and duration of critical events during visual circuit maturation in mouse and primates. RGC transcriptomes in mice at E12–14 and E16-P0 align with human fetal RGCs at days 52–57 and 67–107, respectively. In general, the order of maturation events is conserved between species. Data adapted from [177]. Figure was constructed using BioRender.

Figure 5.. RGC subtype diversity in primates and mice.

(A) Schematic of the ratio of RGC types in primate and mouse. Primates have approximately 18 known RGC types, the majority of which are midget RGCs (60% on average, but 90% in the fovea) and parasol RGCs (16%). In contrast, mice have ~50 RGC types, most of which comprise a small percent of the total. (B) ‘En face’ structural schematic of dendrites for a subset of RGC types and estimated abundance relative to the total number of RGCs. (C) Estimated RGC dendritic area as a percentage of the total retinal area for the most conserved types in mouse and human. While alpha RGCs most closely align with midget and parasol RGCs at the transcriptional level, their dendritic structures and relative coverage area differ dramatically. A single midget cell dendritic area spans ~0.001% of the human retina, while a single alpha RGC occupies ~0.1% of the mouse retina. In contrast, in humans, ipRCGs are approximately 10 times larger than those in mouse. Dendritic tracing images are adapted from[119,178,179]. (D) Schematic of mouse αRGCs and their transcriptionally closest partner among humans RGC depicting the conservation of their laminar targeting, general functional properties (ON versus OFF, sustained versus transient), and brain targeting to the LGN. Figure was constructed using BioRender.

Figure 6.. Current experimental models in mice and humans for understanding visual development and disease.

(A) Schematic showing mouse (upper panel) and human (lower panel) experimental system alternatives for modeling the visual system and its diseases. In mice, axon injury is often modeled using optic nerve crush. Genetically modified mice can also be used to model human visual diseases by either expressing known human mutations associated with these conditions or by the induction of disease-related changes, even if their causal models or mutations are not directly linked to human disease. For example, mice do not normally develop glaucoma or most other RGC-related visual diseases, prompting the development of chronic injury models and induced ocular hypertension models. For studies of human RGCs, human retinal organoids from human pluripotent stem cells (hPSC) can be used. Alternatively, RGCs can be purified for study in a dish or in transplantation models. Human RGC axon regeneration can be modeled using microfluidics chambers or assembloid strategies in which retinal organoids are paired with brain organoids. Human disease-related mutations can be introduced into hPSCs from which retinal organoids or RGCs can be derived. (B) Cartoon representation illustrating the potential differences in genetic risk factor expression between human and mouse RGCs across different RGC-related diseases. In some cases, a conserved RGC type may express a genetic risk factor that is shared between both species. In other cases, the risk factor may be conserved in mouse RGCs but expressed in a different RGC type than that in humans. Alternatively, both the expression of the risk factor and the RGC type in which it is expressed may not be conserved across species. Figure was constructed using BioRender.