Computational modelling and experimental investigation of retinal tissue self-organisation

Abstract

Individual retinal cell types exhibit semi-regular spatial patterns called retinal mosaics. These mosaics could enable uniform sampling of visual information and are formed to varying degrees across cell types. Retinal ganglion cells (RGC) and amacrine cells are notably known to exhibit such layouts. RGCs dendritic arbours also form organised structures, laminating at different levels and exhibiting specific morphologies depending on the considered type. Mechanisms responsible for the formation of such organised structures and their requirements are still not well understood. Mosaic formation follows three main theories: (1) homotypic cells prevent nearby cells from adopting the same type, (2) cell tangential migration, with homotypic cell repulsion, (3) cell death (with RGCs exhibiting high rates of apoptosis). Here, we use BioDynaMo, an agent-based simulation framework, to build a detailed and mechanistic model of mosaic formation. In particular, we investigate the implications of the three theories. We report that the cell migration mechanism yields the most regular mosaics. We also found that cell death can create regular mosaics only if the death rate is kept below 30%, after which cell death have a negative impact on mosaic regularity. We also investigate the implication of intrinsic and extrinsic factors in the development of RGC dendritic tree morphologies. Waves of spontaneous activity sweep across the RGC layer from postnatal day (P) 0- 10 and may drive the development of some cellular features (mosaic, dendritic lamination). Using a combination of immunohistochemistry and pan-retinal activity recording, we report transient clusters of auto-fluorescent cells around the optic disc during the period of cholinergic waves. They migrate towards the periphery between P2-9 and then disappear, coincidentally with the switch from Stage II to stages III waves. Waves origin follow a similar centre-to-periphery developmental pattern. We propose here that these clusters represent activity hotspots and are the sites for wave initiation.