The similarities in the morphologies and functions of the zebrafish and human retinas have made the zebrafish visual system a useful research model [1–4]. Like the human retina, the neuronal cell bodies are precisely organized in three major laminae, the ganglion cell layer (GCL), inner nuclear layer (INL) and outer nuclear layer (ONL) . These three laminae are separated by plexiform layers, the inner plexiform layer (IPL) and outer plexiform layer (OPL), which mainly contain neuronal projections . Furthermore, the zebrafish has a cone-dense retina and thus has rich color vision, providing an advantage over nocturnal rodent retina studies [2, 3]. The organization of the genome and the genetic pathways controlling signal transduction and retinal development are also highly conserved between zebrafish and humans . Because zebrafish are highly tractable to both genetic and chemical manipulation, many genetic and chemical screenings have been performed [1–7]. From these screenings, a number of genes and chemicals have been identified that could affect the structures and functions of the vertebrate retina [1–5].
Retinal research has been facilitated by improvements in imaging techniques. Multiple technical developments have permitted the visualization of retinal cell structures and their dynamics in vitro, ex vivo and in vivo . In vitro approaches such as immunohistochemical analyses allow the labeling of certain retinal cell types . DiOlistic labeling with fluorescent dyes is an ex vivo approach to display arbor morphologies of the retina by labeling single cells discretely . These in vitro and ex vivo approaches can be used with multiple probes to detect changes in several retinal cell types simultaneously [9, 10]. However, these approaches are labor-intensive and have relatively low throughputs. As an in vivo approach, stable transgenic zebrafish lines expressing fluorescent proteins such as green fluorescent protein (GFP) have been used [11, 12]. In these transgenic lines, fluorescent proteins are expressed in specific cell types, such as rod photoreceptors , UV-sensitive cone photoreceptors , subtypes of bipolar cells  and retinal ganglion cells (RGC) [11, 12]. Although these transgenic lines can be used for high-throughput in vivo screening by assessing the changes in the fluorescent signals in the retina [11, 12], the assessments are usually restricted to the cells expressing the fluorescent proteins.
Another technique for in vivo imaging is vital staining of the zebrafish retina using fluorescent small molecules [16, 17]. Zebrafish larvae absorb small molecules present in the surrounding water through their skin and gills . Fluorescent sphingolipids such as Bodipy-ceramide have been used as labeling agents for in vivo imaging of the zebrafish retina [19, 20]. Fluorescent sphingolipids are inserted into the plasma membrane of many cells in zebrafish, allowing the cellular and axon layers of the zebrafish retina to be visualized . However, since fluorescent sphingolipids also localize to the interstitial fluid of zebrafish larvae, the whole larvae are stained en masse . Therefore, in vivo imaging of the zebrafish retina using fluorescent sphingolipids requires staining for several hours and at high concentrations (50-100 μM) to achieve visualization.
In this study, we screened fluorescent coumarin derivatives that could resolve these problems. It has been shown that coumarin derivatives can reach the mammalian brain by passing through the blood-brain barrier (BBB) and have potential as therapeutic agents for autoimmune encephalomyelitis  and amyloid imaging agents for Alzheimer's disease . Since the BBB and blood-retinal barrier (BRB) are both endothelial barriers where tight junctions between the endothelial cells seal the vascular lumen , we hypothesized that coumarin derivatives would also be delivered into the retina. This screening identified four coumarin derivatives that are suitable for in vivo imaging of the zebrafish retina.