The visual system has been a classical model to study sensory processing for decades. The visual system has received particular interests for many reasons, including but not limited to that (1) the visual information is highly dimensional (e.g. luminance, contrast, shape, color, movement and etc.) and thus requires delicate encoding mechanisms, (2) our own species, Homo sapiens, heavily relies on the vision for daily life, and (3) the visual cortex displays robust and clearly tuned responses even in anesthetized animals, which was a big advantage for experiments in early years.
Conventional studies of the visual system has been focused on cats and later expanded to monkeys. In particular, landmark works of David Hubel and Torsten Wiesel on cats and monkeys in 1950s-70s not only provided a prototype for later studies on the visual cortex, but also deeply influenced the research on other sensory systems. The choice of carnivores and primates as model organisms for vision research is largely because that they are agitate predators or fruit eaters, which require the sharp vision for survival. Most rodents, on the other hand, were believed to have very poor vision because they are nocturnal and preys, so that they may have qualitatively different visual systems compared to carnivores and primates.
As a consequence, although rodents, especially mouse, have been popularized in many branches of molecular biology and neuroscience (e.g. the study on learning and memory), its value in studying the visual system was not well appreciated until the discovery that cells in the primary visual cortex (V1) in mice show robust orientation selectivity, a signature feature of V1 in carnivores and primates. Although much larger than that in cats and monkeys, the receptive field of mouse V1 cells shares surprisingly similar structures as in other species.
Visual processing starts with the phototransduction in photoreceptors, in which photons interact with opsins to induce electric current through a G-protein mediated signaling cascade. While signals generated by photoreceptors, within their dynamic range, may be largely view as a ‘pixel’ array, nonlinear processing occurs at every following step, as soon as the synaptic transmission from photoreceptors to bipolar cells.
The outcome of retinal processing is subsets of retinal ganglion cells (RGCs) with diverse response properties. In mice, 22 types of RGCs have been described so far based on their distinct functions and/or morphologies, and it will not be surprising if more are unearthed in future. Different RGC groups extract different features of the visual information, and channel into various visual centers in the brain. It has been shown that Hox10d+ RGCs, including 3 subtypes of ON direction-selective ganglion cells (DSGCs) and 1 subtype of ON-OFF DSGCs, specifically project to the AOS. These RGCs prefer slow moving patterns and are tuned to movement direction, which is consistent with the function of the AOS, i.e. image stabilization by vestibule ocular reflex during the animal’s self-motion. At least 80% RGCs projects to the SC in rodents (, a number much higher than that in primates, which may reflect a greater importance of subcortical processing in rodent vision. Another primary target of RGCs is the visual thalamus (i.e. dLGN), which further relays the information to V1.
In mice, nearly all retinal inputs in the dLGN are from axon collaterals of SC-projecting RGCs. The percentage of RGCs that project to the dLGN in mice is still not completely clear. Early studies with the retrograde labeling suggested that retinogeniculate cells comprise up to ~35% of mouse RGCs. However, a recent study with single cell electroporation showed that the vast majority of RGCs terminate in both SC and dLGN in mice. Recent genetic and anatomical studies have revealed two subtypes of ON-OFF DSGCs (DRD4+ and TRHR+) that project to both the SC and dLGN. However, given the large number of SC/dLGN projecting RGCs, the retina must send highly diverse information to these two structures.