In retinitis pigmentosa, retinal detachment, age-related macular degeneration, and glaucoma, retinal neuronal cells are damaged by a common mechanism, apoptosis. Because apoptosis is an active process that requires de novo expression of a "death message", this process can be controlled by inhibiting the expression of the "death message". We first studied whether a retinal ischemia-reperfusion model can be used as a model for retinal neuronal apoptosis. In the retinal ischemia-reperfusion injuries, typical features of apoptosis, including TUNEL-positive cells, DNA ladder formation, and ultrastructural features of apoptosis were found. Using the model, systematic research to identify the "death message" was done by DNA microarray analysis. About 200 messages were found to be up- or down-regulated during the process of retinal ischemia-reperfusion. These genes were divided into four groups: (1) transcription factor genes, (2) cell cycle-related genes, (3) reactive oxygen scavenger genes and (4) molecular chaperon genes. The possible roles of such genes in neuronal apoptosis following retinal ischemia-reperfusion injury were studied. In the model, reactive oxygen species produced by reperfusion was found to generate lipid peroxides and induced up-regulation of a transcription factor, c-Jun, that further induced aberrant expression of cell cycle-related genes such as cyclin D1 in amacrine cells. However, because no controlled expression of cell cycle-related genes takes place in retinal neurons, amacrine cells died by a G1 arrest mechanism. On the other hand, horizontal cells never expressed cyclin D1 and the cells were found to die by necrosis. The study revealed a possible mechanism of retinal neuronal apoptosis and it also became apparent that different types of neurons use different "death messages". Furthermore, the possibility that inhibition of a "death message" sometimes induces necrosis rather than apoptosis was shown. This means that we need to try inhibition of the death mechanism upstream rather than downstream. Administration of thioredoxin, an endogenous reactive oxygen species that blocks generation of lipid peroxides and thus inhibits the death process upstream, was found to be neuroprotective against retinal ischemia-reperfusion injury. Aberrant expression of c-Jun and cyclin D1 was down-regulated by the treatment. Possible roles of caspases were also studied by using the ischemia-reperfusion injury, RCS rat, and excessive light exposure damage in wild type and caspase-1 deficient mice. Also, application of adeno-associated virus that carries Bcl-xL was tested to find possible neuroprotective effects on RCS rats. Our studies showed that caspase-1 played a more important role in the retinal photoreceptors and caspase-3 was important in neurons in the inner nuclear layer. Caspase-2 was found to be a major caspase in the retinal ganglion cell layer. In agreement with the findings, caspase-1 deficient mice showed less prominent light damage than wild type mice. Gene therapy by Bcl-xL was effective to protect retinal photoreceptor damage in RCS rats.