Retinitis pigmentosa is an incurable retinal disease that leads to blindness. One puzzling aspect concerns the progression of the disease. Although most mutations that cause retinitis pigmentosa are in rod photoreceptor–specific genes, cone photoreceptors also die as a result of such mutations. To understand the mechanism of non-autonomous cone death, we analyzed four mouse models harboring mutations in rod-specific genes. We found changes in the insulin/mammalian target of rapamycin pathway that coincided with the activation of autophagy during the period of cone death. We increased or decreased the insulin level and measured the survival of cones in one of the models. Mice that were treated systemically with insulin had prolonged cone survival, whereas depletion of endogenous insulin had the opposite effect. These data suggest that the non-autonomous cone death in retinitis pigmentosa could, at least in part, be a result of the starvation of cones. Retinitis pigmentosa is a type of inherited retinal degeneration. It is currently untreatable and usually leads to blindness. With over 40 reintitis pigmentosa genes identified, it is the most common type of retinal degeneration caused by a single disease allele (RetNet, http://www.sph.uth.tmc.edu/Retnet/). The phenotype is characterized by an initial loss of night vision as a result of the malfunction and death of rod photoreceptors. This phase is followed by a progressive loss of cones. Because cones are responsible for color and high-acuity vision, it is their loss that leads to a reduction in the quality of life. In many cases, the disease-causing allele is expressed exclusively in rods; nonetheless, cones die as well. Indeed, to date there is no known form of retinal degeneration in humans or mice where rods die and cones survive. In contrast, mutations in cone-specific genes result only in cone death. Several theories have been proposed to explain this finding. For example, cone death could be a result of the release of a toxin produced by dying rods or the loss of a trophic factor that is produced by healthy rods 1‐6 . Alternatively, cone death could be caused by microglia that are mobilized initially during rod death 7 or by oxidative stress 8,9 .O xidative stress might also directly harm cones. The constant flow of oxygen through the retinal pigmented epithelium (RPE) to photoreceptors and the loss of rods, which are 95% of the photoreceptors in human and mouse, may result in an overload of oxygen to the remaining cones 10 . Evidence for all of these mechanisms exists in mice, yet none are able to fully explain why cones may survive for many years in the absence of rods in humans. Nonetheless, rodents are a very good model for this type of retinal degeneration. Although the rodent retina lacks a macula, which is the cone-rich and rod-free area that is present in humans, the macula is not involved in the initial phase of the disease. In humans, retinitis pigmentosa starts outside of the macula, where the distribution of rods and cones is similar to that in mice. To determine the common underlying mechanism for cone death in retinitis pigmentosa, we compared four mouse models harboring mutations in rod-specific genes (Pde6b ‐/‐ (ref. 11), Pde6g ‐/‐ (ref. 12), Rho ‐/‐ (ref. 13) and P23H 14 , which carries a Rho transgene that has an amino acid subsitution of histidine for proline at amino acid 23, as occurs in some human cases of retinitis pigmentosa, see Methods). Affymetrix arrays were used to identify common changes in gene expression that accompany cone death. Changes in a substantial number of genes involved in cellular metabolism coincided with the onset of cone death. These changes were suggestive of cones suffering from a shortage of nutrients. We then found that cones showed signs of autophagy, a cellular self-digestion process, which is consistent with prolonged starvation. We also found that several aspects of the insulin/ mammalian target of rapamycin (mTOR) pathway, an important pathway that regulates cellular metabolism, were affected during the period of cone degeneration. As a result of this finding, we increased and decreased the insulin level and measured the survival of cones in one of the models. Mice treated systemically with insulin had prolonged cone survival, whereas depletion of endogenous insulin had the opposite effect. Therefore, cone starvation is a likely contributor to the slow demise of cones in humans with retinitis pigmentosa. Treatments aimed at improving nutrition of cones are thus a plausible therapeutic avenue. RESULTS Photoreceptor death kinetics and microarray analysis To establish a framework for comparing gene expression in four different models of retinitis pigmentosa, we established the equivalent stages of disease pathology through examination of the kinetics of rod (Fig. 1 and Supplementary Figs. 1 and 2 online) and cone (Fig. 2 and