Sensory receptor expansion and neural accommodation in butterfly color vision

The evolution of more complex brains required existing neurons and neural circuits to accommodate new inputs. The genetic and developmental mechanisms that enable such integration are largely unknown. Butterflies evolved more complex retinal mosaics through the addition of a second R7 color photoreceptor per ommatidium (unit eye). In Drosophila , the unique R7 makes a stochastic choice to express one of two opsin genes. In butterflies, the two R7s make two independent stochastic cell fate choices in each ommatidium, producing three ommatidial types instead of the two in flies. Here, we investigate the developmental basis of this change and how the butterfly brain accommodates an expansion in sensory receptor input. We first identified the developmental mechanisms that specify the second R7 cell. We then modified Drosophila retinas to have butterfly-like transcription factor expression, causing recruitment of an additional R7. The two R7s make independent stochastic choices, like butterflies, leading to three stochastically distributed ommatidial types instead of two. In Drosophila , each of the two subtypes of R7s connects to one of two types of Dm8 neurons. Dm8 neurons of both types are born in excess, with Dm8s that are not connected to their cognate R7s undergoing apoptosis. In the presence of extra R7s in butterfly-like retinas, additional Dm8s are retained, leading to two Dm8s per medulla column. We propose a model in which population-level variation in the ratio of R7 subtypes produced in the retina maintains an excess pool of two types of potentially interacting partners in the brain, thus providing developmental flexibility which allows brains to accommodate additional inputs. We demonstrate how excess neurons that are eliminated in normal development facilitated the expansion of color vision in butterflies, and similar cooption events may represent a general mechanism of neural evolution.