flies have the opposite phenotypes of warts: All melted mutant R8 express Rh6, and overexpression of melted leads to expression of Rh5 in all R8 [48]. So, melted is both necessary and sufficient for pR8 fate and Rh5 expression.

This suggests that warts and melted control each subtype through their mutual transcriptional regulation; indeed, they are expressed in mutually exclusive subsets of R8 cells that correspond to the y and p subtypes, respectively [48]. Wherever melted is expressed in R8, rh5 is induced; similarly, warts is coexpressed with Rh6. Therefore, warts and melted repress each other in a transcriptionally bistable negative-feedback loop that ensures that R8 cells only make one choice of expressing Rh5 or Rh6. This bistable loop interprets the R7 signal and instructs R8 to make a robust and unambiguous subtype choice [52] (Fig. 4).

It should be noted that, although all other regulatory factors mentioned so far were transcription factors, Warts and Melted are cytoplasmic signaling proteins (Ser/Thr kinase and PH (pleckstrin homology domain) domain proteins, respectively). This reflects their involvement in interpreting the signal from R7 to decide between two alternate fates from an otherwise equipotent precursor cell. It remains that transcriptional effectors are required within the loop to control the expression of warts and melted.

The involvement of such growth regulators in photoreceptor differentiation is an excellent example of how a signal transduction cassette can be reused during development for an entirely different purpose: Long after the photoreceptors have exited the cell cycle, the warts tumor suppressor pathway is no longer required and is now available to regulate this totally different process of cell specification. It should be noted that the other transcription factors described—Spalt, Prospero, Senseless, Orthodenticle, and Spineless—all have very crucial and totally different functions earlier in development. Some of them even act earlier in photoreceptor specification, before these genes are reused for specifying photoreceptor subtypes.

COMPARISON BETWEEN MAMMALIAN AND DROSOPHILA COLOR VISION RHODOPSINS

What can we learn about color vision from integrating data on mammalian and Drosophila rhodopsins? The vertebrate and invertebrate visual systems for color vision are different and have likely evolved independently, even though they are both based on very old photoreceptor molecules, the rhodopsins, which can even be found in bacteria. However, given the about 950 million years of evolution that separate Drosophila and mammals, the basic principles of function, organization, and development are remarkably similar [53].

Human Color-Sensitive Opsins

Humans have three cone cell populations that detect short (S, blue-sensitive), medium (M, green-sensitive), and long (L, red-sensitive) wavelengths by expressing differentially sensitive opsins [43, 54]. Only one rhodopsin is usually expressed per cone type, although as in rabbit or mice, some coexpression is observed [55]. Dim-light vision and motion detection use the broadly tuned rod rhodopsin, which is evolutionarily a green opsin, likely because this represents the middle of the visible spectrum of light.

This rhodopsin is specific to rods, which represent the majority of the human retina, and comprise almost the entire retina in nocturnal animals. Drosophila outer photoreceptors, like rods, also comprise the majority of photoreceptors as motion is critical for all of the fly visual functions. Outer photoreceptors also express a broad-spectrum rhodopsin (Rh1) that is related to green invertebrate opsins. Vertebrate and invertebrate opsins only share one very distant common ancestor of unknown spectral sensitivity, and it is likely that green opsins have evolved independently in the two branches. Although human cones are concentrated in the fovea (other animals do not exhibit this foveal specialization), the fly color receptors are distributed throughout the retina. However, in both cases, S, M, or L cone and fly color photoreceptors are distributed in a random manner [56]. This is in contrast to the fish retina, which contains four types of color photoreceptors that are precisely ordered, likely because they live in a blurry environment where repetitive patterns are unlikely to interfere with perception by regular arrays.

Vertebrate retinal progenitors are multipotent and undergo successive phases of competence during which they successfully give rise to retinal ganglion cells, then cones and the neurons and glial cells that form the first layer of neural processing in the retina. Rods are produced throughout most of these phases. This depends on the expression of transcription factors that successively specify each type of retinal cell fates. The integration of these intrinsic transcription factor decisions ensures that each cell type is properly specified. Many human retinopathies that result in altered opsin expression are associated with mutations in transcription factors that regulate the specification of rodand cone-specific genetic programs—Pax-6, Crx, Nrl, Nr2e3, and Trβ2—which are discussed below.

Photoreceptor and Rhodopsin Specification in Flies and Mammals:

Parallel Themes

As in flies, the identification of factors controlling retinal development have come both from dissecting the promoter of the final differentiation products, the rhodopsin (e.g., nrl and crx), and from the genetic analysis of patients or mice with retinopathy (crx, nr2E3, tr2Beta). This resulted in the identification of two factors—Pax-6 and Crx—that have revealed similarities between vertebrate and Drosophila retina development that would not have been predicted from the independent evolution of the two visual systems.

Pax-6, the master control gene for eye development in flies, is also a critical eye determination gene in vertebrates. Pax-6 mutations in mammals result in an early block in eye development and diseases such as aniridia. The common use of Pax-6 in fly and vertebrate development was explained by the ancestral role of Pax-6 in regulating photoreceptor determination, which led to its recruitment into controlling most steps in retinal (and lens) development. A second parallel between flies and mammals exists between Crx and Drosophila Otd, although the similarities are more difficult to explain. Crx is an Otd/Otx homeodomain transcription factor that plays a major role in controlling most opsins in vertebrates. Crx overexpression in mice produces more rod-like photoreceptors, while reducing Crx function disrupts photoreceptor morphogenesis [57]. Mutations in Crx occur in several inherited human retinopathies, including Leber’s congenital amaurosis and cone-rod dystrophy 2 [58]. Thus, Crx helps regulate photoreceptor differentiation and opsin expression. A functional analogy can be made with its ortholog in flies, Otd, which controls not only rh3 and rh5 but is also required