Generation of uniform fly retinas

The compound eye of Drosophila melanogaster is composed of about 800 ommatidial units, each containing eight photoreceptor cells (R1–R8). Based on their spectral sensitivities, three subtypes can be distinguished [1]: The ‘pale’ (p) and ‘yellow’ (y) ommatidia [2] are distributed randomly throughout the main part of the eye (Figure 1A). They show different spectral sensitivity, at shorter and longer wavelengths, respectively. The third subtype, consisting of morphologically distinct ommatidia with monochromatic inner photoreceptors, is found in a narrow stripe at the dorsal rim (DRA) of the adult eye and probably detects polarized light [3]. Here, we completely re-designed the stochastic p/y mosaic by manipulating only three genes: spineless [4], the growth regulator melted, and the tumor suppressor warts/lats [5]. By enforcing unambiguous cell fate decisions independently in both R7 and R8 cells, completely homogeneous retinas consisting of only one ommatidial subtype can be created, including unusual opsin combinations that never occur in the wild type. Ommatidial subtypes can be distinguished based on the opsin genes expressed by their central photoreceptors R7 and R8 [6]: R7 cells of p-ommatidia contain a UV-sensitive opsin, Rh3 (Figure 1B), while the R7 cells of y-ommatidia cells contain a distinct UV-opsin, Rh4 (Figure 1C) [7]. R8 cells of p-ommatidia always express blue-sensitive Rh5 [8,9], while the R8 opsin expressed in yommatidia, Rh6, is green-sensitive [10]. Therefore, opsin expression in R7 and R8 cells of the same ommatidium is tightly coupled [8–11]: Rh3 is always coupled to Rh5 in p-ommatidia while Rh4 is found with Rh6 in y-ommatidia. On some occasions, Rh3 can be found in the same ommatidium with Rh6, creating a low percentage of Rh3/Rh6 ‘odd-coupled’ ommatidia (Figure 1C) [11]. The gene spineless (ss) encodes the Drosophila homologue of the vertebrate Dioxin receptor and plays a crucial role in stochastic specification of ommatidia [4]: Loss of ss leads to a complete loss of the y-subtype in R7, while the p-subtype expands. Late overexpression of ss in all R7 cells has the opposite effect, inducing rh4 at the expense of rh3. The tumor suppressor warts (wts) and the growth regulator melted (melt) specify opposite pand y-fates specifically in R8 [5], without affecting opsin expression in R7. melt is expressed in the p-subtype R8 and wts in R8 of y-ommatidia, where they mutually repress each other’s expression. melt is necessary and sufficient for inducing the p-fate in R8, while wts specifies the y-fate in R8. melt and wts thus form an effector system specifying the R8 fates in response to a yet unknown instructive signal from R7 to R8. We have performed additional experiments addressing the epistatic relationship between ss, melt and wts, resulting in a two step model for retinal patterning in Drosophila (Figure 1D; Supplemental data). First, R7 cells make a random but biased cell fate choice: By expressing ss, about 70% of R7 cells develop towards a rh4-expressing fate. In a second step, only rh3-expressing R7 cells send a signal to the underlying R8 cells where expression of rh5 is induced through transcriptional activation of melt and repression of wts, thus creating p-ommatidia. The remaining R8 cells in y-ommatidia express the default opsin rh6, which does not require instruction from, or even the presence of, R7 cells [8,9]. Neither manipulation of the p/y subtype choice in R7 (by loss or gain of ss), nor in R8 cells (by loss or gain of wts/melt) alone are sufficient to create ‘homogeneous retinas’ containing only one ommatidial subtype