We analyze the establishment of the retinal mosaic that supports color vision. The fly eye is composed of 800 unit eyes (ommatidia), each containing 8 photoreceptors (PRs) (R1-R8). The six outer PRs (R1-R6) express the broad spectrum Rhodopsin 1 (Rh1) and mediate motion detection. The two inner PRs, R7 and R8 express different rhodopsins (rhs) and are responsible for color vision when their inputs are compared in the optic lobes. Two subtypes of ommatidia called y and p are randomly distributed throughout the main part of the retina. In the p subtype (30% of ommatidia), R7 expresses UV-sensitive Rh3 and R8 blue-Rh5 and these ommatidia discriminate among short wavelengths. In the y subtype (70%), R7 expresses long UV-Rh4, while R8 expresses green-Rh6, thus allowing better discrimination among longer wavelengths. Although PR development starts at the morphogenetic furrow in the eye imaginal disc, their differentiation only occurs later during pupation. At this stage, expression of the spalt gene distinguishes between inner and outer PRs. The R7 vs. R8 identity decision is achieved by prospero (R7) and senseless (R8). The p and y ommatidial subtypes are then specified stochastically by spineless (see below). Spatial specialization defines distinct regions of the retina: The Dorsal Rim Area involved in polarized light vision (that serves as the compass of the fly) is defined by homothorax while a dorsal region involved in solar vs. antisolar orientation is specified by the Iro-C. [Mollereau Nature 2001; Cook Dev Cell 2003; Tahayato Dev Cell 2003; Wernet Cell 2003; Wernet Current Biol 2007; Mazzoni PLoS Biol 2008]
Control of the stochastic choice between two ommatidial fates: spineless (Bob Johnston) Stochastic events play an important role in many biological processes. The p and y ommatidial subtypes are distributed stochastically in the retina, similarly to human cone PR distribution. R7s first make the p vs. y choice and then impose fate onto R8. The stochastic expression of the spineless gene in R7 determines which ommatidia become p or y. Whereas most stochastic systems are controlled by random activation, stochastic spineless expression is determined by global activation coupled with random repression requiring combinatorial inputs from multiple cis-regulatory elements acting at long range. A transvection-like mechanism coordinates expression of the two spineless alleles to yield a robust frequency of expression. Spineless induces expression of Rh4, repression of Rh3, and inhibition of a signal to R8s. It directly activates the rh4 promoter, whereas it triggers a complex interlocked feedforward loop motif to indirectly repress rh3. These distinct modes of regulation determine differential response thresholds to stochastic Spineless activity. Therefore, this stochastic system is finely tuned to trigger distinct network motifs that determine stochastic or regional expression outcomes by overriding or incorporating underlying biases. [Wernet Nature 2006; Losick Science 2008; Johnston Cell 2011]
Deterministic vs stochastic choice in evolution (Mike Perry) The stochastic expression of Spineless determines the two subtypes of ommatidia. Specification of another PR type used in detecting polarized light is controlled instead in a deterministic, spatial manner by the transcription factor Homothorax. We have identified a fly species (Dolichopodidae, or long-legged flies) in which color vision PR subtypes are specified deterministically in regular, alternating rows instead of a random, stochastic distribution. By examining both the underlying mechanisms and functional consequences of each type of patterning, as well as the changes involved in the evolutionary transition between them, we hope to better understand these fundamentally different types of cell fate specification.
A bistable loop in R8 reinforces R8 fates induced by R7 to R8: (Brent Wells): R7s that do not express spineless commit to the p fate and express rh3; they then send an instructive signal to underlying R8s, which also commit to the p fate and express rh5. In the absence of the R7 signal, R8s commit to the y fate and express rh6. We have identified a cascade of genes required in R8 cells to ensure the correct choice of y vs. p cell fate. warts, which encodes a Ser/Thr kinase tumor suppressor gene, is necessary and sufficient for R8 to adopt the y fate. Another growth regulator, melted plays the opposite role and induces the p fate in R8. warts and melted are expressed in a complementary manner in the yR8 and pR8 subsets, respectively and repress each other's transcription to form a bi-stable loop. Other members of the Warts tumor suppressor pathway, hippo and salvador have phenotypes identical to warts while yorkie and scalloped have the opposite phenotype, as would be expected from their role in tumor suppression. We are studying the signaling mechanisms required for R8 subtype specification, which lead to robust expression of Rh5 in pR8 and Rh6 in yR8. Dpp signaling, initiated in R7, appears to act through the Activin receptor, Baboon (in R8) in a non-canonical pathway to establish R8 fate, likely by interacting with the bi-stable warts/melted loop. To identify other members of the pathway, we are conducting an RNAi-based screen of over 3,000 membrane-associated and signaling molecules looking at subtype expression of p and y R7 and R8. [Mikeladze-Dvali Cell 2005]
Opposite network-level feedbacks in the Hippo pathway for growth regulation and post-mitotic neural fate: (David Jukam) Signaling pathways are re-used for multiple purposes in development. The Hippo tumor suppressor pathway coordinates proliferation and apoptosis in dividing cells via the transcriptional co-activator and oncogene, YAP/Yorkie (Yki), which negatively regulates itself indirectly through negative feedback. We have shown that only a subset of the numerous upstream Hippo regulators that probe the environment to finely tune growth is required upstream of the Hippo core complex for Warts expression and activity in R8 to specify Rh6 fate. Pathway activity is disrupted when Melted represses Warts transcription to specify the opposite fate of Rh5. The R8 Hippo pathway therefore exhibits ON-or-OFF regulation, promoting a binary output and mutually exclusive R8 subtypes. Furthermore, upstream Mer and Lgl are continuously required to maintain R8 neuronal subtypes. This shows that the Hippo pathway is re-implemented for sensory neuron fate by combining canonical and non-canonical regulatory steps.
In post-mitotic R8 fate specification, Yki positively regulates itself through network-level positive feedback to ensure a stable fate decision and generate mutually exclusive expression of Rh5 or Rh6 in R8. Yki and its co-factor Scalloped (Sd) transcriptionally repress warts and promote melted for fate determination while coordinately repressing Rh6 and promoting Rh5. Thus, the homeostatic configuration of the Hippo tumor suppressor network has been re-wired in post-mitotic neurons for positive feedback--a new regulatory mechanism appropriate for an all-or-nothing neuronal fate decision. Altering the feedback architecture of conserved signaling modules provides an efficient mechanism for co-option of signaling networks for diverse purposes in development and evolution. [Jukam Dev Cell 2011]
The cis-regulatory code underlying subtype-specific rhodopsin expression: (Jens Rister) Decoding the 'grammar' of cis-regulatory elements that dictates temporal and spatial gene expression is a major challenge. As terminal differentiation genes, Rhs are particularly well suited for this endeavor, as they likely follow a simpler regulatory code than genes that are dynamically expressed at several stages of development and that often require several distant cis-regulatory modules. Indeed, compact upstream regions of less than 300 bps recapitulate endogenous rh patterns. We also benefit from our deep knowledge of key transcriptional activators and repressors (and their respective binding sites) that control subtype-specific rh expression. Our dissection of the rh promoters has revealed that single base pair changes in regulatory elements that are highly conserved among rh genes can be critical for proper subtype-specific expression. Furthermore, we have identified novel motifs that prevent co-expression of rh5 and rh6 and are currently screening candidate TFs that bind to them obtained from a yeast-one hybrid screen performed in cooperation with the lab of Simon Sprecher (Fribourg), who studies rh5 and rh6 regulation in the larval eye. As these rhs are expressed at an earlier stage of Drosophila development, this allows us to investigate the underlying cis-regulatory grammar in different cellular and developmental contexts. [Mazzoni Neuron 2005; Sprecher Nature 2008]
Maintenance of mutual repression of rhodopsin genes (Daniel Vasiliauskas): The Warts/melted bistable loop ensures that the decision in R8 to express Rh5 or Rh6 is robust. However, we have shown that the presence of a functional Rh6 protein is also required to maintain exclusion of Rh5 from yR8 photoreceptors. In rh6 mutants these photoreceptors are specified correctly, and Rh5 expression is initially restricted to the p R8s. However, as the fly ages without Rh6 activity, yR8s also start to express Rh5, suggesting that Rh6 is required to maintain repression of the rh5 gene. We are investigating the involvement of the phototransduction cascade and developmental mechanisms in this pathway and are also screening for other genes that function to maintain exclusive Rhodopsin expression in the fly photoreceptors. [Vasiliauskas Nature 2011]