The gene encoding Rh6 was cloned using a Calliphora complementary DNA with significant homology to previously characterized fly rhodopsins [19]. Rh6, like Rh5, shares features consistent with the hallmarks of known rhodopsin molecules. The chromophore linkage sites, an N-glycosylation site, and cysteine residues to form a disulfide bridge are all present in Rh6, which also contains conserved proline residues in the transmembrane domain and the cytoplasmic domains responsible for G protein binding.

Rh5 belongs to the blue family of invertebrate opsins and shares the highest amino acid similarity (59%) with locust S. gregaria lo2 opsin [18], also a blue opsin. Rh6 is a green opsin that is distant from the UV opsins. Other flies, such as Musca and the blowfly (Calliphora), which have a similar retinal organization as Drosophila, also have Rh5 and Rh6 orthologs [27]. Calliphora also has two R8 photoreceptor subtypes, pR8 and yR8, that are reported to have slightly different morphologies, a situation undetected in Drosophila melanogaster [28]. Drosophila Rh5 and Rh6 share higher amino acid identity with their Calliphora, honeybee, and mosquito orthologs than with other Drosophila rhodopsins, suggesting that rh5 and rh6 genes were both present in a common ancestor rather than having evolved independently in each fly lineage [27]. However, Rh5 is slightly more distant from the UV class—it shares 32% similarity with Rh3 and 33% with Rh4—than is Rh6, with 41% homology with Rh3 and 44% with Rh4 [17]. Together, this and other sequence evidence such as intron-exon structure, imply that Rh3/Rh4, Rh5, and Rh6 are ancestral rhodopsins, while a recent event in higher flies (after the mosquito lineage evolved) led to the duplication into the current Rh3 and Rh4.

On the other hand, Rh5 is more divergent than Rh6 from Drosophila Rh1 (31%) and Rh2 (31%), which are themselves divergent members of the green class of invertebrate opsins (Fig. 3). Rh1 and Rh2 appear unique to the higher dipteran lineage and are replaced by several green-type opsins in bees and butterflies. They may have evolved from an ancestral Rh6-like green opsin and have evolved to acquire a very broad spectrum of absorption through the acquisition of a second retinal chromosphore that absorbs blue light and transfers it to the retinal inside the rhodopsin pocket [23]. Their evolution is likely linked to the appearance of an “open rhabdome” in these flies: The separation between the rhabdomeres of the different photoreceptors allowed for the utilization of a very broad-spectrum opsin such as Rh1.

DEVELOPMENT AND PATTERNING OF RHODOPSINS FOR DROSOPHILA COLOR VISION

Mutually Exclusive Rhodopsin Expression

Only after the Drosophila color-sensitive opsin genes had been cloned and protein localization determined could their genes and promoters be manipulated using the powerful genetic tools available in Drosophila. This allowed for the discovery of new genetic pathways and mechanisms that coordinate the complex expression pattern of rhodopsins in the fly retina.

The “one-receptor molecule per neuron” rule in sensory systems ensures that the system can discriminate among sensory inputs by allowing each neuron to act as a

Fig. 3. Phylogenetic relationship among color-sensitive rhodopsins. The phylogenetic tree displays the broad evolutionary relationship among opsin sequences. Vertebrate and invertebrate color-sensitive opsins likely evolved from a single common ancestor and later events created the subfamilies of green-, red-, blue-, and ultraviolet (UV)-sensitive opsins. GF = goldfish; Rh rhodopsin; ZF = zebrafish.

specialized unit of perception [29–33]. In the fly retina, photoreceptors also express “one rhodopsin per photoreceptor” to compare responses to different light wavelengths and create color contrast. In the Drosophila eye, the decision to express one of four rhodopsin molecules in inner photoreceptors, Rh3 or Rh4 in R7 and Rh5 or Rh6 in R8, establishes the basis for the retinal mosaic of photoreceptor subtypes and hence the basis for color vision [34]. It should be noted that coexpression of different opsins exists in bees and butterflies, although its significance is not clear and might correspond to the lack of the broad-spectrum Rh1 in these species. Even in mice, green and blue opsins are often found in the same cones, perhaps reflecting the fact that mice are nocturnal animals and practically color-blind.

Although the distribution of p and y ommatidia is random, the 70:30 y:p ratio is constant from retina to retina across a large number of flies that diverged over 120 million years ago and rarely deviates more than 5% in either direction (35, 36, D. Vasiliauskas and C. Desplan, personal communication). This implies that photoreceptor subtype specification and color-sensitive rhodopsin expression are tightly regulated during development.

The known mechanism for how photoreceptor subtypes are specified is, so far, one of the most detailed descriptions of how neurons express sensory receptors in a mutually exclusive manner. Next, we discuss recent results to illustrate the use of seemingly baroque and diverse mechanisms to control neuronal fate specification, rhodopsin expression, and retinal patterning, all resulting in one receptor per neuron.

Transcription Factors Specify Outer from Inner Photoreceptors and Distinguish R7 from R8

The first step leading to mutually exclusive rhodopsin expression is the specification of a photoreceptor as belonging to the inner versus outer class. Eye development starts at the morphogenetic furrow, a wave of determination that moves across the developing eye imaginal disk during the third-instar larva life and leaves in its wake regularly spaced clusters of eight photoreceptors that will become each ommatidium. R8 is the first photoreceptor to be recruited, and it will recruit neighboring cells to become pairs of photoreceptors (R2/R5, R3/R4, and R1/R6) through sequential induction by the epidermal growth factor (EGF) receptor pathway. R7 is the last cell to be recruited, by a process that requires a dedicated signaling pathway, the Sevenless receptor tyrosine kinase and its ligand Boss [13]. It takes at least 4 days after R8 is first determined for R8 to complete its pale or yellow subtype specification, then differentiate and express Rh5 or Rh6.

Which events in photoreceptor determination occur between the larval stage and the terminal differentiation in late pupal life, when all photoreceptors and subtypes are present? Cell–cell signaling and cell-specific transcription factors initially determine the identity of photoreceptors as outer or inner [37, 38]. The spalt gene complex (sal) directs R7 and R8 toward an inner photoreceptor state away from an outer photoreceptor “default” state. In sal mutants, all photoreceptors develop as “outer,” and all express Rh1. Once R7 and R8 are specified as generic inner photoreceptors, they continue toward more restricted fates. Further distinctions between R7 and R8 are achieved by the R7-specific expression of the transcription factor prospero, which is required to establish R7 fate and to repress R8 features (in particular Rh5 and Rh6 expression). A parallel role is ascribed to the zinc finger protein Senseless in R8 [39, 40].

The analysis of R7and R8-specific rhodopsin promoters provides initial insight into mutually exclusive rhodopsin expression. Transgenic analysis shows that minimal promoters, often less than 300 bp in length, are sufficient to recapitulate Rh3, Rh4, Rh5, and Rh6 expression in vivo [39, 41, 42]. Although in Old World monkeys the exclusion mechanism between the M (green) and L (red) opsins is due to the regulation by a locus control region (LCR) distally located upstream of both genes, a similar mechanism can be ruled out in Drosophila because the four color rhodopsin genes in flies are not clustered and can be on distinct chromosomes [43–45]. Furthermore, even rhodopsin promoter transgenes are regulated appropriately and are expressed in the same exclusive subset as the endogenous gene with activity that they report, indicating trans-regulatory mechanisms.

All four rhodopsin promoters contain a common rhodopsin conserved sequence 1 (RCSI)/palindromic 3 (P3) site near the TATA box; this site has been shown to provide general eye or photoreceptor-specific expression [46]. The site is recognized

by the homeodomain of Pax-6, the “master control gene for eye development.” This function as a regulator of photoreceptor determination likely represents the ancestral role of Pax-6. Upstream from the RCSI/P3 site are rhodopsin-specific sequence elements common to the rhodopsins expressed in each photoreceptor (R7 vs. R8) or in each subtype (y or p). These appear to confer R7, R8, as well as subtype specific regulation. For instance, a conserved 11-bp element, seq56, found in both the rh5 and rh6 promoters is recognized by the R7-specific transcription factor Prospero. Binding by Prospero protein to seq56 in R7 prevents the expression of these R8 rhodopsins in R7 [39]. Indeed, when prospero or when the binding sites are mutated, Rh5 and Rh6 are expressed inappropriately in R7. Thus, Prospero restricts R8 rhodopsins to their proper cell type, R8.

Another highly conserved site is found in the promoter of pR7 and pR8 rhodopsins, rh3 and rh5, respectively, which are always found in the same ommatidium. The homeodomain transcription factor Orthodenticle (Otd) binds to these sites, which are required for the p subtype: Rh3 and Rh5 are no longer expressed in otd eye mutants [42]. However, otd has only a permissive function as it is expressed in all photoreceptors, and its overexpression leads to no effect on rhodopsin transcription; other transcription factors must be required to control the pale R7/R8 subtype. Furthermore, the y pair, Rh4 and Rh6, is not expanded in otd mutants that lack the pR7 and pR8. Therefore, while the analysis of the promoters of Drosophila rhodopsins demonstrated that their mutually exclusive and subtype-specific rhodopsin regulation was transcriptional in nature (as opposed to posttranscriptional regulation at the messenger RNA or protein-trafficking level), genetics was used to unravel the mechanism of rhodopsin exclusion in Drosophila color vision.

A Stochastic Decision Induces Rhodopsins in R7 Photoreceptor

The specification to either the p or y ommatidial subtype is a two-step process that begins in R7 (Fig. 2). The instructive event that determines the 30:70 ratio of rhodopsins results from the stochastic expression of spineless (ss), which encodes yet another transcription factor, this time from the PAS-bHLH (per-arnt-sim, basic helix-loop-helix) family [46]. Only 70% of R7 cells express ss, and Rh4 is later induced in these same cells when they differentiate. The underlying R8 cells then turn on Rh6, which appears to be the default R8 opsin [47]. In those R7 that do not express ss, Rh3 is later induced as the R7 default state. However, these Rh3-expressing cells then instruct their R8 neighbor to express RH5 (see Fig. 4). It appears that the presence of ss causes Rh4 expression and prevents R7 from sending a signal to R8, whereas the absence of ss allows the R7-to-R8 signal to occur. Thus, ss is necessary and sufficient for the yR7 (Rh4, Rh6) subtype fate. The stochastic expression of spineless in a subset of R7 photoreceptors is the primary event that leads to the creation of the retinal mosaic of rhodopsin expression used for color vision.

A Bistable Feedback Loop Specifies R8 Photoreceptor Subtype and Expression of Rh5 and Rh6

Strict pairing of rhodopsins in p and y R7/R8 subtypes could result from either a genetic program intrinsic to a particular ommatidium or from signaling events between

Fig. 4. Mutually exclusive expression of color-sensing rhodopsins in R8. Schematic model for how the double-negative feedback loop directs rhodopsin expression in R8 photoreceptor. When R8 receives the signal from Rh3 expressing R7, the growth gene melted is expressed and suppresses warts, allowing for expressing of Rh5. If the signal is not received in R8, the tumor suppressor warts is expressed, which represses melted and promotes expression of Rh6. Thick black lines represent occurring genetic actions; thin gray lines represent genetic actions not occurring [48, 52]. Rh rhodopsin, ss spineless.

R7 and R8 that coordinate subtype specification. When R7 cells are genetically removed in sevenless mutant flies, all R8 express Rh6 [17, 18], suggesting that an R7 cell is required for Rh5 but not for Rh6 expression in R8. An R7-to-R8 signal model was thus proposed to explain the coordination between the rhodopsins in R7 and R8. How does R8 interpret the R7 signal and select its subtype as defined by expression of Rh5 or Rh6? While the identity of this signal remains elusive, genetic experiments have implicated the tumor suppressor warts/D-lats and a growth regulator melted in the postmitotic specification of R8 subtype fate [48] (Fig. 4).

Warts is a serine/threonine kinase and the Drosophila homolog of the human large tumor suppressor (LATS) tumor suppressor genes that function to coordinately regulate proliferation and apoptosis in developing tissues [49, 50]. Warts misexpression in all photoreceptors induces the expression of Rh6 in all and only R8 cells. Conversely, warts loss of function leads to expression of Rh5 in all R8 [48]. This strongly suggests that warts is necessary and sufficient for the yR8 fate and for Rh6 expression.

Whereas warts limits growth, melted is involved in fat metabolism as a positive regulator of the insulin/TOR (target of rapamycin) signaling pathway [51]. In R8, it opposes warts function to control Rh5 expression. Misexpression of melted and melted mutant