A new rhodopsin in R8 photoreceptors of Drosophila: evidence for coordinate expression with Rh3 in R7 cells

The insect eye consists of a large number of elementary units called ommatidia. Each ommatidium is organized as a structurally and functionally defined group of photoreceptors. The sophisticated organization of photoreceptors in each ommatidium mediates several functions such as image formation, color vision and sensitivity to polarized light.

One of the major criteria for distinguishing functionally different photoreceptors is the type of connections to the brain. This difference allows the grouping of the eight photoreceptor cells of the Drosophila ommatidia into two subsystems which project to two different optic lobes (Kirschfeld and Franceschini, 1968).

The outer photoreceptors R1-R6, which are mainly responsible for image formation contain a visible light (495 nm)-sensitive opsin, Rh1. They have large rhabdomeres spanning the entire thickness of the retina layer. R1-R6 axons project to the lamina, which is their first synaptic target in the optic lobe. Their large rhabdomeres capture light with high efficiency and work in a broad range of light intensities. Thus, functionally, one can compare them to the rod photoreceptors of the vertebrate retina.

On the contrary, the inner photoreceptor cells R7 and R8 are located one on top of the other, thus occupying only the apical (R7) or the basal (R8) half of the retinal layer. These cells have smaller rhabdomeres and may represent a high acuity system (For review see: Hardie, 1985). Their axons project to the medullar region of the optic lobe.

The stratified organization of the photoreceptor cells in the retina of Drosophila may be considered ancient in evolution since it is also observed in distant and primitive insects such as cockroach (Blattodea) and springtail (Collembola) as well as in primitive arthropods (For review see: Autrum and Thomas, 1983; Goldsmith, 1964; Goldsmith and Bernard, 1974; Menzel, 1983). It is possible to identify combinations of R7/R8-like cells in many species (honeybee, locust, dragonfly (Wolken, 1975)). One explanation for the evolutionarily conserved morphological coupling between R7 and R8 is that they may have evolved from a primitive system designed to compare sensory input between two cells in order to detect polarized light. This configuration is found in the dorsal marginal region of the Drosophila retina which is specialized in detection of the polarized light and serves for navigation (see: Hardie, 1985). In this region, both R7 and R8 contain the same opsin, Rh3, and their rhabdomeres have orthogonal orientation of their microvili which harbor Rhodopsin (Wunderer et al., 1989). The R7 and R8 cells in the remaining part of the Drosophila retina express distinct rhodopsins. They are likely to have evolved later and to participate in color detection and phototactic behavior (Fischbach, 1979).

There are four known rhodopsins genes in Drosophila, the expression of which obeys a common rule — ‘one receptor cell, one type of receptor molecule’. Besides the visual systems of vertebrates and invertebrates, this phenomenon is also observed in the other sensory systems such as the vertebrate olfactory system (Ngai et al., 1993; Ressler et al., 1994). In Drosophila Rh1, encoded by the ninaE gene, is present in the outer R1-6 photoreceptors. rh2 is exclusively expressed in ocelli, which represent a distinct visual system (Pollock and Benzer, 1988). rh3 and rh4 are expressed in R7 (Fortini and Rubin, 1990), while no opsin was known in the R8 photoreceptors. An interesting property of the R7 cells is the mutually exclusive expression of rh3 and rh4. rh3 is expressed in 30% of the ommatidia (also called p) while rh4 is expressed in the remaining 70% (called y), indicating that there are two distinct classes of ommatidia with a stochastic distribution (Kirschfeld et al., 1978). In fact the system appears quite flexible. For instance, expression of a rh1-like gene in both R7 and R8 cells of the central part of the male eye of Musca has been shown to be important for detection of the flying female (Franceschini et al., 1981; Hardie, 1985). The biological significance of the y and p subclasses is not clear, but they may represent a highly advanced system mediating color image formation. It has also been suggested that the presence in R7y of a filtering pigment that absorbs in the same wavelength as Rh4, i.e. in the blue, may protect the R7y cells from very high intensity light (Hardie, 1985). A sensitizing pigment is also present in these cells, which shifts the wavelength from the to the blue. It is intriguing that the same sensitizing pigment is present both in the rh4-expressing R7 and in the underlying R8y cells. These pigments are absent in the R7p cells, which detect UV light directly through Rh3. rh1, rh3 and rh4 are all also expressed in the larval visual system (Pollock and Benzer, 1988), the Bolwig organ, which is a likely derivative of the ancestral ommatidia (Paulus, 1988). It is not clear whether any of these genes are co-expressed in the same cell in the Bolwig organ.

Until now, no rhodopsin had been detected in R8 cells. However, electroretinogram recordings in rdgB, sevenless (sev) double mutants, and spectrophotometric studies in ora, sev double mutants, which lack both the R1-6 (rgdB, ora) and the R7 (sev) photoreceptor cells indicate that there is a photopigment system with multiple forms in R8 (see Hardie, 1985). This blue-green photopigment system is likely to represent one (or several) R8-specific rhodopsins. In an attempt to understand the regulation of non-overlapping rh3 and rh4 expression, and the correlation with the presence of two morphologically and physiologically different classes of R7/8 p and y cells, we cloned a new R8-specific opsin, rh5. Strikingly, this gene is expressed in the R8p, the subset of R8 cells that are associated with rh3-expressing R7 cells. Thus, the definition, based on functional criteria in larger flies, of the two classes of p and y ommatidia with different R7/8 pairs corresponds to the expression of two sets of mutually exclusive rhodopsin genes in Drosophila. The striking molecular coordination between expression of rh3 in R7 and rh5 in R8 suggests that there is cellular coordination between these two cells and that signaling takes place either early, at the time of cell fate determination of the photoreceptors, or later, when the rhodopsin genes are first expressed. Interestingly, we also found that rh5 expression disappears in sev mutants, which lack R7 cells. Recently, another group has also identified the same gene (Chou et al., 1996).

To find the rhodopsin gene(s) from R8 photoreceptor cells, we used a PCR-based approach. We generated six divergent primers against four highly conserved regions in dipteran Rhodopsins. Those regions were found by computer alignment using Megalign (DNAstar). The first conserved region is a part of the second extracellular loop: VPEG(Y/N)LT, oligonucleotide: GTICCIGARGGITAYYTNAC (20-mer, degeneracy: 4,096). The second conserved region is part of the third cytoplasmic loop: QAKKMNV, oligos: CARGCIAARAT-GAAYGT; reverse, ACRTTCATYTTYTTNGCYTG (20-mer, degeneracy: 64). The third conserved region is an internal part of the transmembrane helix VI: WTPY(G/L)V(M/I), oligos: direct: TGGACICCITAYGGNGTNAT; reverse, ATIACICCRTANGGNGTCCA (20-mer, degeneracy: 512). The last region was a part of the transmembrane domain VII: VY(A/G)ISHP(R/K)Y, oligo, reverse TAIGKIGGRTGIGAHATIGCRTANAC (26-mer, degeneracy:12,288).

Genomic DNA from D. melanogaster Canton S strain was used as a template in our PCR experiments with the divergent primers. The PCR conditions were as follows: 200 pM of each of the primer, 100 ng of genomic DNA, 1 u Taq polymerase in 100 ×l reaction; 35 cycles with denaturing step at 95°C for 30 seconds, annealing at 50°C for 45 seconds and extension at 72°C for 1 minute in a Thermocycler (Perkin Elmer GeneAmp 2400). All six possible combinations of the divergent primers were tested. The resulting fragments were size-selected by agarose gel and subcloned directly into the pCR ^™II vector using the TA-cloning system (Invitrogen).

After analysis of the clones obtained by all primer combinations (by comparison with the NCBI database), the presence of new sequences, highly related to opsin genes, was detected for each primer pair. The overlap of several fragments enabled us to establish the presence of one new rhodopsin gene. The largest fragment (0.45 kb), was labeled by T7 primer extension with the Klenow fragment of DNA polymerase in the presence of 32P dATP (New England Nuclear) and used to screen a D. melanogaster head cDNA library (λShlox-1, Novagen) by replica filter (MAGNA, Micron) hybridization. The hybridization conditions were as follows: 50% formamide, 6× SSC, 5× Denhardt’s solution, 0.1 mg/ml single strand heterologous DNA, 1-5 ×Ci/ml of radioactive probe. The clones were selected after hybridization and the pSHlox plasmid carrying the rhodopsin-like cDNA was excised from the λ phage according to the instructions of the manufacturer (Novagen). Analysis of the protein was performed by comparison with the database of known Rhodopsins.

A genomic library, λFIX-II (Stratagene) was used for isolation of the genomic region of the new rhodopsin 5 gene. The rh5 pSHlox plasmid, labeled as above, was used as a probe for the genomic library screening. Restriction mapping was performed by the FLASH hybridization procedure (Stratagene). The resulting positive clones were part of a single genomic region corresponding to three rh5 exons.

The genomic location of rh5 was established by hybridization of squashed polytene chromosomes with one of the λFIX-II genomic clones. Third instar Drosophila larvae were fixed in 45% acetic acid and the salivary glands were dissected. The polytene squashes were performed by the standard technique. The rh5 phage DNA was labeled by digoxygenin (DGG) in the presence of Klenow fragment (random-primer kit, Boehringer) and hybridized with the squashes under the following conditions: 50% formamide, 4× SSC, 0.2 mg/ml _ssDNA, 10 ng/×l of DGG-labeled phage DNA, 10-15 hours, at 37°C. Following hybridization, the slides were washed for 10 minutes at 37°C in 2× SSC, then in PBS-Tween 20, and in PBS. Finally, they were incubated with alkaline phosphatase-coupled anti-DGG antibodies. The signal was developed in the presence of BCIP/NBT(as a substrate for the alkaline phosphatase).

Drosophila were decapitated on CO₂ pad. Heads were embedded directly in OCT media and frozen in dry ice. Sections of about 10 ×m were transferred to slides and air dried for 30 minutes. They were fixed in 4% paraformaldehyde for 10 minutes, washed three times with 1× PBS, proteinase K (20 ×g/ml) treated for 5 minutes, washed three times with 1× PBS and postfixed with 4% paraformaldehyde for 10 minutes. After a brief wash with 1× PBS, the sections were prehybridized in hybridization solution (50% formamide, 5× SSC, 200 ×g/ml sonicated salmon sperm DNA, 100 ×g/ml tRNA and 50 ×g/ml heparin adjusted to pH 5.0) for 2 hours at room temperature. Digoxigenin (or fluorescein)-labeled probes were added to the hybridization solution and sections were hybridized at 70°C overnight. They were then washed briefly with 5× SSC,1 hour with 0.2× SSC at 70°C and 1× PBS at room temperature for 10 minutes, and incubated with anti-Dig-POD and anti-fluorescein-AP anti-bodies for 4 hours at room temperature before carrying out the color reactions.

The four known Drosophila opsin genes share significant homologies in several structurally and functionally important regions. This allowed us to use a PCR-based approach to identify new rhodopsin genes. We designed PCR primers that are able to recognize the known rhodopsin genes and are also likely to amplify unknown opsin genes. We generated six divergent primers against four regions highly conserved in dipteran opsins and amplified genomic DNA with these primers and each pair of primers amplified bands of expected lengths. We eliminated the sequences corresponding to known opsin genes by digesting the amplified bands with 6-cutter restriction enzymes specific for rh1, rh2, rh3 and rh4. Then, we re-amplified the digested bands, cloned and sequenced the fragments. All six chosen pairs of primers identified a new gene corresponding to an unknown opsin. The amplified fragment was used to screen an adult cDNA library, which led to the isolation of a 1.35 kb Cdna clone named λShlox-2.1. Screening of a genomic library with the λShlox-2.1 probe allowed the recovery of genomic λFIX-II phage clones, spanning approximately a 30 kb long region with the full gene and upstream promoter sequences. We called the new gene rhodopsin 5 (rh5).

Sequence analysis of the cDNA clone λShlox-2.1 indicated a single extended open reading frame, encoding a protein of 382 amino acids with predicted relative molecular mass of 42.8×103. The deduced amino acid sequence shared the highest homology (58.9% similarity) with the locust S. gregaria lo2 opsin (Gartner and Towner, 1995), 58.3% with Drosophila Rh4 and 55% with Rh3 opsins (Fig. 1). The seven putative transmembrane domains and the hydrophobicity plot (data not shown) were almost identical to those of the other opsins. No significant gaps or substitutions as compared to the homologues listed above were detected in the conserved regions. Finally, the highly conserved Lysine at position 322 in hydrophobic domain VII was also found. It represents part of the likely chromophore-binding site (Ovchinnikov, 1982). This enabled us to conclude that the rh5 gene that we had cloned encodes a new bona fide Drosophila opsin protein.

Rh5 shares common features with other rhodopsin proteins (Fig. 2). A conserved N-glycosylation site (NDS) was found at position 14, as in Rh2. The C-terminal cytoplasmic portion of the protein contained several Serine and Threonine residues, which may serve as sites for light-dependent phosphorylation. Finally, the two Cysteine residues found at positions124 and 201 are highly conserved and are thought to be involved in formation of a disulfide bond between extracellular loops II and III (for review see Gartner and Towner, 1995; Maden, 1995).

An unusual difference between Rh5 and related proteins was found in the second cytoplasmic loop, between transmembrane domains III and IV. This part of the protein has been shown to be important (along with the cytoplasmic loop between transmembrane domains V and VI) for G-protein binding (Arendit et al., 1989). Twelve amino acids in this region (positions 155-166) showed a very high homology (73.7% similar) with locust opsin Lo2, but only 26.3% and 31.6% similarity to Rh4 and Rh3, and even less to all other opsins (Fig. 3). This region was also very divergent from the other known locust opsin, Lo1 (21.1%). Unfortunately, the expression patterns of Lo1 and Lo2 are not known. However, if Rh5 and Lo2 are true functional homologues, it is quite attractive to consider their extremely high conservation in the mentioned cytoplasmic loop as an internal property of R8-specific opsins.

Comparison of sequences of the cDNA and genomic DNA revealed the presence of two introns. Intron 1 is 58 bp long; intron 2 maps to the same position as in the Drosophila rh1 and rh4 genes.

We used in situ hybridization on polytene chromosomes with a labeled rh5 genomic fragment (Fig. 4A) and mapped rh5 to the 33B region. Although there are several interesting genes in the region (oph, esc, prd), none has a phenotype consistent with the loss of a Rhodopsin. The rh5 gene does not map close to any other rhodopsin gene. In fact, all five known rh genes occupy distant locations (Fig. 4B) and none of them form a cluster. This strongly suggests that control of their mutually exclusive expression does not involve regulation of a cluster of genes through a locus control region (Wang et al., 1992).

Several features of rhodopsin promoters are highly conserved. In particular, all known rh promoters contain a bona fide TATA box and a site called RCS-1 (or P3) (Fortini and Rubin, 1990; Mismer and Rubin, 1989; Sheng et al., 1997), positioned 20-40 bp upstream from the TATA box. Both a TATA box and a P3/RCS-1 sequence could be unambiguously identified in the promoter region of rh5 (Fig. 5), allowing us to define the start of transcription within 10 bp, i.e. 55-65 bp upstream of the ATG. Another copy of a P3-like sequence was found 437bp upstream of the ATG. The P3 sites appear to be binding sites for the dimeric form of the Pax-6 homeodomain (Sheng et al., 1997). We also identified another element that was highly reminiscent of RUS3A (Fig. 5), a conserved element of the rh3 promoter (Fortini and Rubin, 1990). This sequences contains an element corresponding to a putative binding site for homeodomain proteins bearing a Lysine at the critical position 50 of their homeodomain (K50), such as Bicoid, Orthodenticle, Sine oculis and D-goosecoid (Cheyette et al., 1994; Goriely et al., 1996; Wilson et al., 1993).

Because the known opsin genes are expressed in the photoreceptors R1-6 (rh1), in two subsets of R7 (rh3 and rh4), in the Bolwig organ (larval visual system, rh1, rh3 and rh4) and in the ocelli (rh2, Pollock and Benzer, 1988), we were very interested to know whether the rh5 gene could be the missing rhodopsin specifically expressed in R8 cells. We performed in situ hybridization on adult head sections with an antisense riboprobe using the cDNA as a template. A strong signal was detected in the lower part of the retina, where the R8 photoreceptors are located, under the layer of R7 cells (Fig. 6A). This signal was only present in a subset of the R8 cells, in no more than 50% of the ommatidia (Fig. 6A-D). This is similar to the proportion of ommatidia that express rh3 in R7, and to the proportion of one of the two classes of R7/R8 (R7/R8p) identified by physiological methods in Musca and Callyphora (see Hardie, 1985). Thus, it appeared that we had cloned a new gene specifically expressed in a subset of R8 cells.

A fascinating question was whether rh5 was expressed in the subset of R8 cells associated with the R7 cells expressing rh3. For this purpose, we performed a double in situ hybridizations with rh5 and lacZ probes (Fig. 6B) in a fly line carrying a transgene expressing the lacZ reporter gene under the control of 343 bp from the rh3 promoter (Fortini and Rubin, 1990). We observed an exact correspondence between the R7 cells in the upper layer expressing lacZ (rh3) and the R8 cells in the lower layer expressing rh5 (Fig. 6B). By contrast, when we performed double stainings for the presence of rh4 and rh5 transcripts (Fig. 6C), we could observe complete non-overlapping expression. This was most clearly visible in tangential cross sections (Fig. 6D) where we could see both R7 and R8 cells from each ommatidium in the same section: a single R8 cell (brown), expressed rh5, and was surrounded by five ommatidia containing rh4-expressing R7 cells (blue), which did not express rh5 in their R8 cells. Thus, we conclude that the R7 and R8 cells in each subclass of ommatidia (p or y) differentiate in concert to express a coordinate pair of opsin genes.

The coordination of rh3 and rh5 expression in R7 and R8 cells, respectively, implies that there is communication between these two cells. The inductive mechanism by which R8 induces R7 cell fate determination during morphogenetic furrow progression is well documented (Rubin, 1991; Zipursky and Rubin, 1994). Interaction between the products of boss and sevenless mediates activation of the Ras pathway in R7 (Simon et al., 1991). However, if the sev-boss interaction is sufficient for the appropriate development of all R7 cells, there is no indication that it may be involved in the distinction between the p and y classes. Therefore, another signaling system between these two cells must exist, in which either R7 or R8 may induce subtype specification. Another alternative would be that an unknown system defines two classes of ommatidia, from the y or the p type, which is later interpreted by the R7 and R8 cells independently. To address this problem, we analyzed the expression of rh5 gene in a situation where R7 is absent. We used in situ hybridization to detect rh5 expression in the retina of sevenless flies (Fig. 7). Although we observed normal levels of rh5 expression in all sectioned wild type heads processed in parallel, no rh5 expression was detected in any of the sev mutant eyes. This indicates that the R7 cells are necessary at least for the formation of the R8p subtype and for the correct expression of rh5.

We have identified rh5, a new rhodopsin gene in Drosophila which is expressed in a subset of R8 photoreceptors (p). It is very likely that there is still another gene (rh6?) expressed in the remaining R8 (y) cells. The presence of the seven transmembrane domains as well as several other features highly conserved among opsins allow us to consider Rh5 as a true rhodopsin with all major characteristics of the photoreceptor molecules. This includes the retinal-binding site, the Cysteine residues allowing formation of a S-S bond between extracellular loops I and II, the putative sites of phosphorylation near the C-terminus of the protein and the consensus sequence for N-dependent glycosylation. The protein is most closely related to the Rh3 and Rh4 proteins of Drosophila, but is clearly distinct from them (Fig. 1). Phylogenetic relationships among Rh3, Rh4 and Rh5 suggest that rh3 is a primitive gene from which the R8 rhodopsin evolved (rh5). If this was the case, formation of the subclasses of ommatidia (rh3 or rh4 expressing R7) would be the latest evolutionary acquisition. In fact, the closest relative of rh5 is lo2, a gene isolated from locust S. gregaria (orthoptera) that shares interesting features with it and is likely to share R8-specific features. In particular, the high conservation of their second cytoplasmic loop (Fig. 3) is unexpected. This region diverges extensively from that of all other opsins. Such a large difference in sequence of a functionally important region may cause differences in binding of G-proteins in the R8 cells, as compared to R1-7. This difference may affect the whole signal transduction cascade and the sensitivity of the R8 photoreceptors. It may also suggest binding of a different G-protein.

A very extensive dissection of the promoters of rh1,rh3 and rh4 in Drosophila, both in melanogaster and in virilis, has provided a good understanding of the organization of the various rhodopsin promoters (Fortini and Rubin, 1990; Mismer and Rubin, 1989). They appear to contain three types of elements: a basal promoter element, composed of a consensus TATA box, found in general at –33bp from the start of transcription; a photoreceptor-specific element, called RCS1 or P3 (Wilson et al., 1995); finally, an element allowing expression in subtypes of photoreceptors (RUSx; Fortini and Rubin, 1990). Recently, we have demonstrated that the P3-like sequence is the site of action of the Pax-6 transcription factor (Sheng et al., 1997). Although Pax-6 is considered as the ‘master regulator’ of eye development because of its extensive conservation in sequence and genetic function from flies to mammals, this conservation was highly unexpected since the classical concept of the phylogeny of the photoreceptive organs is that they have evolved independently at least 40 times (Salvini-Plawen and Mayr, 1977). The presence of a conserved Pax-6-binding site in the promoter of opsin genes suggests that its initial role was to control photoreceptor-specific gene expression (Zuker, 1994, Sheng et al., 1997). We therefore expected to find a P3 site just upstream of the TATA element, a location that is highly conserved in all Drosophila opsin genes. The function of a second P3-like site found 437 bp upstream of the initiating ATG is not clear. So far, no such site has been found in rh promoters which, for the most part, require less than 300 bp for full expression (Fortini and Rubin, 1990). Finally, the putative K50 homeodomain-binding sites also conserved in rh3 may represent a feature of the rh3/rh5 subclass (p). This site may also be the site of action of a R7/8p-specific factor. Alternatively, since rh3 is the opsin expressed in the R8 cells in the dorsal margin (where it is expressed in both R7 and R8 in order to detect polarization of light; Hardie, 1985), these TAATCCC sites may be the target for a homeoprotein specific to the R8 cells, both in the dorsal margin and in the main part of the retina. Analysis of the promoter of the other R8-specific opsin (rh6?) will test these models.

One of the most conserved feature of the photoreceptors of different arthropods is the organization into two systems, which project to different regions of the optic lobe. If the shortfiber system is quite homogeneous, the long-fiber system appears to be more specialized and to fall into several subclasses. A possible biological implication of the coordinate expression of sets of opsin genes in R7 and R8 cells in each particular subclass is the ability to compare information between the two superimposed cells. This speculation seems to be reasonable since both cells project to the same region of the optic lobe (medulla) where computing inputs coming from R7 and R8 is necessary, for instance for detecting polarized light. The role of the two subclasses of ommatidia in the main part of the retina is not clear. A simple explanation for the coordinate expression between rh3 and rh5 (rh4 and rh6?) is that absorption by the upper cell (R7) restricts the light spectrum available to the underlying R8 cell and thus dictates the nature of the possible photopigments in R8. Since the R7 cells of the two subtypes have different absorption spectra (due to the presence of accessory pigments in R7y), this requires the coordination of the sensitive pigments in R8. Otherwise, the stochastic distribution of photopigments in R8 under the different screening shields of the R7 cells might cause misinterpretation of visual information. Obviously, additional information about the development and arrangement of photoreceptor cells is required to understand the biological meaning of this coordination.

Considering the conservation of the Rhodopsin molecules (see Fig. 1) and of the retinal structures, it is possible to propose two alternative modes of concerted evolution between photoreceptors and rhodopsins. In the first model, the dorsal margin of Drosophila has the most primitive configuration. This system is poised to detect polarized light by comparing the inputs of R7 and R8 cells with orthogonal oriented microvili, but containing the same, UV-sensitive Rhodopsin (Rh3) (Fortini and Rubin, 1991; Hardie, 1985). This system has evolved to give the current organization found in the main part of the fly retina, where different opsins are coordinately expressed in subclasses of ommatidia. In this case, polarized light is no longer detected (the microvili are not orthogonal and the rhabdomeres are twisted), but the division of ommatidia into subclasses with coordinate rh expression has allowed acquisition of a new function, which is likely to be color vision. Thus, it is probable that, from the ancestral opsin, initially expressed in all inner photoreceptors, new opsins have evolved in R7 and R8 cells. Since Rh3 and Rh4 are more similar than Rh3 and Rh5, they must represent the product of the latest gene duplication event. This means that the functional distinction between R7 and R8 preceded the division of ommatidia into subclasses. An alternative model suggests the recruitment of the R8-specific opsin before specialization of the ommatidia of the dorsal margin. The ancestral ommatidium, in this case, had different photopigments R7 and R8 cells.

However, a common features to all photoreceptors is the mechanism for transcriptional exclusion among rh genes This phenomenon of exclusion, which is widespread in sensory systems (e.g. the olfactory system) is very common among opsin genes from other species, including vertebrates. For instance, the old world primates have the green and red opsin genes located within 10 kb. They are controlled by a common Locus Control Region (LCR) (Maden, 1995; Wang et al., 1992) which allows expression of a single gene of the cluster. Allelic exclusion, which is often essential to avoid expression of two different genes of a single cluster from the two chromosomes (e.g. for the olfactory receptor genes; (Chess et al., 1994; Wong et al., 1993), is obtained by X inactivation (the cluster is on the X chromosome). The dispersion in the chromosome of the Drosophila rh3, rh4 and rh5 genes strongly suggests that regulation of their exclusive expression is unlikely to involve control by an LCR.

The most striking observation reported here is the fact that rh5 is expressed in a subset of R8 cells, which corresponds exactly to the R7 cells expressing rh3. This makes it very likely that there is a mechanism for synchronization of rh gene expression between the two cells. Two possibilities exist. In the first one, two subclasses of ommatidia would be specified early, at the time of the determination of the eight photoreceptor clusters. They would later on express either the rh3/rh5 pair, or the rh4/putative rh6 pair. It must be noted that rh gene expression occurs at least 4 days later than photoreceptor determination. In favor of this model, enhancer trap lines have been identified that are expressed in a subset of R8 cells just posterior to the morphogenetic furrow (Treisman et al., 1995; Jessica Treisman, personal communication). Unfortunately, their expression does not last long enough to overlap rh expression. This may either represent an early distinction of different R8 cells during photoreceptor differentiation, or may be simply due to variable expression of a weakly expressing line.

The second model is that the distinction between the two types of R7/R8 pairs appears late, when the rh genes are first transcribed. Expression of a specific rh gene (in R7 or in R8) may be a stochastic event, which is reinforced by an unknown mechanism, while expression of the other rh gene would be repressed in this cell. Communication between the R7 and R8 cells expressing specific rh pairs (at least one signal going from R7 to R8, as shown by loss of rh5 expression in sev mutants) would permit coordination of expression of the rh genes. It is possible that the Rhodopsin molecules themselves mediate this communication, in a manner that is similar to what has been observed for the olfactory system. The olfactory receptor molecules serve as guidance signals for projection to specific targets in the olfactory bulb (Mombaerts et al., 1996).

Although the absence of rh5 in sevenless mutants does not provide evidence in favor of either one of the models, it may reflect the existence of a positive feedback loop in the interaction between the R8 and R7 cells. Since the R8 cells do not develop properly (rh5 is not expressed) in the absence of R7, a signal from R7 must be necessary to complete retinal development and for the formation of the different subclasses of ommatidia. The sevenless-boss pathway, or some part of it, may be involved in this interaction. Alternatively, a totally distinct pathway may lead to coordination of rh gene expression in R8. It is interesting to mention that physiological data indicate that there are still functional R8 cells in sevenless mutants (see Hardie, 1985). Indeed, a photopigment could be detected in the retina of flies lacking both functional R1-6 cells (ninaE mutants) ands R7 cells (sev mutants). Thus, an unknown rhodopsin (rh6?), which is predicted to be expressed in the remaining subset of R8cells (R8y; in concert with rh4 in R7y), could be expressed in all R8 cells in sevenless flies. If this is true, the R7 cell controls the expression of rh5 in R8, and thus the determine of the particular subclass (p) of ommatidia. In other words, the decision for transcriptional exclusion between rh3/5 and rh4/6 would occur in R7, downstream of the bosssev signaling and upstream of the suggested feedback control from R7 to R8. Finally, it must be noted that expression of rh3 in R8 cells of the dorsal margin is not affected in sevenless mutants (Feiler et al., 1992). This indicates that at least some of the R8 development is not affected by R7 cells.

In conclusion, the discrimination between rh3 and rh4 (and between rh5 and putative rh6) expression, and the coordination between pairs of rh genes in neighboring photoreceptors provides a powerful paradigm for studying not only the mechanism of transcriptional exclusion often found in sensory systems, but also for studying the concerted evolution of rhodopsin genes in achieving specific functions such as color vision.

We are very indebted to Ulrike Gaul for her constant support and for her input into the project. Her intellectual and technical contribution were invaluable. We would like to also thank Charles Zuker for his very helpful support and for discussions, Roger Hardie, and Gerry Rubin for communicating their deep knowledge of the eye and of rhodopsin, Paul Towner for help with the locust rhodopsin, and John Pollock for discussions and for rh probes. We are very grateful to Jessica Treisman for all her constant availability for discussing the work and for revising this manuscript. We also thank Jean-René Huynh and Boris Lorberg for significant help with the cloning, and to Philippe Beaufils and Ali Tahayato for sharing many excitements in this project. Thanks to the other members of the Desplan and Gaul laboratories for discussions and support. Finally, we are also grateful to Steve Britt for his good spirit during discussions about this work. Accession numbers: U80856 DMU 80855 (promoter), U80667 DMU 80667 (cDNA).