Ectopic expression of seven-up causes cell fate changes during ommatidial assembly

How the diversity of neurons is generated during neurogenesis is one of the central questions in developmental biology. The compound eye of Drosophila offers an excellent model system for studying the genetic control of specification of neuronal identities. The Drosophila eye is a hexagonal array of ∼800 ommatidia, or unit eyes, each containing 8 photoreceptor neurons, R1 through R8, and 12 non-neuronal accessory cells. Individual photoreceptor neurons can be uniquely identified by their morphology and the stereotyped position that they occupy (reviewed by Tomlinson, 1988; Ready, 1989). Since no lineage restrictions exist within the 20 cells that constitute an ommatidium, cells acquire their identity by responding to signals from neighboring cells within the ommatidium (Ready et al., 1976; Lawrence and Green, 1979; Wolf and Ready, 1991). A number of genes have been identified that are required for correct specification of cell fates during ommatidial assembly. Through mosaic analysis one can identify the cells in which a gene activity is required, and deduce its role in the cell-cell interactions that mediate cell fate decisions. Molecular features of the genes isolated so far are consistent with the proposed cell-cell interactions and induction mechanisms (reviewed by Banerjee and Zipursky, 1990; Hafen, 1991; Rubin, 1991)

The specification of the R7 neuron has been studied in the most detail and is at present the best understood (reviewed by Rubin, 1991). R7 is the UV-sensitive photoreceptor that synapses in a layer of the optic lobe distinct from all other photoreceptor neurons in each ommatidium (reviewed by Hardie, 1986). In normal development, the R7 neuron differentiates from a cell that occupies a fixed position in the ommatidial cluster, between R1 and R6. Several genes have been identified that are involved in selecting a single photoreceptor precursor as the R7 photoreceptor. Loss-of-function alleles of sevenless (sev), bride of sevenless (boss) and seven in absentia (sina) all transform the R7 precursor cell to a non-neuronal cone cell (Tomlinson and Ready, 1986, 1987b; Reinke and Zipursky, 1988; Carthew and Rubin, 1990). The sevenless gene encodes a receptor tyrosine kinase (Hafen et al., 1987; Bowtell et al., 1988; Basler and Hafen, 1988; Simon et al., 1989), whereas boss encodes a transmembrane protein that is expressed in the R8 cell and is the ligand for sev (Hart et al., 1990; Kramer et al., 1991). Genetic analyses indicate that the sev tyrosine kinase acts through activation of the ras pathway (Simon et al., 1991; Fortini et al., 1992; Rogge et al., 1991; Bonfini et al., 1992; Gaul et al., 1992). Increased levels of ras activity in the cone cells achieved by either a ligand-independent allele of sev (Basler et al., 1991), ectopic expression of boss (Van Vactor et al., 1991), expression of activated Ras1 (Fortini et al., 1992), or by reduction of the activity of Gap1 (Gaul et al., 1992; Rogge et al., 1992; Buckles et al., 1992) all result in transformation of cone cells to R7 neurons. In contrast, loss-of-function alleles of rough (Tomlinson et al., 1988; Heberlein et al., 1991; Van Vactor et al., 1991) and seven-up (svp) (Mlodzik et al., 1990b) cause more than one photoreceptor precursor to adopt the R7 fate without affecting cone cell differentiation. In particular, loss of svp⁺ function results in the cell autonomous transformation of four outer photoreceptor cells, R3/R4/R1/R6, towards R7-like cells (Mlodzik et al., 1990b).

The predicted svp protein shares homology with members of the steroid receptor family (reviewed by Evans, 1988; Green and Chambon, 1988), suggesting that it acts as a ligand-responsive transcription factor (Mlodzik et al., 1990b). Two human homologues of svp have been identified that share extensive homology in both the DNA-binding domain and the ligand-binding domain (Miyajima et al., 1988; Wang et al., 1989; Ladias and Karathanasis, 1991). A striking aspect of svp expression is that, despite its apparent structure as a receptor, there is complete coincidence between the cells that express svp and those that require its function (Mlodzik et al., 1990b). This is in contrast to the sev receptor tyrosine kinase, which is required only in the R7 cell but is expressed in most photoreceptor precursors as well as cone cells (Tomlinson et al., 1987; Banerjee et al., 1987). The expression pattern of sev does not play a major role in restricting R7-forming potential, since ectopic expression of sev in all cells under heat-shock promoter does not cause other cells to adopt the R7 cell fate (Basler and Hafen, 1989a; Bowtell et al., 1989a). The restriction of sev activity is achieved by local presentation of its ligand, the boss protein by the R8 cell. There is also a restriction in the ability of sev-expressing cells to internalize the boss protein to the R7 precursor, but the significance of this restriction in controlling R7-forming potential is not clear (Kramer et al., 1990; Van Vactor et al., 1991; Cagan et al., 1992). The ligand for svp is not identified, nor its distribution known.

Here we have tested the functional significance of the svp expression pattern by analyzing the consequences of ectopic expression of svp. We observe a variety of cell fate transformations within an ommatidium including both the loss and gain of R7 cells. Our results indicate that the spatially restricted expression of svp plays an essential role in controlling the number of R7 cells that form within an ommatidium.

P-element constructs containing hs-svp1 and hs-svp2 genes were made by cloning a 1.7 kb EagI fragment of pc162.1 and a 2.4 kb EagI-ClaI fragment of pc162.2 (Mlodzik et al., 1990b), respectively, into the polylinker region of the CaSpeRhs vector (Thummel and Pirrotta, 1991). P-elements containing sev-svp1 and sev-svp2 genes were made by first inserting a 0.57 kb BamHI fragment of CaSpeR-hs containing the trailer sequence of the hsp70 gene into the BamHI site of SE8/DM30 (Bowtell et al., 1989b) and then inserting the 2.9 kb EcoRI-ClaI fragment of pc162.1 and the 2.7 kb EcoRI-ClaI fragment of pc162.2, respectively, into the ClaI site located upstream of the BamHI site in the SE8/DM30 vector. sev-svpΔMlu has an insertion of a stop codon linker (New England Biolabs) at the MluI site of sev-svp2, truncating the protein at residue 273. sev-svpΔSal has a deletion of a 1 kb SalI-ClaI fragment truncating the protein 17 residues before the divergence point of the two isoforms.

Germ-line transformation was done using ry⁵⁰⁶ and w¹¹¹⁸ as host strains and pπ25.7wc (Karess and Rubin, 1984) as a helper plasmid. Secondary jumps to new locations were made using a strain carrying a genomic source of transposase activity (Robertson et al., 1988).

Antibody stainings of imaginal discs were performed as described (Tomlinson and Ready, 1987a) except that in most cases the peripodial membrane was not removed. Affinity-purified rabbit antibody against BarH1/BarH2 proteins (Higashijima et al., 1992) was a kind gift of K. Saigo. Monoclonal antibody anti-boss1 (Kramer et al., 1991) was a generous gift of L. Zipursky. Monoclonal antibody against β-galactosidase was purchased from Promega. Monoclonal antibody against elav protein was made in the Rubin laboratory monoclonal antibody facility. Sections of adult retinae were made according to Tomlinson and Ready (1987a).

The following enhancer trap marker lines were used as cell-type-specific markers; BB02 (Hart et al., 1990) and rO156 (U. Gaul unpublished), which express β-galactosidase late in R8 differentiation, H214, which expresses β-galactosidase strongly in the R7 cell (Mlodzik et al., 1992), and rI533, an insertion in the Gap1 gene (Gaul et al., 1992).

Clones homozygous for a null allele of svp, svp^e22 (Mlodzik et al., 1990b), were made by mitotic recombination catalyzed by the FLP-FRT system (Golic and Lindquist, 1991). A 24 hour egg collection was taken from the crosses between males of w¹¹¹⁸ hsFLP1/Y; 75AE svp^e22/TM6B Hu, Tb and females of w¹¹¹⁸; 75A M(3) w¹²⁴/TM6B Hu Tb. hsFLP1 has an insertion of the P[ry⁺;hsFLP] element on the X-chromosome (Golic and Lindquist, 1989). 75A is an insertion of P[>w^hs>] element (> denotes an FRT site) at the base of 3R (Golic and Lindquist, 1991; K. Golic personal communication). 75AE has an excision of the w^hs gene catalyzed by FLP-mediated recombination from the 75A insertion and has one copy of FRT left. 40 hours after the end of the egg collection period, the vials containing larvae were submerged in a 37°C waterbath for 2 hours. Female Tubby⁺ larvae were selected as wandering third instar and their discs were processed for antibody staining. Tubby⁺ male siblings were used as controls that do not produce svp mutant clones. The svp mutant clones were visualized using an anti-BarH1/BarH2 antiserum as a molecular probe.

Neuronal differentiation of the eye starts in the eye imaginal disc of the third instar larva as an indentation of the disc, called the morphogenetic furrow, moves in a posterior-to-anterior direction. Pattern formation starts in the furrow as cells are recruited to form ommatidial clusters containing photoreceptor precursors (Tomlinson and Ready, 1987a; Wolf and Ready, 1991). From the morphogenetic furrow, rosettes of cells consisting of four to five core cells surrounded by a ring of 10–15 cells emerge with regular spacing. Each group of cells is then transformed into a precluster containing five postmitotic photoreceptor precursors, R8, R2, R5, R3 and R4. After a wave of mitosis, three more photoreceptor precursors R1/R6/R7 and four cone cells join the cluster successively. Each of the eight photoreceptor precursors and the cone cells occupies a stereotyped position within the cluster. Based on expression of neuron specific antigens, it has been inferred that photoreceptor precursors initiate neuronal differentiation in an invariant sequence; R8 initiates differentiation first, followed by R2/R5, R3/R4, R1/R6 and finally R7 (Tomlinson and Ready, 1987).

Since an eye disc contains ommatidial clusters at different stages of differentiation, it should be possible to identify the developmental stage that is sensitive to the ectopic expression of svp by applying a pulse of svp expression throughout the eye disc. Two classes of cDNAs, called type 1 and type 2, have been identified from the svp locus. Type 1 cDNA encodes a protein that shares homology with both the DNA-binding domain and the ligand-binding domain of steroid receptors, whereas type 2 cDNA diverges from type 1 in the middle of the putative ligand-binding domain (Mlodzik et al., 1990b). To express these two svp isoforms in all cells in the imaginal disc, we generated transformant lines that carry P elements containing svp cDNAs that were placed under the control of the heat-inducible hsp70 promoter (Fig. 1). Fusion genes employing type 1 and type 2 cDNAs will be called hs-svp1 and hs-svp2, respectively.

When late third instar larvae carrying these fusion genes were exposed to a brief (1-2 hours) heat pulse, 5 to 20% of the animals failed to pupate or died as early pupae. Those that survived to adulthood had no obvious defects in external morphology, except that their eyes had a stripe of rough region running dorsoventrally. In retinal sections of such eyes, a stripe of ommatidia with abnormal numbers of photoreceptor cells was found, the stripe often being wider than the region of the rough exterior. The width of stripe with abnormal ommatidia was often more than 10 rows. Since a new row of ommatidia is produced approximately every 2 hours (Basler and Hafen, 1989b), if ubiquitous expression of svp affected a single developmental stage, we would have expected to see a relatively narrow stripe of abnormal ommatidia, e. g. one or two rows. The broadness of the affected region suggests either that the svp protein expressed under heat-shock control has a long perdurance, or that ectopically expressed svp interferes with multiple differentiation steps.

In retinal sections of wild-type eyes, three classes of photoreceptor cells can be distinguished by their morphology. The outer photoreceptor cells R1 through R6 have rhabdomeres of large diameter that project throughout the depth of the retina, whereas the two classes of central photoreceptor cells, R7 and R8, have rhabdomeres of small diameter, the former located in the apical retina, the latter in the basal retina. Within the stripe of abnormal ommatidia in heat-shocked hs-svp flies, ommatidia with different phenotypes were observed: these include loss of outer photoreceptor cells, loss of central photoreceptor cells, appearance of extra central photoreceptor cells and appearance of extra outer photoreceptor cells (Figs 2, 3). Ommatidia with different phenotypes appeared ordered in an anterior-to-posterior manner, forming narrow substripes within the broader stripe of abnormally constructed ommatidia (Fig. 3). For example, in a hs-svp1 eye, loss of one or two outer photoreceptor cells was seen in the anteriormost region of the stripe, partially overlapping a region of ommatidia that lacked the central photoreceptor cells R7 and R8. In the more posterior region of the affected stripe, ommatidia with extra central photoreceptor cells were present. This was followed by a row that contained mostly normal ommatidia, and further posterior was a region that lacked R7 but had an extra photoreceptor with the morphology of an outer photoreceptor cell (Figs 2B, 3D). Formation of substripes of ommatidia with specific phenotypes in a defined order suggests that each phenotype is caused by affecting a specific differentiation process that takes place in a stereotyped position in the developing eye imaginal disc. hs-svp2 retinae also contained substripes of ommatidia with mutant phenotypes similar to those seen in hs-svp1 retinae (Figs 2C, 3E-H). The order of specific substripes in hs-svp2 retinae, however, differed from that in hs-svp1 retinae (see for example the position of extra R7 phenotype, relative to that of the loss of R7 phenotype), indicating that superficially similar phenotypes are not necessarily caused by the same cellular mechanisms.

Previous analysis of loss-of-function phenotype of svp showed that svp prevents R3/R4/R1/R6 cells assuming the R7 cell fate (Mlodzik et al., 1990b). It is thus possible that ectopic expression of svp in the R7 precursor would prevent its differentiation and either transform it towards an R3/R4/R1/R6 cell fate or cause the loss of the R7 cell. Indeed, we found substripes of ommatidia that show such phenotypes in heat shocked hs-svp retinae (Figs 2, 3).

The posteriormost region of the affected stripe in hs-svp1 and hs-svp2 retinae contained ommatidia that lacked a photoreceptor with normal R7 cell morphology in the apical sections at the level that R7 cell is present in wild-type ommatidia. Even in such abnormal ommatidia, the cell corresponding to R7 can be identified by comparison with flanking normally constructed ommatidia. The rhabdomere of the affected R7 cells (R7T) are larger than those of the normal R7 cells, resembling those of the outer photoreceptor cells. Moreover the R7T rhabdomere was not in its normal central position, but is located between those of R1 and R6. Serial sections revealed that the R7T rhabdomere indeed extended throughout the depth of the retina, as those of normal R1-R6 subtypes (Fig. 2B,C,I,L). These phenotypes are indistinguishable from those observed in flies that express the rough protein in the R7 cell, transforming R7 cell into R1-R6 subtype (Basler et al., 1990; Kimmel et al. 1990). We conclude that ectopic expression of svp, like that of rough, causes transformation of the R7 cell to outer photoreceptor cells. Since the width of the ommatidial rows with this specific phenotype was approximately two rows, the R7 cell must be sensitive to ectopic expression of svp for at least 4 hours.

In addition to the transformation of the R7 cell to an outer photoreceptor cell described above, heat-shocked hs-svp1 and hs-svp2 retinae contained another substripe that lacked the R7 cell. This substripe was found two to five rows posterior to the anterior margin of the affected stripe (Figs 2, 3). Examination of basal sections revealed that most ommatidia that lacked R7 also were missing R8 (Figs 2,3). Of a total of 13 ommatidia arranged in a row in a heat-shocked hs-svp1 retina, we found 7 ommatidia that lacked a central photoreceptor in apical sections, all of which also lacked the R8 in basal sections (Fig. 2). Two possibilities exist as to how the R7 cell was lost. First, expression of svp in the R7 precursor might have prevented its differentiation as a R7 neuron in a cell autonomous manner. Alternatively, since R8 is known to induce R7 differentiation by expressing boss on its surface (Kramer et al., 1991; Van Vactor et al., 1991), the loss of R7 could be a consequence of the loss of R8. We explored the latter possibility first by examining whether or not boss expression was affected by ectopic expression of svp. Following a 1 hour heat shock, third instar larvae were reared at 22°C for various periods and then stained with an anti-boss monoclonal antibody (Kramer et al., 1991). In wild-type larvae that were heat shocked, as well as in non-heat-shocked hs-svp animals, boss protein could first be seen in R8 three rows posterior to the morphogenetic furrow and remained detectable towards the posterior end of the eye disc (Kramer et al., 1991). When eye discs from hs-svp larvae were fixed 2 hours after heat shock, no abnormalities in boss expression were detected. However, a chase of 9 hours or longer resulted in appearance of a stripe of ommatidia where boss expression was absent or greatly reduced (Figs 4, 5). Since boss has an essential role in the induction of R7, we expect failure of R7 differentiation to occur in such ommatidia. The width of the boss-negative stripe was much narrower than the stripe of abnormally constructed ommatidia seen in retinal sections but correlated with the width of the substripe that lacks central photoreceptor cells. This means that there is a narrow developmental window in which ubiquitous expression of svp results in the loss of boss expression.

The induction of R7 through a boss-sev interaction takes place eight to ten rows behind the furrow (Basler et al., 1989; Mullins and Rubin, 1991; Hart et al., 1991). We wished to determine whether ectopic expression of svp causes an early defect in the initiation of boss expression and/or R8 differentiation, or if it causes a loss of boss expression at the time that the boss-sev interaction occurs. The distance between the morphogenetic furrow and the boss-negative stripe increased with longer chase periods; chases of 12 and 20 hours produced boss-negative stripes 7 and 11 rows posterior to the furrow respectively (Fig. 5). This meant that a new row of ommatidia was produced every 2 hours, indicating that there was no significant effect in advancement of the morphogenetic furrow after heat-shock treatment, at least between these two time points. Extrapolation of such data indicated that the time that svp expression affected boss expression is one to two rows behind the furrow, anterior to the position at which boss protein could first be detected (Fig. 5). We conclude that ectopic expression of svp results in a failure to initiate boss expression, and that proper expression of boss in such cells is never recovered following heat shock.

In order to determine whether ectopic expression of svp resulted in specific repression of boss expression or a general defect in R8 identity, we examined the behavior of enhancer trap marker lines that express β-galactosidase in the R8 cell. Two enhancer trap lines BB02 (Hart et al., 1990) and rO156 (Ulrike Gaul, personal communication) express β-galactosidase in R8 from row 10 on throughout imaginal disc development, independent of boss function. Third instar larvae carrying these marker genes, as well as hs-svp1 and hs-svp2 transgenes were heat shocked for 1 hour and stained for β-galactosidase activity in 40-hour-old pupae. With both hs-svp1 and hs-svp2, there were one to two rows in which most ommatidia failed to express these R8 markers (Fig. 4D-F; and data not shown). With BB02, which expresses β-galactosidase also in R7 albeit at a lower level than R8, simultaneous loss of R8 and R7 was observed (Fig. 4E,F). Since multiple independent R8 markers showed loss of expression, it is likely that ubiquitous expression of svp isoforms resulted in a change in R8 identity.

To follow the fate of the affected R8 cell, we used the D120 line, which is an enhancer trap insertion in the scabrous gene (Mlodzik et al., 1990a). In this line, β-galactosidase expression starts in the R8 cell in the morphogenetic furrow, reflecting R8 specification (Mlodzik et al., 1990a; Baker et al., 1990). Although scabrous transcript and protein are expressed only in the morphogenetic furrow region, due to the stability of lacZ transcript and/or protein, β-galactosidase is detectable in R8 towards the posterior edge of the eye disc, thus serving as an excellent marker to follow R8 development. Larvae carrying the D120 insertion and either hs-svp1 or hs-svp2 genes were subjected to a 1 hour heat shock, and stained with anti-boss and anti-β-galactosidase antibodies after various chase periods. With 2 and 4 hours of chase, no abnormality in boss or β-galactosidase expression could be detected (data not shown). At 6 hours after heat shock, ommatidia that should have initiated boss expression failed to stain with boss antibody, yet all such ommatidia still contained R8 that expressed β-galactosidase (Fig. 6B,D). This means that the specification of the R8 precursor was normal and that the R8 cell was present despite its failure to initiate boss expression. With 11 hours of chase, the level of β-galactosidase in R8 in many of boss-negative ommatidia was either low or undetectable, although ommatidia further posterior still expressed β-galactosidase (Fig. 6D,E). Upon a 13 hour chase, all boss-negative ommatidia also failed to express β-galactosidase (data not shown). These results suggest that the R8 precursor undergoes cell death between 6 and 13 hours following the ectopic expression of svp.

The finding that ectopic expression of svp interferes with R8 differentiation prompted us to examine whether svp suppresses R8 differentiation in normal development. We asked whether the loss of svp function results in expression of R8 traits in photoreceptor cells other than R8 itself. Due to embryonic lethality of svp mutants, clones that were genotypically mutant for svp were generated in the eye imaginal discs. We found that in svp clones the BarH1/BarH2 proteins, which are normally expressed in R1 and R6 (Higashijima et al., 1992), are not expressed, consistent with transformation of their fates (Fig. 4G). Within such svp mutant clones, expression of boss was still restricted to a single cell within each ommatidium (Fig. 4H). We conclude that although svp can suppress R8 differentiation, it is not utilized or not essential for restriction of R8 differentiation during normal ommatidial development.

Experiments using hs-svp animals show that ubiquitous expression of svp causes transformation of R7 cell to outer photoreceptor cell fate at a specific stage during ommatidial development. To test if a similar phenotype is produced by specifically expressing svp in the R7 cell, transgenic animals that carry svp cDNAs under the control of sev regulatory elements were generated (Fig. 1). Sequences from the sev promoter and enhancer used in these fusion genes direct high levels of expression in R3/R4/R7 and in the cone cells and lower levels in R1/R6 (Bowtell et al., 1989b; Basler et al., 1990; Kimmel et al., 1990), thus achieving ectopic expression of svp in R7 and the cone cells. Seven lines carrying the sev-svp1 gene and nine lines carrying the sev-svp2 gene were examined for abnormality in ommatidial assembly. Six of the nine sev-svp2 lines showed mild roughening of the eye with one copy of the transgene. Surprisingly, sections of such retinae revealed that the roughening is not caused by the loss or transformation of R7 cells. On the contrary, some ommatidia had an extra photoreceptor with R7-like morphology, i.e. rhabdomere of small diameter located apically (Fig. 7A). When the copy number of the construct was increased, we did observe a small number of ommatidia with extra outer photoreceptor cells and others that did not have an R7 cell with R8 present basally. The majority of abnormal ommatidia, however, had extra R7-like cells, the frequency of such ommatidia and the number of extra R7 cells increasing with the copy number of the sev-svp gene (Fig. 7A,B). To establish the identity of the extra photoreceptor cells, we used a reporter strain where the promoter of the Rh4 rhodopsin gene, which is expressed exclusively in the R7 cells, has been fused to the lacZ gene (Fortini and Rubin, 1990). In sev-svp2 retinae, extra photoreceptors expressed this R7 specific trait indicating that the extra cells have an R7 identity (Fig. 8A,B).

There are two possible mechanisms that could generate this phenotype. First, expression of svp in R7 may have affected its differentiation, but such an effect is compensated by R7 differentiation from other cells. Alternatively, R7 development may not have been affected in these transgenic animals and the phenotype is caused solely by the formation of extra R7 cells. The fate of the R7 precursor and the origin of the R7-like cells were examined using an enhancer trap line H214 that expresses β-galactosidase at high levels in the R7 cell in the eye imaginal disc (Mlodzik et al., 1992). In sev-svp2 larval eye discs, not only the R7 cell but also the cone cells showed strong expression of β-galactosidase (Fig. 8D). We also observed expression of the neuron specific antigens elav (Robinow and White, 1991) and the 22C10 antigen (Zipursky et al., 1984) in the cone cells (Fig. 8G, and data not shown), establishing that cone cells were transformed to R7 cells in this genotype. In discs containing four copies of the sev-svp2 transgene, some nuclei located basally in the disc also expressed elav (Fig. 8I). These may be uncommitted cells that have not yet been recruited to the ommatidial cluster. We conclude that svp type 2 protein expressed under the sev enhancer/promoter does not interfere with R7 differentiation, but rather transforms cone cells towards R7 cells. Despite this transformation, most sev-svp2 ommatidia had four or more cone cells, as visualized with cobalt sulfide staining of pupal discs (data not shown). Thus, additional cells appear to be recruited to become cone cells when cone cells precursors enter the neuronal pathway. This is in agreement with other reports regarding the transformation of cone cell precursors (Basler et al., 1991; Gaul et al., 1992).

In contrast to sev-svp2 lines, none of the sev-svp1 lines showed any abnormality in ommatidial structure in single copy. When the copy number was increased to two or three, a phenotype similar to that seen in sev-svp2 lines was observed (Fig. 7C). Thus the effect of sev-svp1 appears to be weaker than that of sev-svp2.

Three genes, sev, boss and sina, are required for specification of the R7 precursor and in respective retinae all ommatidia lack the R7 cell. We examined whether these three genes are also required for the formation of R7 cells developing from the cone cells in sev-svp2 lines. In the presence of the sev-svp2 gene, many ommatidia contained one or two R7 cells even when sev or boss activity was removed (Fig. 7D, and data not shown). Such retinae contained ommatidia that expressed the Rh4/lacZ fusion gene, while in control eyes (sev or boss mutant alone) lacking R7, this R7 specific marker was never expressed (data not shown). On the contrary, in sina; sev-svp2 animals no R7 cells formed, indicating that sina function is required for R7 differentiation in transformed cone cells (Fig. 7E).

To analyze the fate of the R7 precursor and the cone cells in sev; sev-svp2 ommatidia, we examined expression of the H214 lacZ marker in larval eye discs. The cone cells expressed high levels of β-galactosidase as does the R7 cell in wild type, whereas the β-galactosidase level in the R7 precursor was low, if not undetectable, as in sev discs (Fig. 8E). This suggests that the R7 precursor failed to develop as the R7 cell in the absence of sev, whereas R7-like cells differentiating from the cone cells do not require sev activity. Indeed, in clusters that contained elav-positive cone cells, a majority of R7 precursors failed to express elav (Fig. 8H). To test whether the R7-like cells differentiating from cone cell precursors are functional R7 neurons, we tested the UV phototactic behavior of sev; sev-svp2 animals. Animals in which the R7 precursor and the cone cells have switched their fates showed normal phototactic behavior, indicating that transformed cone cells have normal functional properties of the R7 neuron (Fig. 9).

Three genetic situations are known that cause transformation of cone cells into R7: ubiquitous expression of boss that results in activation of the sev pathway (Van Vactor et al., 1991) expression of a boss-independent form of sev (Basler et al., 1991) and reduction of Gap1 activity (Gaul et al., 1992; Rogge et al., 1992; Buckles et al., 1992). Since in sev-svp2 discs, cone cells differentiate into R7 in the absence of boss or sev function, it is unlikely that svp causes its effect by regulation of boss or sev expression and/or function. A possible target for svp is Gap1, which shows highly regulated expression that is confined to photoreceptor cells and cone cells posterior to the morphogenetic furrow (Gaul et al., 1992). Using an enhancer trap insert in Gap1, the rI533 line, we tested whether sev-svp2 down-regulated Gap1 expression in the cone cells. In sev-svp2 discs, cone cells still expressed β-galactosidase (Fig. 8K), indicating that the effect of sev-svp2 was not mediated by repression of Gap1 transcription.

The observation that both isoforms of svp cause similar phenotypes when ectopically expressed raised the possibility that ectopically expressed svp may not require the putative ligand-binding domain for its function. Since many of the steroid receptors function as ligand-independent transcriptional activators when their ligand-binding domain is deleted (reviewed in Evans, 1988), we tested whether truncated svp proteins lacking the ligand-binding domain could produce the same phenotype as that caused by svp type 1 and type 2 isoforms. Three constructs were made: sev-svpΔMlu, in which sev regulatory elements direct expression of a svp protein truncated shortly after the DNA-binding domain, sev-svpΔSal, where svp protein is truncated immediately before the point where type 1 and type 2 isoforms diverge, and hs-svpΔMlu where the same protein as sev-svpΔMlu would be produced under the heat inducible promoter (Fig. 1). Animals that carry up to four copies of either the sev-svpΔMlu or sev-svpΔSal genes were generated, but their ommatidia showed normal morphology in retinal sections. We also failed to detect any phenotype after heat shocking animals carrying the hs-svpΔMlu gene (Fig. 7F and data not shown). Thus C-terminal portions of both type 1 and type 2 isoforms are necessary for production of dominant phenotypes in these misexpression experiments. This requirement for the C-terminal portion could be in the activity of the svp protein, the production of the stable product, or both.

Within an ommatidium, three cell types are known to have the potential to develop as the R7 neuron: the R7 precursor, the cone cells and the precursors of R1 through R6. During wild-type development, however, the R7 precursor is the only cell that adopts the R7 fate. Cone cells are prevented from becoming R7 first because they are not in contact with R8 and are thus unable to activate the sev pathway through sev-boss interaction and, second, because expression of Gap1 in cone cells reduces their intrinsic ras activity. R3/R4/R1/R6 are prevented from becoming R7 by the expression of the svp gene in these cells (Mlodzik et al., 1990b). Although svp⁺ function is normally not required in R2/R5, in rough mutant ommatidia svp is ectopically expressed in R2/R5 and is required to prevent them from becoming R7 (Heberlein et al., 1991). Our results of ectopic expression of the svp isoforms show that proper transcriptional regulation of svp expression is essential for the proper development of two other cell types that have potential to become the R7 neuron, the R7 precursor and the cone cells.

Since expression of svp is restricted to R3/R4/R1/R6, svp could act as a genetic switch between the R3/R4/R1/R6 neuronal type and the R7 type, in a manner similar to the homeotic selector genes. In such a case, one would expect ectopic expression of svp in the R7 cell to interfere with its differentiation, and possibly transform it towards the R3/R4/R1/R6 fate. Indeed, in both hs-svp1 and hs-svp2 animals, we found ommatidia in which the R7 cell was transformed to an outer photoreceptor fate. Curiously, we failed to observe similar transformation in animals carrying sev-svp transgenes, which express svp isoforms in the R7 cell under the control of the sev enhancer/promoter. There are a few possibilities that could explain this discrepancy. First, the R7 to outer photoreceptor transformation in hs-svp retinae may not be caused by ectopic expression in the R7 cell, but may be due to expression in other cells that influence the differentiation of the R7 precursor. Alternatively, under the sev enhancer/promoter control, the svp protein may not have been expressed in the R7 cell in the appropriate temporal pattern to effect this transformation. We favor the latter possibility because the stage that is susceptible to R7-to-outer photoreceptor transformation maps late in ommatidial assembly. In hs-svp1 retinae, the substripe containing transformed R7 was located posterior to the substripe containing extra R7 cells, which is likely to correspond to the cone cell-to-R7 transformation seen in sev-svp retinae. Since expression in R7 directed by sev enhancer/promoter sequences ceases before expression in the cone cells does (Bowtell et al., 1989b), expression of svp in R7 may not have persisted long enough to cause transformation of the R7 cell. In addition, in the case of the rough gene that specifies the R2/R5 cell fate, its ectopic expression in R7 alone is sufficient to transform R7 to the outer photoreceptor fate (Basler et al., 1990; Kimmel et al., 1990) and no data thus far available suggest a requirement for an inductive signal controlling the photoreceptor subtype decision of the R7 cell.

Another phenotype caused by ectopic expression of svp that interferes with R7 differentiation is the loss of R7 seen in hs-svp retinae. It is unlikely that this phenotype is caused by ectopic expression in the R7 cell itself, for the following reasons. First, the loss of the R7 cell is usually accompanied by the loss of R8. Concomitant loss of R7 and R8 suggests that these two events are not caused by independent effects on R7 and R8. Second, ubiquitous expression of svp affects expression of the boss protein which serves as a ligand for the sev receptor tyrosine kinase. Since activation of sev is essential for the R7 precursor to assume a neuronal fate, the effect on boss expression can alone account for failure of R7 differentiation. Third, the ectopic expression of svp in R7 directed by the sev enhancer/promoter does not affect R7 differentiation. Taken together, these data strongly suggest that the loss of R7 caused by the ubiquitous expression of svp is due to the loss of the R8 cell. Whether the loss of R8 is caused by expression of svp in the R8 cell, or due to a nonautonomous mechanism, is not known.

Since R8 is believed to play a central role in initiating a series of induction steps (Tomlinson and Ready, 1987a), it is rather surprising that, in an ommatidium lacking R8, assembly proceeds with minor effects on the induction of outer photoreceptor cells. The sensitive period for the effect on R8 is quite early, one or two rows posterior to the morphogenetic furrow (Fig. 5). We show, however, that hs-svp does not affect specification of the R8 cell, as visualized by the expression of an enhancer trap insertion in the scabrous gene, and that the loss of R8 is likely to be caused by cell death that occurs between 6 and 13 hours after heat shock. Anterior to the stripe of ommatidia that have lost R8, there is another substripe consisting of ommatidia with different phenotypes (Figs 2, 3). This implies that there is yet another step in ommatidial assembly that svp can interfere with, which takes place prior to the effect on R8 differentiation and boss expression. Therefore, it is likely that the R8 cell had already performed its function to induce other cells (e. g. R2/5) prior to the step that is affected by the ectopic expression of svp. Such a process must take place at or near the morphogenetic furrow, possibly prior to the formation of the 5-cell precluster (Banerjee and Zipursky, 1990).

Both the loss-of-function phenotype and the R7-to-outer photoreceptor transformation phenotype caused by hs-svp transgenes is consistent with the idea that svp acts by preventing R7 differentiation. It was thus unexpected that the ectopic expression of svp would cause the transformation of the cone cells towards R7 neurons. R7 differentiation from the cone cells caused by sev-svp transgene expression does not require the function of boss or sev, whereas it is completely suppressed by mutations in sina. Thus, ectopically expressed svp acts downstream of the sev receptor tyrosine kinase, but acts either upstream or in parallel to the sina gene. These epistatic relationships are similar to other conditions that cause the same cellular transformation, i.e., the expression of activated Ras1 (Fortini et al., 1991) and the reduction in the Gap1 activity (Gaul et al., 1992; Rogge et al., 1992; Buckles et al., 1992). Similarities in phenotypes and genetic relationships of svp and the activated ras pathway suggest that svp might act through ras to provide the potential to become a neuron. In support of this view, we have identified alleles of the Ras1 gene among dominant suppressors of rough eye phenotype caused by the sev-svp2 transgene (S. Kramer, F. Birkmeyer, M. M. and Y. H. unpublished). Although phenotypes caused by the loss of svp function indicate that svp is involved in a decision between two neuronal cell types, there is some evidence suggesting that svp plays a role similar to that of sev in providing neuronal fate per se. In ommatidia that are doubly mutant for svp and sev, not only R7 but also some of the outer photoreceptors fail to adopt a photoreceptor cell fate (Mlodzik et al., 1990b). Since sev is not required in R3/R4/R1/R6 in svp⁺ ommatidia, it appears that the role that svp plays in R3/R4/R1/R6 is not simply to control their photoreceptor subtype, but also to decide between neuronal versus non-neuronal fate. Thus the role of svp appears different from that of the rough gene, which specifies a subtype of outer photoreceptors, but does not have the potential to induce neuronal development of cone cells (Kimmel et al., 1990; Basler et al., 1990). It should be noted, however, that many of the dominant phenotypes can be caused by both type 1 and type 2 isoforms, which differ in the putative ligand-binding domain. It is therefore possible that the phenotypes observed are generated in a ligand-independent way, due to a function that is at least in part different from the one that svp performs in R3/R4/R1/R6.

We have shown that spatially restricted expression of the svp gene is essential for execution of its normal function. This result is similar to those obtained with another Drosophila member of the steroid receptor gene family, the tailless gene (Steingrímsson et al., 1991) which, like svp, is expressed in the region of the embryo requiring its function. Ectopically expressed tailless appears to have the same function as the endogenous gene product, suggesting that either the tailless ligand is uniformly distributed, or the tailless function is ligand-independent. On the contrary, ubiquitous expression of the ultraspiracle gene, which is the Drosophila homolog of the retinoid X receptor and is likely to function as a heterodimer with the ecdysone receptor (Yao et al., 1992), do not interfere with normal development (Oro et al., 1992). These differences may reflect differences in strategies that steroid receptors utilize to regulate their functions, such as the distribution of the receptor itself, distribution of the ligand, or the distribution of the receptor’s heterodimeric partner.

We thank K. Golic, P. Lawrence, U. Gaul, M. Fortini and L. Zipursky for providing fly strains, K. Saigo and L. Zipursky for gifts of antibodies, D. Bowtell and C. Thummel for their expression vectors, D. Brunner and E. Hafen for help with phototactic behavior experiments, S. Kramer for help in sev-svp2 experiments, and A. Ephrussi, A. Bejsoveck and S. Roth for valuable comments on the manuscript. This work was supported by a National Institutes of Health Training Grant (S. R. W.) and grants from National Institutes of Health (R01 NS29662) and Pew Scholars Program (Y. H.), and Swiss National Science Foundation (M. M.). C. S. G. and G. M. R. are investigators of the Howard Hughes Medical Institute.