ABSTRACT
The dachshund gene encodes a nuclear protein that is required for normal eye development in Drosophila. In the absence of dachshund function, flies develop with severely reduced or no eyes. We show that targeted expression of dachshund is sufficient to direct ectopic retinal development in a variety of tissues, including the adult head, thorax and legs. This result is similar to that observed with the highly conserved Drosophila gene eyeless, which can induce ectopic eye formation on all major appendages. Here, we show that dachshund and eyeless induce the expression of each other and that dachshund is required for ectopic retinal development driven by eyeless misexpression. These results suggest that the control of eye development requires the complex interaction of multiple genes, even at the very highest regulatory levels.
INTRODUCTION
Molecular and genetic studies of development in Drosophila have helped to decipher the fundamental mechanisms of cell-fate determination and, due to the highly conserved nature of the proteins involved, have had profound implications for our understanding of vertebrate development. In particular, the Drosophila eye has been an extremely informative setting for the study of both cell-cell inductive events and long-range signaling. While our knowledge of the genes and mechanisms controlling morphogenesis and neural differentiation in the Drosophila eye has increased dramatically in recent years, relatively little is known about how retinal tissue is specified in the first place. The most important advance in this area was the discovery that the eyeless (ey) gene is able to initiate an entire cascade of gene activity sufficient to generate complete and properly formed compound eyes in Drosophila (Halder et al., 1995). Ectopic expression of ey during development directs the formation of retinal tissue on all major appendages, including antennae, legs and wings. ey encodes a member of the Pax-6 family of transcription factors that contain both a paired domain and a homeodomain (Quiring et al., 1994). Loss-of-function mutations in ey produce flies that have either reduced eyes or no eyes at all. Moreover, vertebrate homologs of ey, including mouse Small eye (Sey) and human Aniridia, are highly conserved both in sequence and in function. The paired and homeodomains encoded by Sey and Aniridia are more than 90% identical to Drosophila Ey and all three genes are required for normal eye development in their respective species (Ton et al., 1991; Hill et al., 1991; Jordan et al., 1992; Glaser et al., 1992). Moreover, in spite of the divergence of insects and mammals more than 500 million years ago, targeted expression of the mouse Sey gene also induces the formation of ectopic compound eyes in Drosophila. Taken together, these results led to the hypothesis that ey is the master control gene for eye morphogenesis (Halder et al., 1995). However, the identity of genes controlling the expression of ey and the presumed down-stream targets of ey function are not known.
Genes known to function early in Drosophila eye development, based on both their expression patterns and mutant phenotypes, are candidates for regulators or targets of ey activity. One such candidate is the dachshund (dac) gene, which encodes a novel nuclear protein required for normal eye development in Drosophila (Mardon et al., 1994). dac is expressed early during eye development and is required for the first steps of eye morphogenesis and photoreceptor determination. The adult Drosophila compound eye is composed of approximately 750 unit eyes or ommatidia that are arranged in a precise hexagonal array (Fig. 1A). Each ommatidium comprises 8 photoreceptor cells, as well as lens-secreting cone cells, pigment cells and a mechanosensory bristle. The adult eye is derived from the eye imaginal disc, an epithelial monolayer that is specified in the late embryo and grows throughout larval development (Fig. 1E). Photoreceptor cells first differentiate at the posterior margin of the eye disc and appear progressively in a wave-like fashion over a period of about two days (Wolff and Ready, 1993). Anterior movement of this wave of neural differentiation, termed the morphogenetic furrow, is composed of two genetically separable events: first, the furrow must begin moving (termed initiation) and second, the furrow must propagate across the eye disc (termed progression). Several conserved genes have been identified whose function is required for proper furrow progression, both as positive and negative regulators, including hedgehog (hh), protein kinase A and patched (reviewed in Heberlein and Moses, 1995; Bonini and Choi, 1995). hh encodes a secreted signaling molecule that plays crucial roles in pattern formation and cell-fate determination throughout much of the metazoa (reviewed in Ingham, 1995; Johnson et al., 1994; Perrimon, 1995). Ectopic expression of hh anterior to the morphogenetic furrow in the undifferentiated epithelium of the eye imaginal disc leads to precocious photoreceptor development (Heberlein et al., 1995; Lee et al., 1994). In contrast, ectopic expression of hh in the anterior compartment of other imaginal discs does not induce photoreceptor development; instead, pattern duplications of structures specific to each disc (i.e., antenna, leg or wing) are observed (Basler and Struhl, 1994). Thus, while ectopic ey expression is sufficient to redirect differentiation of at least a subset of cells from all imaginal discs to a retinal fate, hh acts further downstream during morphogenesis to pattern cells whose primary fate has already been determined.
In contrast to furrow progression, a largely different set of genes regulate furrow initiation. wingless (wg) encodes a secreted signaling molecule that is required to prevent premature initiation of the furrow away from the dorsal and ventral margins of the eye disc and thus refines the furrow to a linear (i.e., not curved) wave of development (Ma and Moses, 1995; Treisman and Rubin, 1995). Signaling by decapentaplegic (dpp), a transforming growth factor β family member, is likely to be required for furrow initiation since loss-of-function mutations in mothers against dpp (mad) prevent initiation (Wiersdorff et al., 1996). mad encodes a novel protein required downstream of dpp signaling (Sekelsky et al., 1995). Even though dpp is expressed in the furrow throughout progression, dpp may not be strictly required for this process because loss of mad function does not prevent furrow propagation. Like dpp, dac is expressed at the posterior margin of the eye disc prior to furrow initiation and neural development (Fig. 2A). In the absence of dac function, dpp expression remains at the posterior margin of the eye disc, furrow initiation is prevented and adults develop with severely reduced or no eyes (Fig. 1B,F). Although dac is highly expressed immediately anterior and posterior to the furrow throughout furrow propagation (Fig. 2B), dac is not required for furrow progression or dpp expression. However, dac is required for proper construction and assembly of ommatidia into the hexagonal array characteristic of the compound eye (Mardon et al., 1994). Since the dac and mad mutant phenotypes in the eye are very similar, it has been suggested that dac may function downstream of dpp signaling but, unlike mad, is not required for the maintenance of dpp expression (Wiersdorff et al., 1996). Based on its early expression pattern at the posterior margin of the eye disc and its essential role in initiation of furrow movement, dac plays a critical role during the earliest stages of morphogenesis and photoreceptor specification in the eye.
In this paper, we show that targeted dac expression in the antennal and leg imaginal discs induces ectopic eye development. Ectopic eyes have a nearly normal morphology and contain photoreceptor neurons that project axons in a manner similar to photoreceptors in the eye disc. Moreover, we show that ectopic dac expression induces ey expression, suggesting that dac acts upstream of ey during retinal development. Surprisingly, we also found that ey induces dac expression and that dac is required for ey function. These results demonstrate that dac and ey expression are intimately related and that these genes are likely to function together in the control of retinal cell-fate specification at early stages of eye development.
MATERIALS AND METHODS
Fly genetics
All Drosophila crosses were carried out at 25°C on standard media. UAS-dac constructs were generated using a full-length dac cDNA (Mardon et al., 1994) cloned in the EcoRI site of pUAST (Brand and Perrimon, 1993). Flies were transformed using standard techniques (Rubin and Spradling, 1982). 14 transformants were analyzed in detail. dpp-GAL4 flies express GAL4 in the antennal disc and at the A-P compartment boundaries of the leg and wing discs in a pattern that approximates that of dpp (Staehling-Hampton et al., 1994). For an unknown reason, the dpp-GAL4 line does not express GAL4 in the morphogenetic furrow during furrow progression. Instead, GAL4 is expressed specifically at the posterior margin of the early eye disc and then fades away from the central portion of the disc rapidly following furrow initiation (see Fig. 2). Rescue of dac null mutant animals were assayed as follows: dac[3], dpp-lacZ ; dpp-GAL4 / SM6-TM6B flies were crossed to dac[1] ; UAS-dac / SM6-TM6B flies and non-Tb larvae or non-Hu adults were selected as dac[3], dpp-lacZ / dac[1] ; dpp-GAL4 / UAS-dac. Targeted expression of ey in a dac null mutant background was carried out as follows: dac[3], dpp-lacZ ; dpp-GAL4 / SM6-TM6B flies were crossed to dac[1] ; UAS-ey / SM6-TM6B flies and non-Tb larvae were selected as dac[3], dpp-lacZ / dac[1] ; dpp-GAL4 / UAS-ey.
Scanning electron microscopy and histology
Due to severe truncation of the legs and wings, animals carrying both UAS-dac and dpp-GAL4 constructs fail to emerge from their pupal cases. Thus, all adult images were prepared from animals dissected from late pupae. Samples were prepared for scanning electron microscopy and histological sections as previously described (Kimmel et al., 1990; Tomlinson and Ready, 1987).
Immunohistochemistry
Imaginal discs were dissected and stained as previously described (Mardon et al., 1994). All ectopic dac expression studies in this paper were performed using dpp-GAL4 × UAS-dac animals. dpp expression was shown using a β-galactosidase reporter construct specific for imaginal discs (Blackman et al., 1991). Anti-Elav (Robinow and White, 1991), anti-Dac (Mardon et al., 1994), anti-Glass (Moses and Rubin, 1991), anti-Neuroglian (Hortsch et al., 1990) and anti-Ey (Halder et al., 1995) stainings were all performed using the same protocol with the following exceptions. For anti-Glass staining, imaginal discs were dissected in PBS and fixed in PLP (2% paraformaldehyde, 10 mM sodium periodate, 75 mM lysine, 37 mM sodium phosphate pH 7.2) for 25 minutes on ice. For anti-Ey staining, imaginal discs were dissected in PBS and fixed in PEM (0.1 M Pipes pH 7.0, 0.2 mM EGTA, 4% paraformaldehyde) for 25 minutes on ice. All discs were mounted in 80% glycerol in PBS.
RESULTS
Rescue of the dachshund mutant eye phenotype
We sought to further our understanding of dac gene function during development through ectopic expression studies using the GAL4 system (Brand and Perrimon, 1993; Brand and Dormand, 1995). Our first step toward this goal was to establish that our dac cDNA clone encodes a functional protein using rescue of the dac mutant eye phenotype as an assay. Confirming predictions based on our previous analyses, expression of dac specifically at the posterior margin of the eye disc early in development (Fig. 2C,E) is able to fully rescue morpho-genetic furrow initiation in a dac null mutant background (Fig. 1C,G). In this setting, even though dac is not expressed anterior or posterior to the furrow during most of progression, the furrow still moves normally across the entire eye imaginal disc (Fig. 2D,F). However, the requirement for dac function during furrow progression for normal ommatidial assembly is readily apparent in these preparations: rescued eyes are disorganized, both during larval stages and in the adult (Fig. 1C,G). Although ectopic dac expression at the posterior margin of the eye disc in a wild-type (dac+) background had little or no effect on furrow initiation or progression, the adult eyes of such animals are truncated in the dorsal-ventral dimension (Fig. 1D,H). This may be attributable to the strong ectopic expression of dac at the dorsal and ventral margins of the eye disc (Fig. 3D). These experiments confirmed that our dac cDNA encodes a functional protein capable of rescuing furrow initiation in dac mutant animals and that little or no Dac protein is required for furrow progression.
dachshund induces ectopic eyes
In addition to the eye disc, dac is normally expressed only in specific domains of the antennal, leg and wing discs (Fig. 3A-C) and the embryo (data not shown). These restricted patterns of dac expression are important for normal development: a single 30 minute heat shock (hs) of animals carrying both hs-GAL4 and UAS-dac transgenes at any time during development causes complete lethality (data not shown). More subtle phenotypes were produced using a dpp-GAL4 construct to drive dac expression in the antennal disc and at the anterior-posterior (A-P) compartment boundary of the leg and wing discs (Cohen, 1993). In each case, ectopic expression of Dac protein caused strong disruption of the normal dac pattern of expression (Fig. 3D-F), increased cell death and truncation of the resulting appendage (data not shown). Strikingly, about 20% of these animals develop ectopic eyes just ventral to the antenna on the anterior surface of the head (Fig. 4A,B,D). This position in the adult corresponds to the site of strongest ectopic dac expression at the ventral margin of the antennal disc (arrowhead in Fig. 3D). Ectopic eyes derived from the ventral antennal disc often comprise up to 40 or 50 ommatidia (Fig. 4A). Ectopic retinal development can occur in a dac null mutant background where the UAS-dac transgene is the only source of Dac protein (data not shown).
The external morphology of dac-induced ectopic eyes closely resembles that of normal adult eyes, including lens and interommatidial bristle formation (Fig. 4C,D). Sections of ectopic eyes reveal a typical ommatidial structure that includes the densely staining rhabdomeres (the light-sensing organelles of the retina) and pigment granules that serve to optically insulate each ommatidium (Fig. 4E,F). However, the normally precise hexagonal shape of most ommatidia is absent (Fig. 4C,D) and a minority of ommatidia display the characteristic trapezoidal arrangement or number of rhabdomeres (Fig. 4E,F). Ectopic eyes derived from dac misexpression are normally pigmented (Fig. 5B-D).
Ectopic dac expression also induces retinal development on a portion of the body wall (thorax) that is derived from leg imaginal discs. While 95% of animals display at least some red pigment on the thorax just dorsal to the articulation of the leg and body, only 5% present obvious ommatidial structures (Fig. 6). This position again corresponds to the site of strongest ectopic dac expression in the leg imaginal disc (arrow-heads in Fig. 3B,E). In about one percent of cases, pigment is also clearly visible on the leg per se (data not shown). Patches of red pigment are only rarely associated with wing structures and no well-formed ommatidia have been observed. Truncation of the adult appendages demonstrates that ectopic dac expression along the A-P compartment boundary significantly alters the fate of cells in each of the imaginal discs, indicating that dac plays an instructive role during development. Moreover, because ectopic dac expression in the antennal and leg imaginal discs is sufficient to direct some cells to adopt a retinal fate, dac must function at or near the very highest levels of the regulatory hierarchy controlling eye development.
Developmental analysis of ectopic photoreceptors
Ectopic neural development resulting from targeted dac expression is readily apparent in antennal and leg discs during late larval stages. The nuclear protein Elav is detected in all neurons in Drosophila (Robinow and White, 1991) and is not normally expressed in the antennal disc (Fig. 5E). Ectopic dac expression induced formation of Elav-positive cells in the ventral antennal disc in about 20% of samples analyzed, in good agreement with the frequency of ectopic eyes observed in adults (Fig. 5F,G). These cells are likely to be ectopic photoreceptor neurons. Indeed, we found evidence for retina-specific development in the antennal disc by the ectopic appearance of Glass-expressing cells (Fig. 5H,L). glass encodes a zinc-finger protein that is specific for visual system development in Drosophila and is not normally expressed in antennal or leg imaginal discs (Moses and Rubin, 1991; Moses et al., 1989). Ectopic glass expression is also observed in the dorsal leg disc in response to dac misexpression (Fig. 6C,F). Remarkably, ectopic photoreceptors in the antennal disc send out axonal projections that first extend medially and then turn sharply posteriorly (Fig. 5J,K). This is very similar to the pattern of normal axonal projection in the wild-type eye disc where photoreceptor axons exit posteriorly through the optic stalk to synapse in the optic lobe of the larval brain (Fig. 5I). Whether the axonal projections of ectopic photoreceptor cells in the antennal disc are able to find their way to targets in the brain (or elsewhere) is not known.
We looked for evidence of morphogenetic furrow movement by examining the pattern of dpp expression in antennal discs positive for photoreceptor cell development. dpp expression marks the position of the furrow as it progresses across the eye imaginal disc. In the antennal disc, dpp is normally expressed in a sector abutting the ventral margin of the antennal disc (Fig. 5E). In antennal discs that show ectopic neural development, dpp expression appears to have moved away from the ventral margin toward the center of the disc as an elongated stripe (Fig. 5F,G). These results suggest that ectopic dac expression is sufficient to initiate movement of an ectopic furrow at the ventral margin of the antennal disc and induce photoreceptor development.
Relationship between dachshund and eyeless
Ectopic eye development caused by targeted expression of the dac gene is remarkably similar to results obtained with ey (Halder et al., 1995). We therefore sought to examine the relationship between dac and ey using both molecular and genetic approaches. Three independent results suggest that dac functions downstream of ey. First, misexpression of ey in the antennal, leg and wing imaginal discs is sufficient to induce ectopic dac expression and disrupt the normal pattern of dac expression in all discs (Fig. 3G-I). These results suggest that ey activity positively regulates dac transcription, either directly or indirectly, and raised the possibility that dac function may be required downstream of ey. Indeed, we found that targeted ey expression was unable to induce ectopic eye formation in a dac null mutant background (Fig. 7). Finally, if dac functions downstream of ey in a simple linear pathway, then dac function should not be required for ey expression. ey is normally expressed anterior to the morphogenetic furrow in the eye imaginal disc (Fig. 8A) and is not expressed in other imaginal discs (Quiring et al., 1994). We found that ey is still expressed in dac null mutant eye discs, demonstrating that dac is not essential for ey expression (Fig. 8B). However, ey expression is restricted to the posterior margin of the disc, presumably due to the failure of furrow initiation in dac mutants. These results suggest that dac functions downstream of ey and, considering that dac can induce ectopic retinal development, are consistent with the idea that dac may be a direct target of ey function.
No null mutant allele of ey is available (Quiring et al., 1994) and all existing ey alleles cause highly variable phenotypes in the eye, ranging from wild type to total absence in the same animal (data not shown). Thus, we are unable to test whether ey is required for dac expression or function. Surprisingly, however, we found that ectopic dac expression in the antennal disc is sufficient to induce ey expression (Fig. 8C). In addition, the normal pattern of ey expression in the eye disc is disrupted as a result of ectopic dac expression. Extremely weak or no ey expression was observed in the leg or wing discs in response to directed dac expression (data not shown). Thus, dac is unable to drive ey expression at detectable levels in all cells in which it is present. Nevertheless, dac-mediated induction of ey expression in the antennal disc, in addition to the results presented above, suggests that dac may function both upstream and downstream of ey during retinal development.
DISCUSSION
Proper construction of tissues and organs during development requires the precise regulation of complex genetic hierarchies. Only a few cases have been described where expression of a single gene is sufficient to direct the development of complete and properly formed organs or systems. For example, expression of the Sry gene is able to cause genotypically XX mice to development as phenotypic males (Koopman et al., 1991). In two other cases, the LEAFY (LFY) and APETALA1 (AP1) genes each direct precocious flower development in Arabidopsis (Mandel and Yanofsky, 1995; Weigel and Nilsson, 1995). Most recently, targeted expression of the Drosophila gene vestigial (vg) was shown to generate wing-like out-growths from the eyes, legs and antennae of adult flies (Kim et al., 1996). Finally, the highly conserved Drosophila gene eyeless (ey) is sufficient to direct the formation of compound eyes on all the major appendages of flies when ectopically expressed during development (Halder et al., 1995). Most of these genes encode nuclear proteins (Sry, AP1, Vg and Ey) that may control elaborate gene networks governing organ formation. In particular, ey encodes a member of the Pax-6 family of DNA-binding transcription factors that includes mouse Small eye (Sey) and human Aniridia (Quiring et al., 1994). Targeted expression of ey or Sey causes ectopic eye development in Drosophila, suggesting that genetic mechanisms of retinal development may be more highly conserved than previously anticipated (Halder et al., 1995). Since ey, Sey and Aniridia are each required for normal eye development and ey and Sey are functionally conserved, these genes are likely to represent descendants of a common ancestral gene that existed prior to the divergence of insects and mammals more than 500 million years ago. Moreover, these results have forced a reevaluation of the traditional view that insect and vertebrate visual systems evolved independently (Quiring et al., 1994).
We have identified the Drosophila gene dachshund (dac) as another member of this small group of genes known to occupy positions high in the regulatory hierarchies directing organ development. dac is expressed in nearly all cells of the eye imaginal disc throughout larval development and in the lamina of the optic lobe where most photoreceptor axons first synapse in the brain. dac encodes a novel nuclear protein that is necessary for the initiation of eye morphogenesis and proper assembly of the retinal field in Drosophila (Mardon et al., 1994). Here, we have shown that dac is also sufficient to induce properly formed ectopic eye structures in a variety of tissues in Drosophila. Taken together, these results suggest that dac functions at many levels of the genetic hierarchy controlling eye development and that dac can switch on a pathway that is likely to involve thousands of genes (Halder et al., 1995; Thaker and Kankel, 1992).
Targeted expression of dac is sufficient to induce ectopic ey expression in one or more imaginal tissues during larval development. Similarly, ey can induce dac expression in most imaginal cells. Whether a similar regulatory relationship normally exists between dac and ey in the eye disc is not known. In wild-type eye discs, dac and ey expression anterior to the morphogenetic furrow overlap significantly (Figs 2B, 8A). In addition, ectopic expression of either gene at the posterior and lateral margins of the eye disc disrupts the normal pattern of expression of the other (Figs 3G, 8C). Thus, it seems likely that dac and ey contribute to the regulation of each other during normal eye development.
We have shown that dac is required for induction of ectopic eye development by ey misexpression. In contrast, we do not know whether ey is required for dac function because no strong or highly penetrant mutant alleles of ey exist. Although dac does not induce strong ey expression in the leg imaginal disc, dac can efficiently induce pigment development and, in some cases, ommatidial formation in structures derived from the leg disc. Thus, if dac induction of ectopic retinal development is mediated by and requires ey function, then only low levels of ey are required for this process. Alternatively, dac may act either downstream of ey or in a parallel pathway and not require ey function in this regard.
Since dac is able to induce ectopic ey expression in the antennal disc, dac can function as a positive regulator of ey. Three lines of evidence suggest that dac also functions down-stream of ey. First, ey induces dac. Second, dac is not required for ey expression and third, dac is required for ectopic eye induction by ey. Thus, it is possible that dac and ey participate in a positive regulatory feedback loop during eye development. This is reminiscent of the regulatory relationships proposed for wingless, hedgehog and engrailed in the establishment of parasegmental boundaries in the Drosophila embryo (Hooper, 1994; Manoukian et al., 1995) and Sonic hedgehog and Fgf-4 in specifying growth and patterning of the vertebrate limb (Laufer et al., 1994; Niswander et al., 1994). Whether Dac and Ey each act directly or indirectly to control transcription of the other remains to be determined. Since dac is expressed in a wide variety of tissues and locations where ey is not, including the antennal, leg and wing discs, it is clear that neither gene is solely necessary or sufficient for expression of the other. Moreover, the nature of the regulatory relationship between dac and ey must depend upon the local cellular environment.
Targeted expression of dac using dpp-GAL4 induces ectopic retinal development in about 20% of antennal discs and, to a lesser extent, in leg and wing discs. In contrast, targeted expression of ey in the same manner is able to direct retinal development in all imaginal discs with complete penetrance. This suggests that ey may induce the expression of one or a few other genes that facilitate eye development and that are not efficiently induced by dac. However, neither gene is able to induce photoreceptor development in all cells in which it is expressed. For example, ectopic dac expression is unable to efficiently induce retinal development along any part of the A-P compartment boundary of the wing imaginal disc. Similarly, targeted ey expression fails to induce photoreceptor development along the ventral portion of the A-P compartment boundary of the wing disc (data not shown). These results suggest that dac and ey do not act alone in the control of gene expression or retinal cell-fate specification. Instead, these genes are likely to require the coop-eration of, or be inhibited by, other factors that are expressed in a spatially or temporally restricted pattern during development. Genes acting early in retinal development are potential candidates for such factors, and include dpp, sine oculis (so) and eyes absent (eya). Both so and eya encode nuclear proteins that are expressed early in the eye imaginal disc and are required for normal eye development (Bonini et al., 1993; Cheyette et al., 1994). Further experiments examining the relationships among these and other genes will be required to decipher the molecular and genetic mechanisms controlling retinal cell-fate specification.
Although a vertebrate homolog of dac has not been identified, highly conserved dac homologs have been isolated from several invertebrate species, including Drosophila virilis, the flour beetle Tribolium and the butterfly Precis coenia. The amino acid sequences predicted from these dac homologs are 60 to 85% identical to the Drosophila melanogaster Dac protein (W. Shen, G. Mardon, unpublished data and T. Heanue and C. Tabin, personal communication). Since these species are 60 to 250 million years diverged from Drosophila melanogaster and most of the known genes required for eye development in Drosophila are highly conserved in vertebrates, it is likely that one or more vertebrate homologs of dac exists (Beverley and Wilson, 1984).
ACKNOWLEDGEMENTS
We thank Hugo Bellen, Andrea Brand, Kwang Choi, Walter Gehring, Corey Goodman, Georg Halder, Tiffany Heanue, Ulrike Heberlein, Kevin Moses, Norbert Perrimon, Gerry Rubin, Cliff Tabin, and the Bloomington Drosophila Stock Center for generously providing fly stocks, plasmid DNAs, antibodies, protocols and unpublished data. We also acknowledge Hugo Bellen and Yuchun He for help with fly transformations, Joiner Cartwright, Susan Robbins, Mary Hsiao, John Hicks and Jim Barrish for assistance with scanning electron microscopy, Thurl Harris for help in plasmid construction and Zhihuan Zhang for fly work and disc staining and Hugo Bellen, Kwang Choi and Jim Musser for comments on the manuscript. This work was supported by grants from the National Eye Institute (R01 EY11232-01), the Baylor Mental Retardation Research Center (P30 HD24064) and the Retina Research Foundation.