dachshund encodes a nuclear protein required for normal eye and leg development in Drosophila

The adult epidermis in higher dipteran insects such as Drosophila is derived from specialized structures in larvae known as imaginal discs. Set aside during mid-embryogenesis as small groups of 10 to 70 cells each, imaginal disc primordia grow and differentiate during larval development as epithelial monolayers. Metamorphosis of imaginal discs during early pupal life generates the adult cuticle and appendages, including the eyes, legs and wings (reviewed in Cohen, 1993). While a morphological description of these events has been well-doc-umented, the genetic and molecular mechanisms controlling pattern formation and morphogenesis of imaginal discs are just now being unraveled.

Eye imaginal disc development creates a highly ordered array of differentiated cells from a pool of equivalent, unde-termined cells (Ready et al., 1976; Lawrence and Green, 1979). This process generates the adult eye, which consists of about 800 unit eyes or ommatidia. Each ommatidium comprises eight photoreceptors, termed R1-R8, as well as a dozen accessory cells, including lens-secreting cone cells, pigment cells and a mechanosensory bristle. During late larval and early pupal development, photoreceptor cell determination and differen-tiation sweep across the eye imaginal disc following an inden-tation in the epithelial monolayer termed the morphogenetic furrow (Ready et al., 1976). The morphogenetic furrow is char-acterized by an apical-basal shortening of cells, synchroniza-tion of the cell cycle, coordinated gene expression and initiation of neural differentiation (Wolff and Ready, 1991; Heberlein et al., 1993; Ma et al., 1993). The furrow begins at the posterior margin of the eye disc and proceeds anteriorly as a linear front, traversing the disc during a period of about 2 days. Immediately posterior to the furrow, the first presump-tive photoreceptor cells become apparent by both morpholog-ical and molecular criteria (Tomlinson and Ready, 1987a; Wolff and Ready, 1991). As the furrow progresses across the eye disc, photoreceptor cells are recruited in a stepwise fashion such that all stages of ommatidial assembly can be viewed in a single disc preparation (reviewed in Wolff and Ready, 1993). The molecular mechanisms controlling cell-fate determination and movement of the morphogenetic furrow in the eye imaginal disc are the subjects of this report.

Two genes that are involved in progression of the morpho-genetic furrow, decapentaplegic (dpp) and hedgehog (hh), are also key determinants of pattern formation in other imaginal discs as well as in the Drosophila embryo. Both dpp and hh encode phylogenetically conserved, secreted proteins that act as concentration-dependent morphogens responsible for organizing tissue polarity throughout development. dpp specifies dorsal-ventral pattern formation in the embryo (Ferguson and Anderson, 1992) and encodes a Drosophila homolog of the mammalian transforming growth factor β (TGFβ) family of cytokines that regulate cell proliferation and differentiation (Padgett et al., 1987). The segment polarity gene hh is required for the specification of compartment boundaries in the Drosophila embryo and imaginal discs (Ingham et al., 1991; Tabata and Kornberg, 1994; Basler and Struhl, 1994). Ver-tebrate homologs of hh exhibit organizing activity during mouse central nervous system and chick limb development (Echelard et al., 1993; Riddle et al., 1993). During eye imaginal disc development, dpp is expressed specifically in the mor-phogenetic furrow and this expression is correlated with pro-gression of the furrow. Mutations that stop furrow progression also abolish dpp expression. In particular, hh function is required for both dpp expression and furrow progression (Heberlein et al., 1993; Ma et al., 1993). hh is expressed posterior to the furrow in determined photoreceptor cells and it has been proposed that hh protein diffuses anteriorly to pos-itively regulate dpp expression in the furrow. Consistent with this interpretation, ectopic expression of hh in the anterior compartment of the wing imaginal disc is sufficient to induce dpp expression in cells bordering the area of ectopic hh expression (Basler and Struhl, 1994). In the eye disc, dpp and/or hh may diffuse several cell diameters further into the unpatterned epithelium anterior to the furrow to regulate other genes required for undetermined cells to become competent to adopt a neural fate. Indeed, hh function may be required for expression of at least three other genes involved in eye devel-opment: glass, scabrous and hairy (Ma et al., 1993). glass and scabrous are normally expressed in and posterior to the furrow while hairy is expressed anterior to the furrow (Moses and Rubin, 1991; Baker et al., 1990; Mlodzik et al., 1990; Brown et al., 1991). Since hh is expressed in differentiated photore-ceptors about three columns posterior to the furrow, these data suggest that hh is required for expression of genes that are expressed earlier, in developmental terms, than hh itself. This is possible because all stages of photoreceptor development are in progress simultaneously in a single eye disc and hh is a secreted protein capable of traveling backwards in develop-mental time (i.e., anteriorly) by means of diffusion (Ma et al., 1993). In this manner, hh is thought to control progression of the morphogenetic furrow.

In contrast to furrow progression, little is known about the mechanisms controlling initiation of the morphogenetic furrow. The furrow begins moving (i.e., initiates) during the late second or early third instar larval stage. However, the signal that causes the furrow to begin moving is not known. It seems likely that a distinct initiation mechanism precedes propagation: while hh is essential for furrow progression, it cannot be involved in furrow initiation because it is not expressed in the eye disc prior to neural differentiation. In addition, not all cells of the eye imaginal disc differentiate as retina: the periphery of the eye disc monolayer gives rise to a portion of the adult head cuticle surrounding the retinal field (Cohen, 1993). Again, the mechanisms controlling the distinc-tion between retinal and cuticle fates in the eye disc are not known. In this paper, we describe mutations in dachshund (dac), which prevent cells at the posterior margin of the eye disc from following a retinal differentiation pathway. dac encodes a novel nuclear protein which is expressed at the posterior margin of the eye imaginal disc prior to furrow initiation and anterior to the furrow during furrow progression. In the absence of dac function, cells at the posterior margin of the eye imaginal disc appear to follow a cuticle differentiation pathway and are unable to respond to the pattern propagation signals required for furrow movement. Furrow initiation fails and little or no photoreceptor cell differentiation is observed. Curiously, the requirement for dac function in retinal cell-fate determination is not observed for cells located in more anterior portions of the eye disc. These results suggest that posterior margin cells possess an identity distinct from that of other cells in the eye imaginal disc from an early stage of development.

The hypermorphic Egfr allele used was Elp^B1(Baker and Rubin, 1989). The recessive viable Egfr allele top¹ was a gift of Kathryn Anderson. The P-element insertion in the dachshund genomic locus was originally called rK364 and has been renamed dac^P. The sup-pressor of Elp phenotype did not separate meiotically from this P-element in over 1300 progeny scored. In situ hybridization of a probe representing P-element sequences to salivary gland polytene chromo-somes placed the dac^P element at cytological position 36A on the second chromosome (not shown). An additional nine alleles of dac (dac¹-dac⁹) were obtained from Iain Dawson, Yale University. The dac locus was originally designated as the l(2)36Ae complementation group (Ashburner et al., 1990) but was later renamed dac. Analysis of polytene chromosomes from dac⁴ animals revealed a small deletion that removes the major band at 36A (not shown). This deletion was confirmed by DNA hybridization analysis with genomic probes flanking the site of the P insertion (see below). The dac⁴ mutation is, therefore, a complete loss-of-function or null mutation. dac¹, dac³ and dac⁴ are all null mutations for both the eye and leg phenotypes. The other dac mutant alleles are similar to the dac^P allele with regard to both the eye and leg phenotypes and are assumed to be partial loss-of-function mutations in dac.

Adult flies were prepared for electron microscopy as described by Kimmel et al. (1990). Adult eyes were fixed, embedded and sectioned as described by Tomlinson and Ready (1987b). Acridine orange staining of eye and leg imaginal discs was performed as described by Bonini et al. (1993).

Genomic DNA sequences flanking the P-element insertion in dac were cloned by plasmid rescue (Mlodzik et al., 1990). Beginning with a genomic rescue fragment as a probe, eight independent, overlapping lambda phage clones containing genomic DNA surrounding the P-element were then isolated from a Sau3A partial digestion genomic library cloned in λFix (provided by Kevin Moses). These clones span approximately 30 kb. Ten independent, overlapping cDNA clones derived from this genomic region were isolated from a λgt10 cDNA library prepared from eye imaginal disc RNA (constructed by Alan Cowman). All DNA sequences were obtained using a Pharmacia ALF automated sequencer and analyzed with the Staden software. Two full-length cDNAs (5 kb) were sequenced on both strands using synthetic oligonucleotide primers. The corresponding sequences in genomic DNA were also determined and the putative splice junctions deduced (Mount et al., 1992). Approximately 300 bp from both ends of the ten independent cDNAs were also sequenced, as well as the splice junctions flanking exons two, three and four (see Fig. 2). The site of the P-element insertion was determined by sequencing the ends of the genomic rescue fragment.

The P insert in dac is an ‘enhancer trap’ element that contains the bacterial lacZ gene (Mlodzik and Hiromi, 1991). β-galactosidase activity staining of discs prepared from third instar dac^P heterozygote larvae reveals high levels of expression in both eye and leg imaginal discs (not shown). Monoclonal antibodies were raised against two non-overlapping, bacterially expressed dac proteins corresponding to amino acids 149-368 and 378-599 (see Fig. 2). The monoclonal prepa-rations are designated mAbdac1-1 and mAbdac2-3, respectively. Both antibody preparations detect the same expression patterns as that seen with β-galactosidase activity staining in the dac^P line. Although dac protein staining is still observed in imaginal discs prepared from larvae homozygous for the weak dac^P allele, this staining is at sig-nificantly lower levels as compared to wild-type (not shown). This is presumed to be full-length, wild-type dac protein, since the P-element is inserted well 5′ of the open reading frame (Fig. 2A). This result is consistent with the partial loss-of-function phenotype of the dac^P mutant. As expected, no dac protein is detectable in discs prepared from dac⁴ deletion mutant larvae or in dac null mutant clones (Fig. 5). The two other phenotypic dac null alleles, dac¹ and dac³, are also protein nulls. In contrast to the dac⁴ deletion mutant, DNA blot analysis indicates that the genomic DNA encompassing the dac¹ and dac³ null alleles is grossly intact (not shown). These results suggest that loss-of-function mutations in the transcription unit defined by the P-element insertion (dac^P) are responsible for the mutant phenotypes described. In addition, the null phenotype for this locus is unambigu-ously defined by these mutants.

Imaginal discs from second or third instar larvae were dissected in PBS (0.1 M phosphate pH 7.2, 150 mM NaCl), fixed for 15 minutes in 4% formaldehyde in PBS, and washed three times 10 minutes in PBS. Primary antibody incubations were performed in PAXDG (PBS containing 1% BSA, 0.3% Triton X-100, 0.3% sodium deoxycholate, 5% normal goat serum) for 2 hours at room temperature or overnight at 4°C. mAbdac2-3 antibodies were used at a 1:5 dilution, rat mono-clonal anti-ELAV antibodies were used at a 1:1 dilution and rabbit anti-β-galactosidase polyclonal antibodies (Organon Teknica) were used at a 1:5000 dilution. Following the primary antibody incubation, discs were washed three times 10 minutes in PAXDG. Discs were then incubated in PAXDG for 2 hours at room temperature with a 1:200 dilution of one of the following horseradish peroxidase-conju-gated secondary antibodies: goat anti-rat (for the ELAV primary), goat anti-mouse (for the mAbdac primary), or donkey anti-rabbit (for the β-galactosidase primary). After washing three times 10 minutes in PBS, discs were incubated 10 minutes in PBS with 0.5 mg/ml diaminobenzidene (DAB), 0.003% hydrogen peroxide and 1.5 mM NiCl2 for intensification. The double-label experiment for ELAV and β-galactosidase shown in Fig. 4 was performed as follows: eye discs were incubated with both primary antibodies together as described above. Following the PAXDG washes, discs were incubated first with goat anti-rat secondary and then developed using DAB with NiCl2. Following three 10 minute washes in PAXDG, discs were then incubated with donkey anti-rabbit secondary and then developed using DAB without NiCl2 intensification.

Mitotic clones in eye imaginal discs were generated using the FLP/FRT system essentially as described by Xu and Rubin (1993) with some modifications. dac null alleles, dac³ and dac⁴, were recom-bined with an FRT sequence at cytological position 40A. In addition, these stocks carried a single copy of either the dpp-lacZ reporter construct (Blackman et al., 1991) or a P-element containing the lacZ gene inserted in hh (line rJ413 or P30; Ma et al., 1993). Mitotic recom-bination was induced during the first instar larval stage with a 1 hour heat shock at 38°C to induce expression of the FLP recombinase carried on the X chromosome. Eye imaginal discs were dissected from wandering third instar larval animals and fixed as described above. β-galactosidase activity in discs was detected using the X-Gal substrate as described by Simon et al. (1985). The position of homozygous dac null mutant clones was determined by staining discs with anti-dac monoclonal antibodies or with anti-ELAV as described above.

Specification of appropriate spacing between ommatidial founder cells in the morphogenetic furrow is crucial to the first step of neural determination and pattern formation in the eye disc. Although this process is not understood, mutations in the Drosophila homolog of the epidermal growth factor receptor, Egfr, prevent normal spacing and differentiation of photore-ceptor cells in the developing eye (Clifford and Schüpbach, 1989; Xu and Rubin, 1993). We reasoned that a genetic screen for modifiers of Egfr function might identify other genes involved in this process. Such a screen is facilitated by the dominant Ellipse (Elp) allele of Egfr (Baker and Rubin, 1989; 1992). In contrast to the highly ordered arrangement of ommatidia in wild-type animals, Elp heterozygotes have reduced, rough eyes (Fig. 1A,B). Elp behaves as a hypermor-phic allele of Egfr because the Elp phenotype is suppressed by a deletion for this locus. Screens for mutations that dominantly enhance or suppress the Elp rough eye phenotype are likely to identify gene products that affect the activity of Egfr.

A collection of 1170 P-element lines were generated in our laboratory and selected for expression in eye imaginal discs (U. Gaul, L. Higgins and G. M. R., unpublished data). A screen of these lines identified one strong suppressor of the Elp eye phenotype. A single copy of this P-element insert completely suppresses both the external rough eye phenotype of Elp het-erozygotes (Fig. 1C) as well as the disorganized ommatidial structure observed in sections of such eyes (not shown). Excision of the P-element reverts both the suppression of Elp and the semi-lethality of the chromosome. Additional alleles that fail to complement the P-element mutant were identified in other genetic screens for modifiers of the Elp eye phenotype (Iain Dawson, personal communication). Due to the recessive leg phenotype associated with these mutants (described below), this locus has been named dachshund (dac). One mutant allele, dac⁴, is a small deletion encompassing the entire dac locus and is therefore a null mutation (data not shown). These results suggest that loss-of-function mutations in dac reduce the activity of the hyperactive Elp allele of Egfr. Two other dac mutants, dac¹ and dac³, are also phenotypic and molecular null mutations (see Materials and Methods). All of the experiments described in this paper were performed with the dac⁴ deletion mutant and at least one of the other null alleles. In all cases, identical results were obtained with each of these null mutants.

Although dac mutants have greatly reduced viability, adult homozygotes of both the P-element, dac^P, and the null alleles are routinely observed. The eyes of dac^P homozygotes are reduced and roughened (Fig. 1E), while the eyes of dac null mutant homozygotes are either severely reduced or, in about 50% of adults examined, absent (Fig. 1D, H). Although many of the ommatidia in dac^P homozygotes are normally con-structed, 50% have either too few or too many photoreceptor cells. The overall array of ommatidia is disrupted, contributing to the rough external eye phenotype of these flies. None of the few ommatidia in dac null mutant eyes has a normal morphology (not shown). In contrast to the compound eye, the external morphology of the adult ocelli appears normal in all dac mutants.

A second phenotype of dac flies inspired its name: dachshund mutants have short, little legs (Fig. 1L-P). The wild-type adult leg is composed of ten discrete segments: the coxa, trochanter, femur, tibia, five tarsal segments and the claw (Fig. 1L). While dac^P homozygote legs appear to have a normal proximal and distal morphology, the intermediate segments are fused and condensed (Fig. 1M). The same intermediate segments, the femur, tibia and proximal three tarsi, are severely condensed in null mutant legs. In contrast, the coxa, trochanter, fourth and fifth tarsal segments, and the claw appear to develop normally, even in the null mutant (Fig. 1N-P). Upon eclosion from their pupal cases, these mutants are unable to locomote normally and quickly fall into the food and die. However, if kept away from wet medium, dac homozygotes can live for several days before dying, presumably from dehydration. These helpless homozygotes are able to flail their misshapen legs, albeit to no avail, indicating that at least a portion of the leg neuromusculature develops normally and is functional.

If loss-of-function mutations in dac reduce the activity of the activated Elp allele of Egfr, then the same mutations in dac should enhance the phenotype of partial loss-of-function mutations in Egfr. We looked for such an interaction with the weak top¹ allele of Egfr (Clifford and Schüpbach, 1989), which displays a very mild roughening of the adult eye (Fig. 1G). Although we did not observe a dominant effect by loss-of-function mutations in dac on the top¹ eye phenotype, we did species seen by blot analysis of total imaginal disc RNA (not shown). DNA sequence comparison of genomic and cDNA clones suggests that the locus spans approximately 20 kb and see what appears to be a recessive synergy between the weak dac^P allele and top¹. That is, dac^Ptop¹ double homozygotes have a reduced, rough eye phenotype that is more severe than either mutant alone or what we would expect from simple additive effects (Fig. 1F). This result suggests that dac and Egfr may function in the same pathway or in parallel pathways during eye development.

DNA flanking the P-element insert in dac^P was isolated and used to screen cDNA libraries prepared from eye imaginal disc RNA. The longest cDNAs obtained from our library were 5 kilobases (kb) in length and correspond with the major RNA comprises 12 exons, some of which can be alternately spliced (Fig. 2). The dac P-element is inserted after a position corre-sponding to base pair 11 of our longest cDNA clone (Fig. 2A). DNA sequence analysis reveals a single long open reading frame encoding a predicted protein of 1081 amino acids, which shares no significant similarity to any sequence in the databases (Fig. 2B).

Expression of dac is readily apparent prior to imaginal disc morphogenesis. In the eye disc, dac is expressed at the posterior margin prior to initiation of the morphogenetic furrow (Fig. 3A). This expression pattern is similar to that of dpp at the same early stage of development (Fig. 3B). Strong dac expression is also detected in the unpatterned epithelium preceding the morphogenetic furrow as it moves anteriorly across the eye disc (Fig. 3C,D). Posterior to the furrow, dac expression is detected primarily in photoreceptors R1, R6 and R7 as well as the cone cells (not shown). dac is expressed only in that part of the eye disc fated to become retina; the periphery of the disc destined to become head cuticle does not express dac (Fig. 3D). The ring pattern of dac expression in the leg disc is established at an early stage of leg disc development (Fig. 3E), well before the characteristic epithelial folds of a mature leg disc are seen (Fig. 3F,G). In all cases, dac protein appears to be localized to the nucleus.

In addition to strong expression in eye and leg discs, dac protein is found in the third antennal disc segment and the wing imaginal disc (Fig. 3C,H). dac expression is also detected in the central nervous system of embryos (Fig. 3I,J) and in the optic lobe of the larval brain (see cover picture). In spite of these distinct patterns of expression, no obvious morphologi-cal phenotype is observable in the embryonic CNS or the adult antennae and wings in dac null mutant animals. The function of dac in these structures remains to be determined.

The expression pattern of dac during larval leg disc develop-ment is consistent with the mutant phenotype of the adult leg. The leg imaginal disc is composed of concentric folds of epithelia such that the outermost portions of the disc give rise to the most proximal segments (i.e. the coxa) in the adult leg while the most distal structures (i.e. the tarsal segments and claw) are derived from the central portion of the disc. As shown by antibody staining, dac is strongly expressed specif-ically in the presumptive epithelium that is fated to give rise to the femur, tibia and proximal tarsal segments (Fig. 3E-G), the same structures most severely affected in dac mutant legs (Fig. 1M,N). There is no dac expression detected in either the center or the periphery of the leg imaginal disc, and the struc-tures derived from these portions of the disc appear to develop normally in dac mutant flies (Fig. 1O,P).

Although the adult legs of dac mutants are greatly shortened, leg imaginal discs prepared from dac mutant third instar larvae appear morphologically normal, both in size and in their con-centric segmental structure (not shown). However, a significant increase in cell death in dac mutant leg discs is revealed by staining with acridine orange (Spreij, 1971), a dye that is actively excluded from living cells (Fig. 4G,H). This cell death is limited primarily to the region of the disc where dac is expressed and most likely accounts for the failure of disc eversion during early pupal stages. By 6 hours into pupal development, wild-type leg discs have undergone rapid elongation, primarily as a result of cell-shape change (Condic et al., 1991). Such elongation does not occur in dac mutant pupal legs, consistent with the adult leg phenotype (not shown). Thus, there is a failure of morphogenesis in dac mutant leg discs during late larval and early pupal development.

The near or total absence of ommatidia in complete loss-of-function dac mutant adult eyes (Fig. 1D,H) could be the result of either an initial failure of photoreceptor determination during larval development or from a degenerative event following neural determination and differentiation, such as in the glass mutation (Moses et al., 1989). We examined the state of neural development in dac mutant eye discs using a mono-clonal antibody that recognizes the nuclear ELAV antigen (Robinow and White, 1991). ELAV is expressed in all neurons in Drosophila and is apparent immediately posterior to the morphogenetic furrow in the eye imaginal disc. In contrast to the highly ordered array of photoreceptors seen in wild-type eye discs, fewer neurons differentiate in dac^P imaginal discs (Fig. 4A,B). Close examination of these discs reveals that highly variable numbers of photoreceptors per ommatidial cluster are already present at this stage (not shown) and can account for the adult phenotype of dac^P homozygote eyes. Moreover, null mutant eye discs have just a few ELAV-positive clusters or none at all (Fig. 4C). When present, these presumptive photoreceptors are always located near the posterior margin of the disc, where neural development normally begins (Fig. 4D). These results demonstrate that normal photoreceptor development is blocked in dac null mutants.

We examined furrow movement in wild-type and dac mutant imaginal discs using a lacZ reporter construct for dpp (Blackman et al., 1991). dpp is expressed specifically in the morphogenetic furrow and can serve as a marker for the position of the furrow in the eye disc (Heberlein et al., 1993). While the morphogenetic furrow has normally progressed about half-way across the eye disc by the late third instar larval stage, dpp-lacZ expression remains at the posterior margin of dac null mutant discs (Fig. 4A,C). In about half of the dac null mutant eye discs examined, there is a small amount of furrow movement observed at the posterior midline of the disc, near the optic stalk. This movement is usually associated with the appearance of a few ELAV-positive cells (Fig. 4C,D). In all other cases studied, there is neither furrow movement nor neural differentiation (not shown). Thus, dac does not appear to be required specifically for neural differentiation. Instead, dac function is required for normal movement of the morpho-genetic furrow.

In the weak dac^P mutant, furrow movement is uneven. Normally, the furrow moves across the eye disc as a linear front with neural development beginning immediately in its wake (Fig. 4A). In contrast, the partial loss-of-function dac^P mutation prevents the furrow from moving normally in lateral regions of the eye disc. This results in a curved furrow that has advanced further at the midline than at the periphery of the disc (Fig. 4B). Nevertheless, neural development closely follows the furrow in all cases examined.

Eye discs of third instar dac null mutant larvae are approx-imately normal in size, suggesting that the failure of furrow movement is not the result of inadequate cell proliferation (Fig. 4C). Nevertheless, in the absence of furrow movement, cells in dac mutant eye discs fail to adopt a neural fate and are likely to remain in an undifferentiated state. The fate of these undif-ferentiated cells is death. As detected by acridine orange staining, a large increase in cell death is seen in dac mutant eye discs (Fig. 4E,F).

To analyze further the role that dac plays in furrow movement, we made homozygous dac null mutant clones of cells in a heterozygous dac background and examined the phenotype of such clones in both larvae and adults. Dramati-cally different results were obtained depending on the position of these clones in the developing eye. dac clones occurring anywhere in the eye but not including the posterior margin (‘interior’ clones) always give rise to ommatidia (Fig. 5A). This suggests that the morphogenetic furrow is able to progress through patches of dac mutant cells given the chance to first initiate movement in wild-type tissue. In contrast to interior clones, posterior margin dac mutant clones usually fail to develop into retinal tissue; head cuticle is found in its place (Fig. 5B,C). In about one third of cases examined a few mutant (white-minus) ommatidia are found at the borders of dac clones (Fig. 5L) and about 5% of posterior margin clones will give rise to a substantial patch of mutant ommatidia (Table 1). The reason for this variability is not known.

dac mutant clones in larval discs are consistent with the observed adult phenotypes. While neural differentiation is detected in interior dac clones (Fig. 5D), few or no ELAV-positive cells are found in posterior margin dac clones (Fig. 5E,F). Using the dpp-lacZ reporter as an assay, we found that the furrow indeed progresses through dac null mutant clones, as long as they do not include the posterior margin of the eye disc (Fig. 5G). However, the rate of furrow movement seems to be slowed in such clones as compared to the surrounding wild-type tissue. Interestingly, this effect on the rate of furrow progression exhibits local domineering non-autonomy: the furrow is also slowed in the geno-typically wild-type epithelium immediately adjacent to dac mutant clones (not shown). This effect on furrow movement correlates with the appearance of additional photoreceptors in geno-typically wild-type ommatidia located next to dac mutant clones in adult eyes (Fig. 5J). Con-sistent with the phenotype of whole dac null mutant eye discs (Fig. 4C), the furrow does not initiate movement in clones that include the posterior margin of the eye disc. Once the furrow has moved past such a clone in the surrounding wild-type tissue, normal furrow progression anterior to the clone resumes (Fig. 5H,I,M). Thus, our clonal analyses indicate that loss of dac function prevents furrow initiation at the posterior margin of the eye disc and slows but does not prevent movement of the furrow during progression.

Although the morphogenetic furrow is able to move through patches of dac mutant cells, sections of dac mutant clones demonstrate that dac is required for normal ommatidial assembly: null mutant clones are clearly disrupted with most ommatidia containing an abnormal number and arrangement of photoreceptors (not shown). Nonetheless, complete and properly formed ommatidia composed entirely of dac mutant cells are occasionally observed (Fig. 5K), sug-gesting that there is no absolute requirement for dac in any given photoreceptor. In addition, mosaic ommatidia at the border of clones – those comprising both wild-type and dac mutant cells – are most often abnormally constructed, sug-gesting that dac is required cell autonomously for ommatidial assembly (not shown). That is, dac-dependent activity from surrounding wild-type cells is unable to rescue the aberrant omma-tidial structure of dac mutant clones.

Additional support for a role for dac during furrow progression comes from genetic analyses of interactions between dac and hedgehog (hh). hh is expressed in differentiating photoreceptors and is required for progression of the morphogenetic furrow (Ma et al., 1993; Heberlein et al., 1993). A lethal loss-of-function P-element insert in hh (rJ413) dominantly enhances the recessive eye phenotype of the weak dac^P allele (Fig. 1E,I). Similarly, loss of one copy of dac acts as a dominant enhancer of the recessive eye phenotype of the viable hh¹ mutation (Fig. 1J,K), which presents a significant reduction of the anterior portion of the eye (Mohler, 1988). These genetic results suggest that dac and hh act in the same or in parallel pathways during eye development.

We examined hh expression in wild-type and dac mutant imaginal discs using a lacZ enhancer trap insert in hh that faith-fully reproduces the normal expression pattern of hh (Ma et al., 1993). Since little or no neural development takes place in dac null mutant eye discs and hh is expressed in the eye specifi-cally in determined photoreceptors, it is not surprising that hh is not expressed in dac mutant eye discs (Fig. 6A,B). However, hh expression in other imaginal discs, including the antennal, leg and wing discs, is not affected by loss of dac function (for example, compare the antennal discs shown in Fig. 6A,B).

Expression of hh in differentiating photoreceptors lacking dac function was examined in homozygous dac null mutant clones in eye imaginal discs. We found that hh expression is normal in the absence of dac activity using a single copy of the viable hh enhancer-trap insertion P30 as an assay (Fig. 6C). Thus, dac function is not cell autonomously required for hh expression. The dominant enhancement of the adult dac mutant eye phenotype by a single copy of the lethal hh mutation rJ413 (Fig. 1I) is also apparent during larval development. As judged by the near absence of ELAV or hh expression in dac null mutant clones, loss-of-function of one copy of hh by the rJ413 lethal insertion results in delayed or reduced neural differen-tiation in such clones (Fig. 6D).

Substantial progress has been made toward elucidating the molecular mechanisms controlling pattern formation and mor-phogenesis during development of both vertebrate and inver-tebrate species. Genetic and molecular studies of embryonic development in Drosophila have been particularly fruitful in identifying and understanding the function of phylogenetically conserved genes that are essential for organizing tissue polarity, specifying segment boundaries and controlling cell-fate determination. In many cases, intercellular signaling is achieved by diffusion of a secreted protein acting over several cell diameters (Bryant, 1993). For example, the product of the segment polarity gene hh acts as a true morphogen in Drosophila, specifying cell fates in the embryo in a concen-tration-dependent manner (Heemskerk and DiNardo, 1994).

Vertebrate homologs of hh have recently been shown to exhibit organizing activity in the chick limb and in central nervous system development of mice (Riddle et al., 1993; Echelard et al., 1993). Similarly, nodal, a member of the TGFβ superfam-ily of secreted proteins, is expressed in the mouse node at the anterior of the primitive streak and is required for mesoderm formation (Zhou et al., 1993). A Drosophila homolog of TGFβ, dpp, acts as a morphogen controlling dorsal-ventral pattern formation in the embryo (Ferguson and Anderson, 1992).

Eye imaginal disc development in Drosophila provides an excellent system for deciphering mechanisms of pattern formation and morphogenesis. Organization of the unpatterned epithelium of the eye disc into a highly ordered array of dif-ferentiated cells follows movement of an indentation across the epithelial monolayer termed the morphogenetic furrow. Furrow progression is not associated with cell migration; instead, furrow movement is detected as a wave of changes in cell shape, cell cycle, gene expression and neural differen-tiation. Progression of the furrow requires the function of hh and is likely to involve dpp. The developmental events sur-rounding morphogenetic furrow movement in Drosophila may be analogous to those associated with the Spemann organizer in Xenopus and Hensen’s node and the zone of polarizing activity in chick: organizing centers determine cell fates of adjacent tissues through the action of secreted morphogens (reviewed in Slack, 1991; Riddle et al., 1993).

hh plays a central role in progression of the morphogenetic furrow during eye disc development (Ma et al., 1993; Heberlein et al., 1993). In this case, hh protein is expressed by determined neural cells posterior to the furrow and is thought to diffuse anteriorly, thereupon controlling expression of other genes required for furrow movement. One such putative target is dpp (Heberlein et al., 1993). Removal of hh function during furrow progression abolishes both the expression of dpp in the furrow as well as furrow propagation. dpp or hh, in turn, are likely to regulate the expression of other genes in the unpat-terned epithelium anterior to the furrow. Transplantation experiments have suggested that positional information suffi-cient to direct neural development is present in the morpho-logically undifferentiated epithelium immediately anterior to the furrow in the eye disc (Lebovitz and Ready, 1986). More recently, molecular analyses have shown that several genes are indeed expressed just anterior to the furrow, including string, eyes absent, hairy and atonal (Alphey et al., 1992; Bonini et al., 1993; Brown et al., 1991; Jarman et al., 1994). However, none of these genes expressed anterior to the furrow has been shown to participate directly in either initiation or propagation of the morphogenetic furrow.

While some of the genes involved in furrow progression have now been identified, little is known about the events sur-rounding initiation of the morphogenetic furrow. In this paper, we present evidence that dachshund (dac) is expressed at the posterior margin of the eye imaginal disc prior to movement of the morphogenetic furrow and is required for cells at the posterior margin to respond to pattern propagation signals such as those encoded by dpp and hh. This may reflect a require-ment for dac to control a cell-fate choice between retina and cuticle. Similar to other genes controlling pattern formation in multiple tissues, such as hh and dpp, dac is also required for normal leg morphogenesis.

In the absence of dac function, flies develop with severely shortened legs. dac is expressed in early larval leg imaginal discs and this expression continues into pupal development. dac activity is necessary to specify the identity of a subset of segments in the leg: the femur, tibia and proximal three tarsi. Loss-of-function mutations in dac result in legs with these intermediate structures fused and condensed. Thus, dac is required in those leg segment primordia in which it is expressed for cells to adopt a fate appropriate to their position. Greatly increased cell death is apparent in third instar dac mutant leg discs, prior to the elongation that occurs during early pupal development. Thus, dac function is required prior to leg disc eversion. dac is likely to act downstream of a proximal-distal patterning signal, such as Distal-less, to establish segment boundaries or identities (Cohen and Jürgens, 1989). The dac mutant leg phenotype is consistent with a model in which establishment of proximal and distal structures during limb development precedes elaboration of intermediate positional information (Cohen, 1993).

In the absence of dac activity, the morphogenetic furrow remains at the posterior margin of the eye imaginal disc. However, furrow movement is not prevented in dac mutants by a lack of sufficient epithelial tissue. Eye discs of dac null mutants are approximately normal in size, indicating that cell proliferation in the eye disc is not inhibited in the absence of dac function or furrow progression. This stands in contrast to other eyeless or reduced-eye mutants, such as eyes absent and Lobe, where a failure of retinal development is associated with larval eye discs that are significantly reduced in size (Bonini et al., 1993; Heberlein et al., 1993). While proliferation is normal, cells of dac mutant eye discs either fail to receive or perhaps respond to the normal cues specifying a neural identity. These cells eventually die, possibly as a result of inap-propriate cell-fate specification. Alternatively, these dying cells may reflect the action of a programmed cell-death mechanism designed to remove undifferentiated cells during development, thereby preventing disruption of normal tissue organization (Ma et al., 1993).

Although loss of dac function prevents all movement of the morphogenetic furrow in posterior lateral regions of the eye disc, some furrow movement at the midline of the disc is observed. Here, the furrow is often able to move a small distance and this is sufficient to allow differentiation of a few photoreceptor cells. These cells develop as apparently normal photoreceptors, able to support the determination of the cone, pigment and bristle accessory cells found in wild-type ommatidia. Thus, dac is not required specifically for photore-ceptor differentiation. In support of this view, photoreceptor development also occurs in interior dac mutant clones in the eye, indicating that there is no cell-autonomous requirement for dac in photoreceptor determination.

The eye phenotype of weak dac mutants is different from other mutants that affect furrow movement. For example, the partial loss-of-function hh¹ allele eliminates dpp expression and furrow progression first in the middle of the eye disc, giving the adult eye a kidney-shaped appearance (Fig. 1J and Heberlein et al., 1993). In contrast, in weak dac mutants, dpp expression is unaffected and the furrow progresses nearly normally at the center of the eye disc but is delayed at the periphery (Fig. 4B). This difference in furrow progression may reflect a requirement for dac, but not hh, during furrow initiation.

Comparison of the phenotype of dac and hh mutant clones at the posterior margin of the eye disc suggests that dac plays an essential role in either furrow initiation or retinal cell-fate determination: while a few ommatidial columns are observed in hh mutant clones located at the posterior margin of the eye (Ma et al., 1993), dac mutant clones that include the posterior margin usually fail to give rise to retinal tissue (Fig. 5B,C,L). Consistent with this adult phenotype, the dpp-lacZ reporter expression remains at the posterior margin in dac mutant clones in the eye disc (Fig. 5H,I,M). Thus, our clonal analysis indicates that dac is cell autonomously required for initiation of movement of the morphogenetic furrow away from the posterior margin of the eye imaginal disc. Moreover, the failure of furrow initiation in dac mutant clones is unlikely to result from either a failure of photoreceptor differentiation or the absence of hh expression, since diffusion of hh protein from nearby wild-type photoreceptor cells is unable to rescue furrow initiation in dac mutant clones. Thus, dac is clonally required for cells at the posterior margin of the eye disc to respond to pattern propagation signals normally sufficient to cause furrow movement. This may reflect a direct requirement for dac in furrow initiation. Alternatively, dac may be required for an earlier step of cell-fate determination such that clones of dac mutant cells at the posterior margin of the eye disc are unable to differentiate as retinal tissue.

In addition to an essential role in early furrow movement, dac is also required for some aspect of ommatidial assembly. dac mutant clones in the adult eye are comprised of mostly abnormal ommatidia, even at the border of such clones. This is in sharp contrast to hh mosaic studies, where even large hh mutant clones are normally constructed (Ma et al., 1993; Heberlein et al., 1993). That is, unlike furrow progression, dac function is required cell autonomously for ommatidial assembly. Since dac is a nuclear protein, it may be involved in regulating other genes functioning in this process. One candidate is the Egfr gene. dac mutants were isolated by their ability to dominantly suppress the rough-eye phenotype of the activated Ellipse allele of Egfr. We have also shown that loss-of-function mutations in dac and Egfr display an apparent recessive synergy during eye development. Both results suggest that dac may positively regulate Egfr. Consistent with the possibility that dac controls Egfr expression and that this effects part of the dac mutant phenotype, Egfr is expressed anterior to the morphogenetic furrow (Zak and Shilo, 1992) and is required for normal photoreceptor determination (Clifford and Schüpbach, 1989; Xu and Rubin, 1993).

Three lines of evidence suggest that a primary initiation signal acts at the posterior midline of the eye disc. First, as judged by dpp expression, the furrow normally initiates movement at the midline of the eye disc. Second, neural-specific markers, such as ELAV and hh, first appear at the posterior midline. Third, in the absence of dac function, the only movement of the furrow occurs at the posterior midline. Since this movement is observed in about half of the discs examined, dac may serve some role in receiving this primary initiation signal but is not absolutely required for this step. This initial step cannot require hh function because hh is not expressed until several rows of ommatidia posterior to the furrow. The initial signal that triggers furrow movement may be transmitted directly via the optic stalk, thereby localizing initiation to the posterior midline. Alternatively, there may be a diffusable primary initiator, such as an ecdysone pulse (Richards, 1981), that is received only by specialized cells at this position.

The expression patterns of dpp and ELAV indicate that the furrow does not begin moving away from the entire posterior margin of the eye disc at once. Instead, the furrow advances as a linear front such that movement away from the curved posterior margin occurs over time as a function of the distance from the midline of the eye disc. The simplest mechanism to control propagation of pattern formation and neural differen-tiation in this manner would be by diffusion of a secreted signaling molecule such as hh. Transmission of this signal does not require dac activity since normal furrow movement resumes lateral to dac mutant clones located at the posterior margin of the eye (Fig. 5). That is, lateral propagation of signals for furrow movement away from the posterior margin of the eye disc is not prevented by intervening clones of dac mutant cells.

Our clonal analysis demonstrates that dac function is cell autonomously required for cells at the posterior margin of the eye disc to respond to pattern propagation signals from sur-rounding wild-type tissue. This requirement is not found in cells located in interior portions of the eye disc. In most posterior margin dac clones, both in larvae and in adults, there is a morphological boundary that forms between dac mutant clones and the adjacent retinal field (Fig. 5F,L), suggesting that dac mutant cells that fail to develop as retina possess different adhesive properties than those of the surrounding cells. The eye imaginal disc is fated to give rise not only to the retina but to a portion of the surrounding head cuticle as well (Cohen, 1993). A boundary similar to that observed in posterior margin dac clones is normally seen at the periphery of wild-type eye discs (Fig. 3D), presumably representing a physical distinction between cuticle and retinal fields. Such boundaries are not observed in interior dac clones. These observations suggest the possibility that dac may be involved in controlling a choice between a cuticular versus a retinal fate. Specifically, dac may be required to suppress a cuticle fate in cells derived from the posterior margin. In the absence of dac function, clones that originate from a posterior margin cell follow a cuticle pathway and are usually unable to respond to the pattern propagation signals, such as dpp and hh, that are present but apparently ineffective. Consistent with this model, dac is not expressed in the peripheral margin of the eye disc destined to become cuticle (Fig. 3D). Thus, dac is at least indirectly required for initiation of the morphogenetic furrow and the onset of neural differentiation. However, this requirement is not absolute since mutant ommatidia bordering posterior margin clones, although usually few in number, are frequently observed. Moreover, if a dac clone is derived from an ‘interior’ cell of the eye disc, these cells appear to respond relatively normally to differen-tiation signals. This difference in responsiveness to pattern propagation signals suggests that posterior margin cells may be different from other cells in the eye disc from an early stage of development. Consistent with this view, both dpp and dac are expressed specifically at the posterior margin of the eye disc prior to furrow initiation or neural differentiation.

In addition to dac, the product of the dpp gene may also be involved in some aspect of furrow initiation. This seems likely for three reasons. First, like dac, dpp is expressed along the posterior margin of the eye disc, prior to furrow initiation (Masucci, et al., 1990; Ma et al., 1993). Second, mutant dpp clones located at the posterior margin of the eye disc result in a reduced retinal field and fail to give rise to ommatidia at the posterior border of such clones (Heberlein et al., 1993). Finally, dpp expression is unaffected in a dac mutant back-ground, indicating that dpp is likely to either be genetically upstream of dac or act in a parallel pathway. Since dac plays at least an indirect role in furrow initiation, these observations suggest that dpp may also be involved. However, a direct role for dpp in this process remains to be demonstrated.

Consistent with the proposal that there are distinct mecha-nisms for furrow initiation and progression, dpp expression is hh-independent prior to initiation of the morphogenetic furrow. dpp is expressed at the posterior margin of the eye disc prior to neural determination and this expression persists in a dac mutant background. Since hh is expressed in the eye only in cells posterior to the morphogenetic furrow and there is little or no furrow movement in the absence of dac function, dpp expression must not require hh activity prior to furrow initiation. However, once the furrow has begun moving across the eye disc, the product of the hh gene is required in differ-entiating photoreceptors for continued expression of dpp in the furrow and for furrow propagation.

We thank Iain Dawson for generously providing dac mutant alleles and other data prior to publication. We also thank everyone in the Rubin laboratory for both technical and intellectual support during this work. In particular, Tanya Wolff for her expertise in the biology of eye imaginal disc development, Smadar Admon and Kelli Lopardo for their assistance in generating monoclonal antibodies, Ulrike Gaul for providing the P-element insert in dachshund, Todd Laverty for chromosomal in situ analysis, Don Pardoe for scanning electron microscope analysis, Dianne Fristrom for discussions of leg develop-ment, and Iain Dawson, Steve Harrison, Ulrike Heberlein, Alex Kolodkin, Pat O’Farrell, Andrew Tomlinson, Jessica Treisman and Tanya Wolff for helpful comments on the manuscript. G. M. was supported by an NIH postdoctoral fellowship.