The Drosophila eyes absent gene directs ectopic eye formation in a pathway conserved between flies and vertebrates

The Drosophila eye, although structurally distinct from the vertebrate eye, shows striking parallels at the molecular level. Many genes that function in eye formation in the fly have homologs that are expressed during vertebrate eye development (Quiring et al., 1994; Zuker, 1994; Oliver et al., 1995). Among these is the Drosophila eyes absent (eya) gene which encodes a nuclear protein that, in the fly, functions prior to the first notable differentiation event-morphogenetic furrow formation – in eye progenitor cell development (Bonini et al., 1993; Leiserson et al., 1994). eya is the founding member of a class of vertebrate Eya genes, with homologs showing expression in the eye and other tissues (Abdelhak et al., 1997; Duncan et al., 1997; Xu et al., 1997; Zimmerman et al., 1997). Human mutations at the Eyal locus have been identified that result in defects in organ formation (Abdelhak et al., 1997).

eya functions at a similar time and place as the fly eyeless gene. eyeless is a counterpart of the human ANIRIDIA and mouse Sey (Small eye) genes, which are Pax-6 family members containing a paired-box and homeobox and likely function as transcription factors (Quiring et al., 1994). Mutated eyeless results in loss of the fly eye (Quiring et al., 1994); similarly, mutation of the human or mouse counterparts leads to eye malformation and reduction of the eye in the extreme, and cataracts in mild forms (Hogan et al., 1986; Hill et al., 1991; Ton et al., 1991; Glaser et al., 1992; Jordan et al., 1992; see Hanson and van Heyningen, 1995). Expression of the eyeless cDNA in the fly using tissue-specific elements can direct the formation of ectopic eyes in the antennae, legs and wings (Halder et al., 1995a). The mouse Sey cDNA, when introduced into the fly, can similarly direct ectopic Drosophila eye development, suggesting potential conservation of fundamental molecular features by which these homologs act in eye formation. Sey mouse mutants have reduced expression of the Eya1 and Eya2 genes in eye progenitor tissue (Xu et al., 1997), suggesting that Eya genes might be mediators of Pax-6 function for eye formation in vertebrates.

Since the vertebrate Eya homologs show expression in eye tissue, it is tantalizing to speculate that the vertebrate and fly genes are functional homologs. Here, we address the level of functional conservation between fly eya and a vertebrate Eya homolog. We then used the fly to define the relationship between eya and eyeless that was suggested by vertebrate work: that the eya gene may be a conserved mediator of eyeless function in eye formation. We found that not only is eya essential for eyeless function, but that the eya gene itself can serve as a master control gene for eye formation. These data indicate striking functional conservation of the genetic pathway of eye formation between flies and vertebrates, and suggest a model of combinatorial gene functions for eye formation.

Fly strains were grown on standard molasses, yeast and cornmeal medium at 25°C. UAS-eya transgenics were made by subcloning the full-length Drosophila eya typel cDNA into the pUAST vector (Brand and Perrimon, 1993). UAS-Eya2 mouse transgenes were made by subcloning a predicted full length mouse Eya2 subclone (Zimmerman et al., 1997) into the pUAST vector. Flies were transformed using standard transgenic techniques (Rubin and Spradling, 1982). eyeless-GAL4 was made by subcloning a fragment which reports eyeless staining in eye progenitor cells (Quiring et al., 1994), into the GAL4 vector (Brand and Perrimon, 1993), and transforming the construct into flies. Other GAL4 lines were obtained from Drosophila Stock Centers; eya mutant alleles are as previously described (Bonini et al., 1993); eyeless² mutant strain was provided by courtesy of Dr Walter Gehring. GAL4 lines used included T59, T155 and dpp-GAL4, all of which express in the eye portion of the eye-antennal disc. dpp-GAL4 is expressed in the imaginal discs in an expression pattern similar to that of dpp (Staehling-Hampton et al., 1994; see also Shen and Mardon, 1997). eyeless-lacZ lines are as described by Quiring et al. (1994); UAS-eyeless lines are as described by Halder et al. (1995a).

Tissue preparations were fixed with 2% paraformaldehyde in TBS, permeabilized with 0.5% Triton X-100, and stained in primary antibody overnight. Primary antibodies were anti-Eya (Bonini et al., 1993) and anti-Glass (Ellis et al., 1993). After rinsing in TBS, tissue was stained with secondary antibodies conjugated to fluorescein or cyanine-3 (Jackson ImmunoResearch Laboratory), rinsed in TBS, then mounted in PDA-glycerol, as previously described (Bonini et al., 1993). Staining with P-galactosidase for eyeless-lacZ expression pattern was performed as described by Quiring et al. (1994). For viewing flies by scanning electron microscopy, flies were critical point dried and scanned at 5 kV. For sections of eye tissue, flies were fixed in glutaraldehyde, embedded in epon and thick sectioned (Bonini et al., 1993).

To test potential functional homology between the vertebrate and fly eya genes, we asked whether a vertebrate Eya homolog was capable of replacing the fly eya gene in the eye developmental pathway. To do this, we required functional replacement of eya activity in the eya² mutant background. The eya² mutant is a viable, eye-specific null for eya gene activity in eye progenitor cells anterior to the furrow (Bonini et al., 1993). These mutant flies are completely eyeless due to complete loss of eye progenitor cells by cell death (Fig. 1A). Thus, any survival and development of ommatidia to the adult eye in this mutant background would indicate functional replacement of fly eya gene activity.

To express the vertebrate gene, the GAL4-UAS system of tissue-specific targetting was used (Brand and Perrimon, 1993). UAS-Eya2 transgenic animals were made using a mouse Eya2 cDNA that is predicted to encode a full-length protein (Zimmerman et al., 1997). The protein encoded by this homolog shows an overall identity of 49% with the predicted fly protein sequence. To express the gene sufficiently early in eye progenitor cell development, we constructed an eyeless-GAL4 transgenic line that drives GAL4 expression in eye progenitor cells prior to furrow formation.

Remarkably, the mouse Eya2 gene restored eye formation to eya² mutant flies (Fig. 1A,B). Sections through the eyes indicated a normal pattern of photoreceptor neurons (Fig. 1C). These data indicate that the mouse Eya2 gene can functionally replace the fly gene in eye formation. These data demonstrate that molecular features of the eye developmental pathway directly dependent on eya gene activity are conserved between flies and vertebrates.

Given this striking level of conservation of eya function, we were interested to determine the relationship between eyeless gene activity and eya gene activity suggested by expression studies in vertebrates. To determine whether eya gene expression occurred upon eye formation directed by the eyeless gene, we generated ectopic eyes with eyeless, and stained the tissue for ectopic expression of Eya protein. Flies bearing a UAS-eyeless insert (Halder et al., 1995a) were crossed to a dpp-GAL4 insert line, which expresses GAL4 in the imaginal disc expression pattern of dpp (Blackman et al., 1991). These animals generated ectopic eyes on the legs, wings and antennal region of the head (Fig. 2A,C; Table 1).

Immunostaining of third-instar larval imaginal discs confirmed that Eya was indeed ectopically expressed in regions where eyeless directed ectopic eye formation: the antennal portion of the eye-antennal disc, leg and wing discs (Fig. 2F,I, data not shown). Of these imaginal tissues, Eya is normally expressed only in eye and ocellar progenitor cells of the eye-antennal disc, and the peripodial membrane of the wing disc; Eya is not normally expressed in cells of the antennal, leg or wing disc proper (Fig. 2H,K, data not shown; Bonini et al., 1993). Activation of eya gene expression by eyeless suggested that eya may indeed be required for formation of eyes by eyeless, similar to the requirement for eya function during normal compound eye development.

To test whether eya gene activity was essential for eyeless-driven ectopic eye formation, we attempted to induce ectopic eyes with UAS-eyeless crossed to dpp-GAL4, but now in the eya² mutant background. As noted, the eya² mutant is completely eyeless and null for the early eye function of the eya gene. If eye formation by eyeless were dependent upon eya gene activity, then ectopic eye formation should fail in the eya² mutant background. Eye formation indeed failed in all tissues of the fly where ectopic eyes had previously developed: the legs, wings and antennal segments of the head (Table 1, Fig. 2B,D). In the imaginal tissues, as anticipated, no ectopic Eya protein expression was detectable in the antennal, leg or wing discs of animals bearing UAS-eyeless in trans to dpp-GAL4 in the eya² mutant background (Fig. 2G,J). These experiments also demonstrated that UAS-eyeless was not able to restore normal eye formation to the eya² mutant (Fig. 2B), indicating that eyeless gene activity cannot replace or substitute for the function of eya in eye development. Taken together, these date clearly indicate that eya gene function is essential for eyeless to form eyes; the eya gene thus appears to be an essential biological target of eyeless gene activity in eye formation.

Given the high conservation demonstrated above of the eya pathway at the functional level between flies and vertebrates, and given the essential role of eya for eyeless function that may well extend between flies and vertebrates, we were interested to determine what were potential effects of the eya gene itself for eye formation. Could eya, like eyeless, mediate eye formation? Previously, we had expressed the eya gene in the fly with a heat shock promoter (Bonini et al., 1993); whereas such expression could restore the eyes to eya mutants that lacked eye formation, no other consistent effects were observed in the rescued animals. Nevertheless, we attempted to express eya at higher levels using the GAL4-UAS system (Brand and Perrimon, 1993), to determine whether it was possible to induce dominant phenotypes that would yield clues to the function of the gene.

A UAS-eya construct was made with the eya cDNA. To determine whether the construct was functional, we attempted rescue of the eya² mutant phenotype using various GAL4 lines that express in the eye progenitor field prior to furrow formation. A number of GAL4 lines were tested, including eyeless-GAL4, <3pp-GAL4, T59, and T155. These GAL4 insertions, when crossed to a UAS-eya insert, could restore to the eya² mutant eyes up to three quarters of normal size. Given that rescue was partial, we attempted to increase eye size by increasing gene dosage of the UAS-eya and GAL4 insertions. In doing this, we found that UAS-eya lines in trans to GAL4 insertion lines were lethal in two doses of the transgenes; for those lines that were lethal at the late pupal stage (dpp-GAL4, T59), the homozygous animals could be observed by dissection of the pupae. This analysis showed that these lines displayed not only rescue of the eye, but also ectopic eye formation in other regions of the animals where GAL4 was expressed with these constructs. The chromosomes bearing the UAS-eya and GAL4 inserts were then crossed out of the eya mutant background and into a normal background for additional analysis. We focused on expression of UAS-eya driven by dpp-GAL4, since this combination led to late pupal lethals that could be readily observed; in these animals, GAL4 expression occurs in the imaginal disc expression pattern of dpp, in the eye and antennal portions of the eye-antennal disc, the leg and wing discs, among other tissues (Blackman et al., 1991).

In a wild-type background, UAS-eya dpp-GAL4 in single copy generated rare examples of ectopic eyes, which resembled normal eyes, on the antennal segment (10% of the animals, Fig. 3A,B). Tangential sections of the ectopic eyes formed indicated that photoreceptor cells developed in a pattern similar to that of the normal compound eye (Fig. 3C). With two copies of UAS-eya dpp-GAL4, ectopic eye formation was induced in the antennal region of the head in almost all animals (96% ectopic eye formation on antennae); 80% also showed ommatidial formation on the legs, and occasionally on the wings. Glass, a photoreceptor-specific gene, was used as a marker to detect development of retinal tissue in the larval imaginal discs. Glass expression is normally restricted to the eye portion of the eye-antennal disc and does not occur in the antennal portion or other imaginal discs (Ellis et al., 1993). Ectopic expression of Glass was seen in the antennal and leg imaginal discs; in these tissues, rosettes of developing photoreceptor clusters similar to the normal pattern were seen (Fig. 3D,E). These data indicate that eya has the capacity to function as a master regulatory gene for eye formation.

These observations raised questions regarding the relationship between the eya and eyeless gene functions during eye formation. Since eya was essential for ectopic eye formation by eyeless (see Table 1; Fig. 2), was eyeless gene function essential for ectopic eye formation by eya? To address this, we first asked whether eyeless gene expression was induced during ectopic eye formation directed by the eya gene. Normally, eyeless expression is restricted to the eye portion of the eye-antennal imaginal disc (Fig. 4A; Quiring et al., 1994). In UAS-eya dpp-GAL4 animals, expression of eyeless occurred ectopically in the antennal region of the eye-antennal disc, in the region where eya directed ectopic eye formation (Fig. 4B). Although eya also directed ectopic eye formation in the leg discs, ectopic eyeless expression was not detectable in that tissue upon eya expression (Fig. 4C); eyeless was capable of autoregulation in leg discs when eyeless itself was ectopically expressed (Fig. 4D). This suggested that eyeless might be required for eya activity to form eyes in some tissues, but dispensible in others. We also determined that eyeless remained expressed in the eye progenitor cells of the eya² mutant (Fig. 4E), suggesting that eya gene function is not essential for the normal expression pattern of eyeless.

To address a functional requirement for eyeless, we investigated whether there were detectable genetic interactions between the eyeless and eya genes. Such experiments are limited by the mutants of eyeless currently available-there are no null mutants for the eye function of eyeless which would allow us to remove eyeless gene activity completely (Quiring et al., 1994). Nevertheless, we attempted to induce ectopic eye formation with eya in a background of reduced eyeless gene activity, by using the eyeless² mutant allele. eyeless² mutant flies show a range of reduced eye phenotypes, with about 30% of the flies missing at least one eye completely. Directed expression of the eya gene in the eyeless² background, did not result in ectopic eye formation in the antennal segments of the head, or the legs (data not shown). These data indicated a dependence on eyeless gene activity in the ability of eya to direct eye development both in the head and in the legs. Thus, eya appears to function both downstream and upstream of eyeless gene activity in eye formation.

This regulatory relationship between the two genes prompted us to ask whether we could detect additional interactions between the genes. To do this, we examined the ability of eyeless to direct eye formation when combined with additional doses of eya gene activity. Animals bearing UAS-eya in trans to dpp-GAL4 show limited dominant effects (see above and Table 2); in animals bearing UAS-eyeless in trans to dpp-GAL4, Eya protein is already highly expressed (see Fig. 2). Nevertheless, ectopic eye formation by eyeless was dramatically enhanced when additional eya gene activity was provided (Table 2, Fig. 5). The ectopic eyes were larger and formed with higher penetrance than with eyeless or eya alone, and eye formation now occurred on the genitalia, a condition never previously observed in individuals with either gene alone (Table 2 and Fig 5B,E). This effect did not appear additive (Table 2). Rather, these data suggest functional synergy between eyeless and eya gene activities in eye formation.

Our data reveal an active role of the eya gene in eye formation, and suggest a model of gene regulatory interactions between eyeless and eya in eye formation in the fly that may extend to their mammalian counterparts.

We found that a vertebrate homolog of eya, the mouse Eya2 gene, can functionally replace the fly gene in eye formation. These data suggest that the role of the eya gene in eye formation has been conserved through evolution, between flies and vertebrates, despite dramatic differences in eye structure between the two (see Zuker, 1994). Such functional homology has been shown for various Pax-6 homologs of the eyeless gene (Halder et al., 1995a; Glardon et al., 1997): we have extended those studies to eya and its homologs, a second gene of the eye developmental pathway. The vertebrate Eya homologs identified to date are all expressed in the developing or adult eye, suggesting all homologs may function in aspects of vertebrate eye formation and maintenance (Duncan et al., 1997; Xu et al., 1997; Zimmerman et al., 1997). For the Eya2 homolog, we demonstrate here a homologous role in the eye developmental pathway.

We have addressed and clarified the relationship of eya gene activity to that of the eyeless gene. Previous data indicate that, normally, eyeless expression precedes that of eya in eye progenitor cells. Whereas eyeless is expressed in the eye primordium from embryonic stages (Quiring et al., 1994), eya expression is initiated during the mid-larval stages (Bonini et al., 1993). We found that eya gene activity was essential for eye formation by eyeless; these data, along with the observation that eyeless remains expressed in fly eya mutants, suggest that eya is downstream of eyeless gene function (Fig. 6). Consistent with this, Eya expression was induced upon ectopic eyeless expression; mammalian Eya expression is also affected by mutation of Pax-6 in the mouse (Xu et al., 1997). eya/Eya thus appears to be essential for eye formation by eyeless /Pax-6. In humans reduction of EYA gene activity may be a critical consequence of mutation in the ANIRIDIA gene, leading to improper eye formation.

In the eya mutant background, we note that eyeless activity was not completely ineffective-although ectopic eye formation did not occur, leg development remained severely affected (see Fig. 2 D,J). This suggests that eyeless is activating gene functions, in addition to that of eya, that are interfering with normal leg development. These additional genes may be upstream of eya in the eye developmental pathway, or in a different branch of the eye developmental pathway (see below and Fig. 6). Furthermore, target genes of eyeless may have functions in addition to those associated with eye development. For example, the dac (dachshund) gene is a target of eyeless gene activity that functions both in eye and leg development (Shen and Mardon, 1997).

We found that eya shares with eyeless the capacity to function as a ‘master control gene’ for eye formation. By this term, we refer to the fact that eya has the capacity to direct the appropriate genetic program of the many genes required for eye development (Halder et al., 1995a). Using loss-of-function eyeless mutants, we found evidence that eyeless activity is required for eya to form eyes-similar to the requirement for eya gene activity in the proper eyeless function in eye formation. Thus, the activities of the eya and eyeless genes appear connected by a regulatory loop, with each functionally required by the other in eye formation (Fig. 6). One qualification of these conclusions is that in leg discs, we were unable to detect eyeless expression upon ectopic eya activity. Nevertheless, genetic studies in the eyeless mutant background indicated that eye formation by eya in legs appeared dependent on eyeless gene activity. Thus, we suggest that eyeless function is required for ectopic eye formation by the eya gene. In Fig. 6, we place eya downstream of, but connected back to, eyeless gene function. Genetically, there is little to argue which gene is first; however, the normal expression patterns of the genes indicate that eyeless expression temporally precedes that of eya during normal eye formation (Bonini et al., 1993; Quiring et al., 1994). In ectopic eye formation, the genes are each essential for the others’ function, thus are interchangeable in placement.

Moreover, we found that eya and eyeless displayed functional synergy in eye formation – the same dosage of eyeless was potentiated when combined with additional eya gene function. This synergy was observed with a dpp enhancer construct, and whether other regulatory elements will mediate a similar level of synergy remains to be determined. However, that eyeless has functions in addition to activating eya activity is also suggested by the severely affected leg morphology observed upon eyeless expression in the eya mutant background (see above). These data suggest that eyeless and eya may function in at least partially distinct pathways for eye formation (Fig. 6). By such a model, expression of either gene alone, if expressed strongly enough, will eventually drive both pathways because they form part of a regulatory loop. However, when both genes are expressed strongly, both pathways will be strongly driven, leading to an enhancement of eye formation compared to that with either gene alone.

Taken together, these data suggest that early events of eye formation proceed not by a simple linear pathway, but rather by a combinatorial code of gene function. Thus, there may be no single ‘master control gene’ for eye formation, but a complex regulatory network of gene activities required to trigger the biological event of eye development. These initial events of eye formation include additional genes, such as dac and sine oculis. The position of dac in this regulatory pathway will be of interest, as dac has also been shown able to direct eye formation (Shen and Mardon, 1997). Given the increase in eyeless expression upon dac-directed eye formation (Shen and Mardon, 1997), dac likely also requires eyeless gene activity to form eyes. This suggests that, minimally, dac and eya are connected through common regulation of, and regulation by, eyeless gene activity.

The relationship between eya, eyeless and other genes central to eye formation conserved in both flies and vertebrates, such as sine oculis/Six-3 (Cheyette et al., 1994; Serikaku and O’Tousa, 1994; Oliver et al., 1995), are also of great interest. Moreover, how these early events of eye determination subsequently merge with pattern formation events of furrow movement (Heberlein and Moses, 1994) and cell cycle regulation (Thomas and Zipursky, 1994), are key aspects of generating a patterned neural structure like an eye. Somehow, these different aspects of the eye developmental process must be triggered by the eye differentiation pathway and coordinately regulated, to achieve this exquisitely organized neural center.

The eya gene has additional roles in development. In flies mutations in eya can be embryonic lethal (Nüsslein-Volhard et al., 1984; Bonini et al., 1993; Leiserson et al., 1994), or result in defects in gonad formation (Boyle et al., 1997), whereas humans mutant in EYA1 show developmental defects of the branchial arches, ear and kidney (Abdelhak et al., 1997). Thus, the eya gene has roles in development of the animal in addition to a function in eye formation. It is thus of interest to determine whether expression of eya has consequences over and above ectopic eye formation. Expression in other tissues or at other times in development may lead to elucidation of additional roles of the gene, as well as insight into the specificity of expression for eye formation. Toward this end, it is rather surprising that genes like eyeless, eya and dac, with roles in the animal in addition to eye formation, should induce eye development when ectopically expressed. What leads to this specificity is of particular interest. With respect to the role of eya in eye formation, loss-of-function eya mutants show death of eye progenitor cells (Bonini et al., 1993).

Taken together, these data indicate that, although the function of eya in eye differentiation is coupled to both differentiation and survival, the most dramatic effects of the gene upon strong expression are in the differentiation pathway. Thus, we anticipate that the relationship of eya activity with cell death will be indirect, through an effect on the pathway of differentiation. What gene activites might be altered in eya mutants, such that the cells become directed down a death pathway, remain to be defined.

With respect to the eye developmental pathway, the biological activities of eyeless and eya in eye formation extend to their mammalian counterparts Pax-6 and Eya. The mouse Sey gene has been shown to function in the fly (Halder et al., 1995a) and we have shown that a mouse homolog of eya has the ability to functionally replace the endogenous fly gene in eye development. These data indicate a remarkable level of conservation of gene function in eye formation between flies and mammals. These data lend support to the idea (Quiring et al., 1994; Halder et al., 1995a,b; see Zuker, 1994) that common genetic pathways may be used for the formation of eyes of widely divergent structure in organisms as evolutionarily distant as flies and man.

We thank our many colleagues in the fly and vertebrate communities, especially in the laboratories of Dr Walter Gehring, Norbert Perrimon, Graeme Mardon, Scott Poethig, and the Drosophila Stock Centers, who have generously provided reagents. We thank Drs Laura Lillien, Anthony Cashmore and Mark Fortini for critical comments. This research has been funded, in part, by grants from the National Eye Institute (EY11259), the John Merck Fund, the University of Pennsylvania Research Foundation (to N. M. B.), and a Vision Center Training Grant (to J. W.).