The Eye Specification Network in Drosophila

Abstract

One of the most exciting revelations in retinal biology is the realization that the molecules and mechanisms that regulate eye development have been conserved in all seeing animals including such diverse organisms as the fruit fly, mouse and man. The emerging commonality among mechanisms used in eye development allows for the use of model systems such as the fruit fly, Drosophila melanogaster, to provide key insights into the development and diseases of the mammalian eye. Eye specification in Drosophila is controlled, in part, by the concerted activities of eight nuclear proteins and several signal transduction cascades that together form a tightly woven regulatory network. Loss of function mutations in several components lead to the complete derailment of eye development while ectopic expression of threse genes in non-retinal tissues can direct the fates of these tissues towards eye formation. Here we will describe what is currently known about this remarkable regulatory cassettee highlight some of the outstanding questions that still need to be answered.

Introduction to the Fly Eye

The compound eye of Drosophila has often been affectionately referred to as a ‘neurocrystalline’ lattice because of its near perfect cellular structure (Ready et al. 1976). This description is certainly exemplified by the structural simplicity observed in both surface views and retinal sections of the adult eye. The adult compound eye is comprised of approximately 800 repeating unit eyes or ommatidia that are organized into a hexagonal array (figure 1A). Each unit eye contains an invariant number of photoreceptor neurons and accessory cells that are can be unambiguously identified by their position within the ommatidium (Dietrich 1909, Ready et al. 1976, Held 2002). Underlying this near perfect cellular architecture is a stereotyped developmental sequence in which each ommatidial unit undergoes an identical series of recruitment steps, gene expression patterns and morphogenetic movements (Ready et al. 1976, Tomlinson & Ready 1987b, Tomlinson & Ready 1987a, Cagan & Ready 1989, Ready 1989, Wolff & Ready 1991, Wolff & Ready 1993, Kumar & Moses 1997). Thus the events that generate a single unit eye are repeated nearly 800 times in the retina (figure 1).

Figure 1.

Structure and Develoment of the Drosphila Retina (A) Scanning electron micrograph of the external surface of the fly eye. (B) Light microsope section revealing the arrangement of ommatidia within the retina. (C) Confocal microscope image of a third larveal instar eye-antennal disc. Red = F-actin, green = ELAV, a pan neuronal marker. (D) A schematic diagram of the ommatidial recruitment series. Anterior is to the right.

The decisions that produce individual photoreceptor neurons and accessory cells occur relatively late in the development of an eye. A much earlier choice - to become an eye or some other tissue- takes center stage during the embryonic and early larval stages of development. During Drosophila embryogenesis small groups of cells (called polyclones) invaginate from the surface of the embryo, develop asynchronously from the remaining cells and ultimately give rise to monolayer epithelia called imaginal discs (Cohen 1993, Held 2002). These cellular sheets serve as templates for the development of all adult derivatives such as the compound eye, antenna, legs, wings, halteres and genitals. This review will describe the regulatory network that regulates the decision to become an eye and the initial steps of pattern formation in the retina, the initiation of the morphogenetic furrow.

eyeless is a Master Regulator

For the better part of the twentieth century the predominant view of eye development was that the eye had a polyphyletic origin, meaning that it evolved independently several times during the course of evolution. In fact it had been estimated that the eye evolved at least 40 and possibly 65 different times (Salvini-Plawen & Mayr 1977). Theoretical measurements have that this is feasible on an evolutionary time scale (Nilsson & Pelger 1994). This was in part due to the observation that there were examples of closely related species having different types of eyes while distantly related organisms had very similarly designed retinas. Much if not all of the analyses were based on morphological traits. However, in the mid-1990s a spectacular set of observations turned the current thinking on its head. The Gehring group in Switzerland had just cloned and sequenced the eyeless gene from Drosophila and showed that it was homologous (at least in sequence) to the mouse Small eye gene (Quiring et al. 1994). They then demonstrated that expression of eyeless in non-retinal tissues such as the antenna and legs could redirect these tissues into adopting an eye fate (Halder et al. 1995a) (figure 2). They further evidenced that the mouse Pax6 protein could not only rescue the eyeless mutant phenotype but could also induce ectopic eyes in Drosophila (Halder et al. 1995a). This was a crucial contribution since it was already known that mutations within the eyeless gene led to an inhibition of eyes in flies and lesions within the human Pax6 gene were the underlying cause of Aniridia, a retinal disorder in which afflicted patients fail to correctly undergo lens development (Hill et al. 1991, Ton et al. 1991). These results led to the immediate recognition of new parallels between invertebrate and mammalian retinal development. The significance of the role played by eyeless in eye development has continued to grow as it has been shown that Pax6 homologs are present in animals representing all the major phyla and has been shown to regulate eye development in all cases that have been experimentally tested (Gehring 1996, Loosli et al. 1996, Callaerts et al. 1997, Glardon et al. 1997, Tomarev et al. 1997, Glardon et al. 1998, Nornes et al. 1998, Callaerts et al. 1999, Nishina et al. 1999, Terzic & Saraga-Babic 1999, Strickler et al. 2001, Gehring 2002, Onuma et al. 2002). It seemed that the eyeless/Pax6 gene occupied the central position within the regulatory hierarchy controlling eye formation. The seemingly universal properties of Pax6 to regulate eye development quickly resulted in the crowning of eyeless as a “master regulator” gene (Halder et al. 1995b, Gehring 1998, Gehring & Ikeo 1999).

Figure 2.

The eyeless Gene is Necessary and Sufficient to Support Eye Formation (A-C) Scanning electron micrographs of wild type and mutant Drosophila heads. (A) wild type. (B) Loss-of-function eyeless mutant. Note the complete loss of retinal tissue. (C) Ectopic eye resulting from mixexpression of eyeless within the antennal segment. Arrow indicates location of ectopic eye. Anterior is to the right.

The evolutionary history of Pax6 activity is of considerable interest since loss-of-function mutations in both Drosophila and C. elegans homologs have been shown to produce defects in head development (Chisholm & Horvitz 1995, Kronhamn et al. 2002). Additionally, both eyeless and twin of eyeless are initially expressed within the entire eye-antennal imaginal disc. The expression of both Pax6 homologs are restricted to the eye disc only later in development, just prior to the initiation of the morphogenetic furrow (Kumar & Moses 2001a). Such expression patterns and mutant phenotypes raise the possibility that the original role for Pax6 was in specifying the head and that its role in eye development is a specialized role that evolved later. It should also be noted that Pax6 is binds to the promoter of the major Drosophila rhodopsin gene (Sheng et al. 1997). This regulation of vision by eyeless is another example of ancient roles for Pax6 that likely predate its role in eye specification.

The Eye Specification Cascade

In subsequent years competition for this coveted title of master regulator grew as several other genes were shown to share many of the same functional traits. As with Pax6 other aspiring master regulators were either absolutely required for eye development, directed ectopic eye formation or both (Treisman & Heberlein 1998, Treisman 1999, Kumar 2001, Kumar & Moses 2001c). One by one genes were added to the list and today it includes another Pax6 gene twin of eyeless (toy), two unique Pax genes eye gone (eyg) and twin of eyegone (toe), the Six family members sine oculis (so) and optix, and eyes absent (eya) and dachshund (dac) genes which are the founding members of the Eya and Dach families (Milani 1941, Hunt 1970a, Hunt 1971, Bonini et al. 1993, Cheyette et al. 1994, Mardon et al. 1994, Serikaku & O’Tousa 1994, Czerny et al. 1999, Seimiya & Gehring 2000). Ironically, many of these mutants had been known to delete the compound eyes for decades, nearly 50 years in some cases. The work by the Gehring group on eyeless spurred a renewed interest in these other genes. Soon evolutionary cousins for each of these eye specification genes were found throughout the animal kingdom including mammals and several human medical disorders can be directly attributed to mutations within the orthologs of the fly eye specification genes (Xu et al. 1997, Zimmerman et al. 1997, Leppert et al. 1999, Azuma et al. 2000, Davis et al. 2001, Hanson 2001, Fee et al. 2002, Heanue et al. 2002, Zhu et al. 2002) (tables 1&2).

Table 1.

Eye Specification Network in Drosphila

Eyeless (ey)	paired domain + homeodomain
twin of eyeless (toy)	paired domain + homeodomain
eye gone (eyg)	partial paired domain + homeodomain
twin of eye gone (toe)	partial paired domain + homeodomain
sine oculis (so)	Six domain + homeodomain
optix	Six domain + homeodomain
eyes absent (eya)	Protein tyrosine phosphatase
dachshund (dac)	winged helix-turn-helix
teashirt (tsh)	zinc finger (C2H2 type)
homothorax (hth)	homeodomain
extradenticle (exd)	homeodomain + helix-turn- helix
Notch (N)	signaling pathway
Epidermal Growth Factor Receptor (Egfr)	RTK signaling pathway
Wingless (Wg)	Wnt signaling pathway
Hedgehog (Hh)	Singling pathway
Decapentaplegic (Dpp)	TGFb signaling pathway

Table 2.

Eye Specification Network Mammalian Homologs and Disease

Fly Gene	Mammalian Homolog	Human Retinal Disease
eyeless (ey)	Pax6	Aniridia
twin of eyeless (toy)	Pax6	Aniridia
eye gone (eyg)	Pax	--------
twin of eye gone (toe)	Pax	--------
optix	Six6	bilateral anophthalmia
eyes abset (eya)	Eya 1-4	congential caaracts
dachshund (dac)	Dach 1-2	--------
teashirt (tsh)	Tsh 1-3	--------
homothorax (hth)	Meis	--------
extradenticle (exd)	Pbx1	--------
Notch (N)	Notch 1-4	Alagille Syndrome
EGF Receptor (Egfr)	waved-2	trichomal, cataracts, RPE cell death
Wingless (Wg)	Wnt	--------
Hedgehog (Hh)	Shh, Dhh, Ihh, Twhh	cyclopia, holoprosencephaly
Decapentaplegic (Dpp)	BMP	anophthalmia, lens induction

The eyeless and twin of eyeless genes encode Pax6 class transcription factors containing a paired domain and a homeodomain. In addition to their roles in Drosophila eye and head development both genes are expressed and function within the developing central nervous system, embryonic brain and mushroom bodies (Kurusu et al. 2000, Noveen et al. 2000, Callaerts et al. 2001, Kammermeier et al. 2001, Adachi et al. 2003). Although both Pax6 genes share many structural similarities and function in the same tissues they appear to play distinct roles in development. Such differences extend even to the regulation of the Pax6 genes as twin of eyeless regulates eyeless but not visa versa (Czerny et al. 1999, Kammermeier et al. 2001, Punzo et al. 2002, Punzo et al. 2004).

The other set of Pax proteins that function during Drosophila eye development is encoded by the eyegone and twin of eyegone genes. Like EY and TOY, these proteins contain a homeodomain but unlike the two Pax6 proteins, EYG and TOE contain only a partial paired domain (Jun et al. 1998). These genes are interesting, in part, because they regulate eye development independently of Pax6 (Hunt 1969, Hunt 1970b, Hunt 1971, Jang et al. 2003, Dominguez et al. 2004, Rodrigues & Moses 2004). These findings are significant because they hint that several parallel pathways are required to correctly build an eye. It might also suggest that the evolutionary history of eye development is tied to the sequential recruitment of these parallel acting pathways. The eyegone and twin of eyegone genes also function in other contexts including the developing salivary glands and thorax (Jones et al. 1998, Aldaz et al. 2003).

The sine oculis and optix genes arose through an ancient duplication event and belong to the Six gene superfamily (Seo et al. 1999, Kawakami et al. 2000). A hallmark of Six family members is the Six domain which has been implicated in protein-protein interactions (Kawakami et al. 2000). As Six family members appear unable to activate transcription despite the ability to bind DNA such interactions are crucial in recruiting accessory proteins. One of the most notable examples is an interaction with EYA which is encoded by the eyes absent loci (Pignoni et al. 1997). The formation of this heterodimeric complex appears to be crucial for the transcriptional activation of target genes and the formation of ectopic eyes. Recent reports have also shown that mammalian Six proteins form complexes with GRO co-repressors and this interaction is crucial for correct vertebrate eye specification (Fisher & Caudy 1998, Chen & Courey 2000, Li 2000, Kobayashi et al. 2001, Zhu et al. 2002, Lopez-Rios et al. 2003). In Drosophila a similar SO-GRO interaction has been described in vitro and genetic interactions between groucho and sine oculis suggest that these two genes cooperate to regulate eye development (Silver et al. 2003). Together these results suggest that subsets of SIX proteins affect eye formation by both activating and repressing transcription of target genes through interactions with EYA and GRO family members. The available data on the role that optix plays in Drosophila eye development is less clear as loss-offunction mutants are not yet available. optix certainly plays a positive role in promoting ectopic eye formation (Seimiya & Gehring 2000). However, this does not appear to be through interactions with eyes absent. In addition, overexpression of optix within the developing fly eye inhibits eye development (J. Kumar unpublished results). How optix exactly fits into the eye specification network is still an open question.

The role that eyes absent plays in eye specification has undergone a renewed level of interest in recent years as it has been reported that in addition to its role as a transcriptional co-activator, EYA also functions as a nuclear protein tyrosine phosphatase (Rayapureddi et al. 2003, Tootle et al. 2003). This represents an unexpected level of regulation within the eye specification cascade. The identification of putative substrates will certainly lead to a deeper understanding of how the SO-EYA complex positively influences retinal development. Interestingly, it appears that high levels of EYA protein are required to promote eye formation and prevent the creation of the SO-GRO repressor complex (Silver et al. 2003). As it appears that EYA and SO are not competing for GRO directly the mechanism for how EYA promotes eye specification remains an interesting paradigm.

Lastly, at the end of the eye specification cascade lies dachshund which encodes a nuclear protein that is distantly related to the Ski/Sno family of DNA binding proteins. It should be noted that although dachshund is required for eye development and can redirect non-retinal tissues into an eye fate, its precise role in specification remains elusive. It had been first suggested that DAC physically interacts with EYA and may in fact be a component of a larger SO-EYA-DAC transcriptional complex (Chen et al. 1997). However, several recent reports have disputed those findings (Silver et al. 2003, Tavsanli et al. 2004). And despite its homology to the Ski/Sno co-repressors it is not clear if DAC has the ability to bind DNA. Thus at the present time dachshund remains an important part of the eye specification network even though its particular role in this process is still somewhat enigmatic.

Over the years a wealth of genetic evidence and a smaller yet growing body of molecular and biochemical data have indicated that these eye specification genes do not function as a linear biochemical pathway but rather they exist within a labyrinth of regulatory interactions in both flies and mammals (figure 3). The network is replete with feedback loops exerting positive as well as negative influences on eye formation. For instance genetic and molecular epistasis experiments have shown that while eyeless regulates the expression of the downstream target gene eyes absent, its own expression is also dependent upon this downstream pathway member (Bonini et al. 1997, Halder et al. 1998). Thus these genes reinforce their own expression through positive feedback mechanisms. Another example of the complexities of eye specification is that the expression of optix is not affected in eyeless mutants and optix does not require eyeless for the induction of ectopic eyes suggesting that parallel pathways may exist for specifying the eye (Seimiya & Gehring 2000). This has led to a blurring of the distinction between the Pax6 genes and other pathway members. Instead of being exalted as a master regulator, eyeless has in some corners been demoted to the status of a lowly “network manager”.

Figure 3.

The Eye Specification Regulatory Network. Eye specification in Drosphila is dependent upon a complicatd network of interactions between signal transduction casades and nuclear transcription factors. This regulatory network consists of many positive asnd negative feedback loops. toy = twin of eyeless, ey = eyeless, toe = twin of eyegone, eyg = eyegone, so = sine oculis, dac = dachshund, eya = eyes absent Egfr = Epidermal Growth Factor Receptor, N = Notch, hh = hedgehog, dpp = decapentalegic.

The real position of Pax6 is likely to be somewhere in between that of master and servant. The genetic, molecular and biochemical data collected so far suggest that the eight known eye specification genes function together, certainly within the same tissue but presumably within the same cells, to specify a retinal cell fate. For example, at the level of transcriptional regulation TOY protein directs the expression of eyeless by binding to sequences within an eyespecific enhancer element and sine oculis expression is directly regulated by the binding of both TOY and EY proteins to its own eye specific enhancer element (Czerny et al. 1999, Niimi et al. 1999). At the biochemical level independent protein complexes have been reported for SO-EYA and EYA-DAC heterodimers (Chen et al. 1997, Pignoni et al. 1997). In addition, molecular epistasis experiments have shown that the expression of each member of the eye specification cascade is dependent upon the activity of at least one other member of the pathway (figure 3). Thus no one particular member can be called king.

While it is likely true that there is not a single control switch for eye development it is fairly clear that the Pax6 genes twin of eyeless and eyeless share several features that confers upon them a special status within the eye determination cascade. First, both genes (along with the other Pax gene eye gone) are the first to be expressed within the eye imaginal disc (Quiring et al. 1994, Jones et al. 1998, Czerny et al. 1999, Kumar & Moses 2001b). These three genes are detected in the eye disc during embryogenesis while the other pathway members are not added to the developing disc until the first and second instar larval stages (Bui et al. 2000, Kumar & Moses 2001a, Kumar & Moses 2001b). A plausible model exists in which the Pax6 genes make the initial determination to adopt a retinal fate but that the other pathway members are subsequently required to reinforce and maintain this decision within the eye imaginal disc. Second, both eyeless and twin of eyeless are demonstrably more potent inducers of ectopic eyes than other genes in the pathway, particularly if one considers the range of tissues that can be redirected into an eye fate. It should be noted that not all tissues are directed into an eye fate by the expression of the Pax6 genes. In fact even within transformable tissues ectopic eye formation can only occur in certain regions of the epithelia (Chen et al. 1999). What makes certain tissues or regions of tissues refractory to the activity of the eye determination genes? One explanation is that these tissues might express negative or inhibitory factors that repress eye development. Another explanation is that the eye specification genes require another set of eye promoting factors that are not normally expressed within other non-retinal tissues. The truth likely lies in between these two possibilities and in fact there is good evidence for both mechanisms.

Inhibitory Mechanisms Delimit the Eye Field

The search for inhibitory molecules and mechanisms has, ironically, been focused on the developing eye imaginal disc itself. Efforts to identify negative regulators have yielded the wingless (wg), teashirt (tsh), homothorax (hth) and extradenticle (exd) genes (Gonzalez-Crespo & Morata 1995, Treisman & Rubin 1995, Pai et al. 1998, Pan & Rubin 1998, Pichaud & Casares 2000b, Bessa et al. 2002, Singh et al. 2002). Collectively these genes form an inhibitory signaling pathway and are thought to delimit eye development in both temporal and spatial manners (figure 4). During development TSH protein is thought to form a heterodimeric complex with ARM (the product of armadillo, a downstream component of the Wingless cascade) (Singh et al. 2002). This complex regulates the transcription of homothorax. EXD protein, which is cytoplasmic in the absence of HTH, binds to HTH and is translocated to the nucleus (Gonzales-Crespo et al. 1998, Pai et al. 1998, Pichaud & Casares 2000a). The HTH-EXD complex is then thought to form a quaternary complex with TSH and EY (HTH-EXD-TSH-EY) (Bessa et al. 2002). It still remains to be determined if this quaternary complex directly represses transcription of downstream genes or it inhibits eye development through the sequestration of EY protein. In any event, these four proteins are co-expressed in a zone anterior to the furrow and it is thought that they serve to define the anterior boundary of the eye. Whether this inhibitory mechanism is used in non-retinal tissues remains an open question. If similar mechanisms are used, are the same proteins used reiteratively in other imaginal discs or there are tissue specific inhibitors that bind to EY protein? While there is no evidence to support one model or the other, it should be noted that EY protein has the ability to bind other homeodomain-containing proteins; thus different proteins could be used to sequester EY in non-retinal tissues (Plaza et al. 2001). Furthermore, the homoedomain of EY has been shown to be a non-essential component during normal and ectopic eye development (Punzo et al. 2001).

Figure 4.

Model for Inhibition of Eye Formation. The eye specification genes are expressed in dynamic patterns within several non-retinal tissuss. Mechanisma must exist to prevent these tissues from being redirected into an eye fate. Also in normal development inhibitory mechanisms must be put in place to limit he size of the eye field. This schematic depicts one potential mechanism for inhibiting precocious retinal development.

The use of inhibitory mechansims to limit retinal development to the eye imaginal disc is unlikely to be the entire story as no loss-of-function mutations in either homothorax, extradenticle or teashirt have been shown to induce ectopic eye development in non-retinal tissues. Thus positive influences are certain to play an essential role in eye formation (see below). It should also be noted that the role of homothorax and teashirt as inhibitory molecules is very much a simplification. Their actual roles in development appear to be much more complicated as (1) ectopic eyes can be induced within the wing only in areas that also express homothorax and (2) overexpression of teashirt can surprisingly redirect antennal tissue into an eye fate (Pan & Rubin 1998, Bessa et al. 2002). Thus it appears that the developmental context within which these genes function is a crucial factor in determining their role in eye development.

Patterning Pathways Function During Eye Development

In addition to inhibitory influences such as these there is clear evidence that normal as well as ectopic eye development is dependent upon the activity of the patterning genes hedgehog (hh) and decapentaplegic (dpp). The first hint came from the observation that only particular regions within imaginal discs have the capacity to be transformed into retinal tissue (Halder et al. 1995a). Careful examinations of gene expression profiles of these domains suggested that the aforementioned signaling molecules and the pathways that they head are crucial requisite components of eye deter-mination (Chen et al. 1999, Kango-Singh et al. 2003). Experimental manipulations that expanded the expression patterns of either hedgehog or decapentaplegic also expanded the areas of tissue competent to support eye formation (Chen et al. 1999, Kango-Singh et al. 2003). It should be noted that although these genes are necessary for the formation of ectopic eyes they, on their own, are insufficient to support ectopic retinal development. The entire expression domain of either patterning gene is not transformable so there are likely to be additional requirements for ectopic eye development that are yet to be identified.

Nonetheless the hedgehog and decapentaplegic genes are also expressed early in the developing eye imaginal disc and loss-of-function mutations inhibit eye development (Heberlein & Moses 1995). Previous studies of hedgehog and decapentaplegic loss-of-function mutants attributed these phenotypes to failures in the initiation of pattern formation (Heberlein et al. 1993, Ma et al. 1993, Heberlein & Moses 1995, Heberlein et al. 1995, Chanut & Heberlein 1997, Dominguez & Hafen 1997, Pignoni & Zipursky 1997, Borod & Heberlein 1998). However, recent reports suggest that in addition to moving the morphogenetic furrow across the eye field, these patterning pathways play a role in eye specification by regulating the expression of eyes absent, a member of the eye determination cascade (Curtiss & Mlodzik 2000, Pappu et al. 2003). eyes absent is an important focal point for regulation within the cascade. The expression of eyes absent is not only regulated by other members of the eye specification cascade and two patterning pathways but the encoded protein itself is regulated through post-translational modifications by receptor tyrosine kinase (RTK) signaling via the cytoplasmic factor mitogen activated protein kinase (MAPK) (Hsiao et al. 2001). Experimental manipulations of the MAPK phosphorylation sites within the EYA protein itself results in a decrease in the ability to rescue eya loss-of-function mutants and to generate ectopic eyes (Hsiao et al. 2001). These results suggest that RTK signaling may either function early during eye determination or later during the specification of photoreceptor cell fates. Supporting these data are the observations that mutations within the mammalian Epidermal Growth Factor Receptor (EGFR) lead to a wide range of retinal defects (Luetteke et al. 1994).

Signaling Pathways Regulate Tissue Identity

Recent evidence has also led to the suggestion that some signaling cascades play a much more central role in imaginal disc specification than previously thought. In both the eye and the wing imaginal disc hyperactivation of the EGFR pathway leads to homeotic transformations in which the eyes and notum are converted into antennae and wings, respectively (Baonza et al. 2000, Kumar & Moses 2001a) (figure 5). Under these circumstances expression of eye genes is abolished in the transformed tissue and transcription of antennal genes is activated (Kumar & Moses, 2001a). These results suggest that in the eye the EGFR pathway functions early to delimit the retinal field to the eye disc proper. This view is supported by a recent demonstration that eyeless positively regulates the transcription of kekkon, a gene that negatively regulates EGFR signaling by binding to the receptor itself (Michaut et al. 2003). One can imagine a model in which EGFR signaling prevents inappropriate eye development in regions that will eventually give rise to the antenna. In the eye disc one role of eyeless may be to tune down EGFR signaling thus allowing for eye development to proceed. Taken together it appears that EGFR functions early to delimit the eye field and again later through eyes absent to promote photoreceptor cell fate specification.

Figure 5.

The Notch and EGF Receptor Signaling Pathways Control Eye and Antennal Tissue Identity. (A,B) Scanning electron micrographs. (C,D) Confocal micrographs. Red = F-actin, green = Distalless, a gene that contols antennal specification. (E,F) Schematic diagrams of wild type and eye-antenna transformations. (A,C) wild type (B,D) Hyperactivation of Egfr leads to the homeotic transformation of the eye into an antenna. Downregulation of Notch signaling leads to the identical phenotype. Note in D that the transformed antenna; expresses DII in the normal antennal pattern. Anterior is to the right.

Decreases in activity of the Notch (N) signaling pathway also lead to eye-antennal transformation as well, thus implicating Notch signaling in promoting an eye fate (Kurata et al. 2000, Kumar & Moses 2001a, Singh & Choi 2003). Any role for Notch in early eye decisions may be conserved throughout all seeing animals, as it is known that mammalian eye development is sensitive to decreases in activity of the pathway (Voas & Rebay 2004). This is exemplified by the observation that mutations within the human Notch2 and/or its ligand JAG1 lead to Alagille syndrome that is characterized by, among other clinical manifestations, defects in anterior chamber development (Lewis 1992, Lindsell et al. 1996, Hanson 2001, Kumar & Moses 2001c, McCright et al. 2001, McCright et al. 2002).

The observation that the EGFR and Notch pathways regulate tissue-wide fate decisions suggests that these pathways lie genetically and molecularly upstream of the eye regulatory network. The homeotic phenotypes associated with the manipulation of signaling levels more closely fit the definition of a master regulator as first put down by Ed Lewis (Lewis, 1992). At the very least these signaling networks appear to function early in retinal specification and may serve to coordinate the activity of the eye specification network. It is a less well-known observation that within the eye imaginal disc the expression profiles of the eight eye specification genes are not completely synchronous in either time or space. During embryogenesis only eyeless, twin of eyeless and eyegone transcripts can be detected within the eye imaginal disc (Quiring et al. 1994, Jones et al. 1998, Czerny et al. 1999, Kumar & Moses 2001a). Transcripts of the other genes accumulate later in the developing eye and it has been recently shown that the expression patterns of all eight genes are not focused within the eye disc until the end of the second larval instar (Kumar & Moses 2001a). At this stage the Notch receptor is also highly enriched within the eye as compared to the antennal disc. This period in the life of the eye disc is critical since manipulation of either EGFR or Notch signaling can redirect the eye into an antennal fate (Kumar & Moses 2001a). It seems likely that the activity of the eye specification network is coordinated through the efforts of patterning and signaling pathways.

During the third larval instar stage the morphogenetic furrow initiates at the posterior edge of the disc and makes its way to the anterior border that separates the eye from the antenna (Ready et al. 1976, Wolff & Ready 1991, Heberlein & Moses 1995). Cells that lie ahead of this advancing patterning wave express all eight eye specification genes (figure 6). Presumably these cells are kept in a ‘pre-eye’ fate until the furrow sweeps over them and forces them to differentiate into the specific cell types that will comprise the retina. Despite the genetic, molecular and biochemical evidence suggesting that the eye network functions as a unit there are several differences in expression profiles. Expression of eyeless, twin of eyeless, optix, eye gone and twin of eye gone expires abruptly at the morphogenetic furrow (figure 6). On the other hand DAC protein remains for several columns behind the furrow but then also fades away (Mardon et al. 1994) (figure 6). In fact only two of the eight proteins (SO and EYA) are present throughout the entire developing eye (Bonini et al. 1993, Cheyette et al. 1994, Serikaku & O’Tousa, 1994) (figure 6). It remains unclear what role if any these three proteins (SO, EYA and DAC) are playing in ommatidial assembly. But along with existing biochemical data it does support the idea that SO and EYA are obligate partners at least within the developing eye (Pignoni et al. 1997). There is also in vitro evidence to support a biochemical interaction between EYA and DAC (Chen et al. 1997). If such a complex forms it is likely to do so only ahead of the furrow. Currently, we have a lot left to learn about how the extant members of this regulatory cascade function together during early retinal determination. There is also quite a bit waiting to be discovered about the role that some of these genes play in cell fate decisions.

Figure 6.

Expression Pattern of the Eye Spectification Genes. (A-C) Confocal micrographs. (D-F) Schematic drawing. (A,D) EY protein is found in cells ahead of the morphogenetic furrow. The expression of twin of eyeless, optix, eyegone and twin of eye gone mimics that of eyeless. (B,E) SO protein is present in all cells within the developing eye disc. The EYA protein is also present in these same cells. (C,F) DAC protein is expressed ahead of the morphogenetic furrow and in several columns of cells posterior to the furrow. Expression in the most posterior regions of the disc is eliminated. MF = morphogenetic furrow. Anterior is to the right.

Eye Specification Genes Function in Other Tissues

An even less publicized aspect of the eye specification network is that complete loss-of function mutations have been shown to affect the development of multiple tissues, not just the eye (Bonini et al. 1993, Cheyette et al. 1994, Mardon et al. 1994, Quiring et al. 1994, Serikaku & O’Tousa 1994, Bonini et al. 1998, Jones et al. 1998). For example, sine oculis and eyes absent have been implicated in both spermatocyte development and polar cell fate during oogenesis (Bai & Montell 2002, Fabrizio et al. 2003). Additionally, sine oculis functions during optic lobe specification and dachshund is a well known player during leg development (Cheyette et al. 1994, Serikaku & O’Tousa 1994, Goto et al. 1999, Inoue et al. 2002, Mardon et al. 1994). The spectacular no-eye phenotypes can be attributed to spontaneously occurring mutations within regulatory regions that control expression just within the developing eye (Quiring et al. 1994, Niimi et al. 1999, Bui et al. 2000). The efforts of many laboratories have collectively demonstrated dynamic expression patterns for each of the so-called ‘eye specification’ genes during many phases of development. For example, during embryogenesis eyeless and twin of eyeless transcripts are present in the embryonic brain and central nervous system in addition to the eye imaginal disc (Kammermeier et al. 2001). On the other hand eyes absent, sine oculis and dachshund are not even expressed in the embryonic eye disc but rather are present within other tissues including the visual primordium, optic lobes, brain and central nervous system (Cheyette et al. 1994, Mardon et al. 1994, Serikaku & O’Tousa 1994, Bonini et al. 1998, Leiserson et al. 1998, Kumar & Moses 2001b) (figure 7). Even more interesting is that the spatial and temporal expression patterns of these genes are often non-overlapping. For instance, despite the superposed expression patterns of sine oculis and eyes absent within the eye imaginal disc and proposed direct biochemical interactions and synergy at the genetic level, SO and EYA proteins can often be found in distinct populations of cells within the developing embryo (Kumar & Moses, 2001b) (Figure 7). The obvious implication is that either these genes have independent roles during development or they have binding partners that are yet to be identified. Each of the other genes are similarly expressed in a wide range of tissues and are suggested to function independent of the other eye specification genes.

Figure 7. Expression Patterns of the Eye Specification Genes During Embryogenesis.

A, Schematic drawing representing the expression pattern of eyeless at stage 13 of embryogenesis. Transcripts of twin of eyeless and eyegone are also present in the eye imaginal disc at this stage. It should be noted that the other eye specification genes are not expressed in the eye disc until the later instar stages of development; B, Schematic drawing representing the expression pattern of sine oculis and dachshund in the head regiosn of the embryo at stage 9. Note that although dachshund is downstream of sine oculis in the developing eye their expression patterns here are non-overlapping; C, Schematic drawing of the expression pattern of eyes absent and sine oculis. Genetic and molecular evidence suggests that during eye development the encoded proteins function as obligate partners. However in the developing embryonic head there are regions in which the two proteins are not co-expressed; D, Schematic drawing of the expression pattern of eyes absent and dachshund. Note that in the developing eye these two proteins are thought to regulate each other and form a biochemical complex. But in the embryonic head their expression patterns largely overlap. Anterior is to the left.

In all fairness it should be recognized that while the vertebrate eye has been a fruitful model system for elucidating the role that the Pax-Six-Eya-Dach network plays in mammals, this network has also been well studied in other developmental contexts particularly during myogenesis (Spitz et al. 1998, Heanue et al. 1999, Ridgeway & Skerjanc 2001, Fougerousse et al. 2002, Ikeda et al. 2002, Laclef et al. 2003a, Li et al. 2003, Grifone et al. 2004). In addition the ‘eye specification’ network has also been extensively studied in mouse models and human patients with Branchio-Oto-Renal syndrome which is characterized by craniofacial defects as well as auditory and kidney abnormalities (Heanue et al. 1999, Xu et al. 1999, Buller et al. 2001, Davies & Fisher 2002, Fougerousse et al. 2002, Pfister et al. 2002, Xu et al. 2002, Laclef et al. 2003b, Xu et al. 2003, Zheng et al. 2003, Brodbeck & Englert 2004, Ozaki et al. 2004, Ruf et al. 2004). The reiterative use of this regulatory network in such diverse developmental contexts is an interesting evolutionary paradigm. It is likely that the entire cascade was co-opted en masse and then modified to fit individual developmental needs. It would be interesting to determine which of role of the ’eye specification’ network is really ancestoral.

Concluding Remarks

It is quite clear from the evidence to date that the eye specification cascade (as we know it) is a conserved unit that functions in all seeing animals including mammals. Astonishingly, several human retinal disorders can be directly attributed to mutations within the human versions of the eye determination genes and the signal transduction cascades that regulate them. This enormous level of conservation over such orders of evolutionary time makes it improbable that the eye evolved on several different occasions. Rather it is likely that the eye evolved once early in history and different evolutionary and ecological pressures have shaped the structure of the eye to fit different environmental and optical constraints, While a lot has been learned over the past decade about how the early decisions in eye development are made, there are still many questions left unanswered. First, have we identified all the members of the eye regulatory network? Probably not; analysis of gene expression using microarrays have shown that literally hundreds of genes are regulated by eyeless and the other extant eye determination genes. How many of these will make it to the level of master regulators is left to be determined.

Second, how do the members of the network interact with each other? To date we have only identified a handful of molecular and biochemical interactions. And even these data cannot explain many of the in vivo observations regarding the expression patterns and phenotypes. A much larger and concerted effort will have to be made to fully understand how these eight known genes interact with each other not to mention the potentially hundreds of new candidates.

Third, how does this network function to promote eye development? We still have a long way to go before we really understand the sequence of events that initially determines the fate of the eye imaginal disc. The cells that give rise to the eye disc are known to be distinct well before the expression of twin of eyeless and eyeless. Thus it remains to be seen if we will ever find a true ‘master regulatory switch’ for the eye. Such a gene might be expected to expressed only within the eye imaginal disc and the onset of its transcription would precede that of all the know eye determination genes.

And finally how has nature produced the magnificent diversity in eye structure that we see today? We know of many genes whose orthologs have been found throughout the animal kingdom but it may be more interesting to look at the genes that have not been conserved during the evolution of the eye. The maturing of the genomics and bioinformatics era certainly will make the discovery of such genes easier. In the end it will be a significant step forward if we can understand how the eye first evolved and how organisms have co-opted genes and pathways to produce such distinct optical structures as the simple eye of man and the compound eye of the fruit fly.