The great diversity of body shapes and the variety of forms and disposition of its constituent organs across phyla contrast with modern genetic and molecular data suggesting that the molecular mechanisms underlying the development of organs are highly conserved between distant animal groups. As form arises from a combination of cell identity determination and patterned cell proliferation, growth, and death, the coordination of these processes lies at the heart of sculpting animal form. Variations of this coordination are the material upon which evolution shapes new species.
The developing compound eye of the fruit fly Drosophila melanogaster has been extensively used for the genetic and molecular studies of eye identity and for the analysis, with single-cell resolution, of how the retina is patterned (for recent reviews, see Treisman and Heberlein, 1998; Gehring, 2002; Frankfort and Mardon, 2002; Voas and Rebay, 2004). It has been only recently that we have started to learn about the mechanisms responsible for the singularization of distinct organs (eye, antenna, and head capsule, Fig. 1) from the common eye–antennal primordium. Still, much less is known about how these distinct organs control their growth, bearing constant, species-specific, size relations with each other. This review sets out to examine Drosophila eye formation to the coordination of organ growth.
DROSOPHILA EYE–ANTENNAL DISC: EMBRYONIC ORIGIN AND LARVAL DEVELOPMENT
The adult fruit fly is built-up piecemeal, its different organs deriving from flat epithelial sacs called imaginal discs. Although three pairs of imaginal discs—clypeolabral, labial, and eye–antennal—contribute to the adult head, most of the head capsule and the major cephalic sensory organs (eyes, antennae, maxillary palps, and ocelli) derive from the eye–antennal discs (see Fig. 1A; reviewed in Haynie and Bryant, 1986; Jurgens and Hartenstein, 1993). This review deals with the development of the eye–antennal disc, which is taken as a representative example of the development of multiple organs from a composite rudiment and because its development has been particularly well studied.
The eye–antennal discs originate from different embryonic segments (antennal, maxillary, and possibly mandibular, intercalary, and labial segments) plus the nonsegmented acron. By the end of embryogenesis, the discs invaginate from the dorsal pouch of the embryo as two small flat epithelial sacs, of approximately 70 cells each (Jurgens and Hartenstein, 1993). During the three larval stages, or instars, the disc grows mostly by cell proliferation. The different organ primordia will be specified in only one of the two layers of the disc—the main epithelium—whereas the other layer, called peripodial epithelium, becomes a squamous epithelium (Fig. 2). The peripodial epithelium will participate in the eversion and fusion of the discs during metamorphosis (Fristrom and Fristrom, 1975; Pastor-Pareja et al., 2004).
The eye–antennal disc cells retain considerable developmental plasticity until late stages of development. Clones of cells, induced as late as the second instar, can still contribute to patches of tissue spanning several organs (antenna, eye, and head capsule), indicating that, at the time of clone induction, the disc cells were not committed (Morata and Lawrence, 1979; Baker, 1978).
Recent analyses of the expression and function of genes involved in the development of the different organs of the eye–antennal disc are coalescing into a model of progressive restriction of “eye-forming competence” and gradual subdivision of the disc primordium into distinct organ domains. This subdivision is led by the interplay between several transcription factors and signaling molecules. Below, we dissect this interplay into (1) the specification of the distinct organ primordia, focusing on the eye primordium; (2) the signals that drive proliferation of the disc and its relation with eye specification; (3) the signaling processes that couple cell proliferation cessation to the gene expression changes that lead to cell differentiation; (4) the dual role of wingless controlling fate specification and proliferation; and (5) possible implications of these mechanisms during the evolution of organ proportion.
DYNAMIC GENE EXPRESSION DOMAINS DEFINE PROGRESSIVELY DISTINCT ORGAN PRIMORDIA WITHIN THE EYE–ANTENNAL DISC
The expression of selector genes regulating fate of the distinct organ primordia within the eye–antennal discs has been studied intensely. This section summarizes the dynamic expression of these selector genes in both time and space.
Selector Gene Expression at First Instar Eye–Antennal Disc
First instar discs express uniformly the Pax6 paralogues eyeless (ey) and twin of eyeless (toy), which act as “eye selector genes” (reviewed in Gehring, 2002). ey is required for identity and eye development, at least in part, by preventing apoptosis (Quiring et al., 1994; Halder et al., 1995, 1998; Czerny et al., 1999), and each of the two Pax6 proteins can induce eye development when ectopically expressed elsewhere in the fly body, even in the absence of the other (Gehring, 2002; Punzo et al., 2004). toy acts upstream of ey by inducing ey's transcription, so their expression patterns are very similar (Czerny et al., 1999). The expression of these genes is maintained on in the eye primordium cells until these commence to differentiate (Fig. 2). Late ey expression is associated with the terminal differentiation of the photoreceptors and needed for rhodopsin genes expression (Halder et al., 1998), but this aspect will not be discussed in this review. The eye selector role of Pax6 is almost universal across the animal kingdom, with the sole exception, so far, during planarian eye regeneration (Pineda et al., 2002). In addition to ey and toy, two additional Pax6-related genes of the Pax6(5a) type, eyegone (eyg) and twin of eyegone (toe), are also expressed in the embryonic eye–antennal disc (Jun et al., 1998; Aldaz et al., 2003; Jang et al., 2003). Their expression is turned off during first instar, and reinitiated during late second larval instar in a restricted dorsoventral domain (Dominguez et al., 2004; Chao et al., 2004). In addition to toy and ey, the TALE-class homeodomain transcription factor homothorax (hth; Rieckhof et al., 1997; Pai et al., 1998) is also expressed in all cells of the first instar eye–antennal disc (see Fig. 2A).
Recently, it has been shown that Optix, the Drosophila homologue of mammalian Six3, is expressed in the eye–antennal disc with a pattern reminiscent to that of ey. Optix is also capable of inducing eye development when ectopically expressed (Seimiya and Gehring, 2000), although this ability is restricted to the eye–antennal disc. The expression of Optix and its eye-inducing ability are both ey-independent and, hence, suggests that Optix leads a parallel eye-specification mechanism (Seimiya and Gehring, 2000). The regulatory and functional relations between Optix and toy have not yet been tested, so it might be that Optix is under the control of toy or requires toy for eye induction. Therefore, this parallel mechanism is still not fully proven and awaits the isolation of toy and Optix null mutations. The mammalian homologues of Optix play key roles during eye specification and growth in vertebrates (Loosli et al., 1999; Zuber et al., 1999; Carl et al., 2002; Lagutin et al., 2003).
Gene Expression Changes at Second Instar Eye–Antennal Disc
During second instar, a series of gene expression changes result in the first molecular definition of the antennal and eye fields (Fig. 2C). First, the expression of ey/toy retracts to the posterior two thirds of the disc, while the homeodomain-encoding gene cut is turned on in the anterior third of the disc (Fig. 2C). The ey/toy and cut complementary domains mark out the territories of the future eye and antenna, respectively (Kenyon et al., 2003; herein, referred to as the eye and antennal discs). The activation of another homeodomain-encoding gene, Distalles (Dll) within the cut domain leads to the coexpression of Dll and hth (Fig. 2C), which together specify the antennal fate (Casares and Mann, 1998; Dong et al., 2000). The complementary domains of cut and ey/toy expression suggest that these transcription factors mutually repress each other to define those domains. In support of this idea, Punzo and coworkers (2004) recently have found that, in the absence of ey and if cell death is prevented, Dll gene is turned on in the eye disc and an ectopic antenna replaces the eye. Therefore, ey expression may antagonize antennal fate in the eye–antennal disc.
Onset of “Early Retinal Genes” Expression During Late Second Instar
At the time of the antenna–eye partition, the expression of the nuclear factor eyes absent (eya; Bonini et al., 1993) starts at the posterior region of the eye disc (Kenyon et al., 2003). Eya is the first of a group of nuclear factors expressed within the presumptive eye field, collectively known as “early retinal genes” (reviewed in Pichaud et al., 2001). This group includes the Six homologue, sine oculis (so; Cheyette et al., 1994; Serikaku and O'Tousa, 1994) and the nuclear factor dachshund (dac; Mardon et al., 1994). The so/Six and dac expression follows that of eya, still during second instar (Kenyon et al., 2003). Expression of both eya and so are activated independently by ey/Pax6 (Halder et al., 1998), and at least in the case of so/Six, this activation is direct (Punzo et al., 2002).
Of interest, although these genes are expressed elsewhere in the fly embryo, they are all first coexpressed in the eye disc during second instar. Because all these genes are required for the initiation of retinal differentiation, it has been proposed that it is this coexpression in second instar that locks-in the eye competence in the ey-expressing cells (Kumar and Moses, 2001; Pichaud et al., 2001). In support of this model, the protein products of the early retinal genes appear to work together in a molecular network: (1) they are expressed in largely overlapping domains and found in the cell's nucleus, (2) they are able to cross-activate each other's transcription, (3) their coexpression triggers ectopic eye formation in other discs, and (4) their protein products physically interact (Pignoni et al., 1997; Bonini et al., 1997; Chen et al., 1997; reviewed in Desplan, 1997).
The molecular mechanisms by which this retinal gene network may operate are beginning to be unveiled: The so/Six family gene so encodes a homeodomain transcription factor that provides a DNA-binding domain (Olivier et al., 1995); dac relates to the Ski/Sno transcriptional corepressors (Hammond et al., 1998) and eya possesses a haloacid halogenase (HAD) domain, which phosphatase activity is required for eya's function in Drosophila (Rayapureddi et al., 2003; Tootle et al., 2003) and mammals (Li et al., 2003). At least in mice, eya phosphatase activity is required to turn Dach1, bound to Six1, from a co-repressor to a coactivator (Li et al., 2003). A similar molecular mechanism might operate in Drosophila eye.
Gene Expression Changes at the Peripodial Epithelium
The peripodial layer also undergoes significant changes in transcription factor gene expression, such as the turning off of ey expression (J. Bessa and F. Casares, unpublished observations; Hallsson et al., 2004), although their regulation and functional meaning, if any, are still unknown. Hallsson et al. (2004) have shown that the peripodial epithelium of the eye disc shows molecular hallmarks of the retinal pigmented epithelium (RPE) of the vertebrate eye. The RPE is an indispensable and highly specialized epithelium of the vertebrate eye. Several transcription factors, including Otx2, Mitf, and Pax6, are required for the onset of RPE specification during development (for a review, see Martinez-Morales et al., 2004). Hallsson et al. (2004) have identified a Drosophila homologue of Mitf (from Microphthalmia transcription factor). Similar to their vertebrate counterparts, dMitf and ey/Pax6 are initially coexpressed in the eye disc, but when the morphological distinction between peripodial and main disc epithelium are well apparent, ey/Pax6 is turned off from the peripodial membrane and dMitf becomes restricted to a region of the peripodial membrane that overlays the region where retinal differentiation is taking place (see below).
In vertebrates, interactions between the RPE and the neural epithelium are important for the patterning and growth of the two layers. Studies from Cho and coworkers (2000) and Gibson and Schubiger (2000) have shown the peripodial epithelium is a source of signaling molecules that contribute to proper patterning and growth of the main disc epithelium. Together these observations have led to the hypothesis of a common origin of the peripodial epithelium and the vertebrate RPE cells during evolution (Hallson et al., 2004).
SIGNALS THAT CONTROL THE GROWTH OF THE EYE DISC
Proliferative growth is continuous from late first instar to late second/early third instar discs. A key proliferative signal is provided by the Notch pathway. Although the Notch receptor is expressed almost ubiquitously in the eye–antennal disc, it is activated by its ligands (Delta and Serrate) only along the dorsoventral compartment boundary of the eye disc. This boundary acts as a signaling center required for the growth of the disc (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998). The steps leading to this localized Notch activation have been deciphered in recent years: The dorsally restricted expression of the signaling molecules wingless (wg) and hedgehog (hh) activate the expression of the Iroquois genes in the dorsal half of the second instar eye disc (Cavodeassi et al., 1999). The three Iroquois (ara, caup, and mirr) proteins act redundantly to repress the expression of the fringe gene, which, therefore, becomes restricted to the ventral half of the primordium (Dominguez and de Celis, 1998; Cho and Choi, 1998; Cavodeassi et al., 1999; Yang et al., 1999). fringe encodes a glycosyltransferase that modulates Notch's interaction with its ligands Delta and Serrate (reviewed in Haines and Irvine, 2003).
The asymmetric expression of fringe is perhaps the most important step in this genetic cascade, because it generates a border of fringe-expressing and fringe-nonexpressing cells—it is at this border that Notch receptor becomes activated by its ligands, but this local activation drives global eye disc growth (Cho and Choi, 1998; Dominguez and de Celis, 1998; Papayannopoulos et al., 1998). Thus, preventing the formation of such fringe+/fringe− border halts disc growth (while creating ectopic borders triggers tissue overgrowth and even the formation of ectopic eyelets; Cavodeassi et al., 1999). A similar dorsal/ventral subdivision, based on the restricted expression of fringe and a border of Notch activation is also established during the development of the wing disc, although the precise activating mechanisms differ (reviewed in Irvine, 1999; Wu and Rao, 1999).
It has been shown recently that the role of Notch in stimulating eye growth is mediated by eyg (Chao et al., 2004; Dominguez et al., 2004; Rodriguez and Moses, 2004). Thus, eyg is expressed in a wedge within the eye primordium straddling the dorsal/ventral boundary from second instar; this expression is under the control of Notch and required downstream of it for eye growth (Chao et al., 2004; Dominguez et al., 2004). Nevertheless, eyg mutant cells of late third instar discs still express eye-determination genes, such as eya, so, and dac; therefore, eye specification can take place in the absence of eyg (Dominguez et al., 2004), although the discs do not grow beyond a second instar size.
The paired DNA-binding domain (PD) of Eyg protein differs from that of canonical Pax proteins, but it structurally resembles the vertebrate-specific alternatively splice variant called Pax6(5a). In eyg, the PD is truncated; therefore, eyg protein binds DNA through its PD C-terminal half. Similarly, the DNA binding ability of the N-terminal PD of the vertebrate-specific Pax6(5a) splice variant is blocked due to the inclusion of 14 extra amino acids (encoded by exon 5a) within this domain. Thus, like eyg, the Pax6(5a) proteins bind DNA through the C-terminal part of the PD (reviewed in Hanson, 2001). Furthermore, and consistent with eyg role in growth control, expression of a cDNA of the human PAX6(5a) isoform triggers dramatic overgrowths in Drosophila imaginal discs (Dominguez et al., 2004). Conversely, expression of ey or its mammalian counterpart PAX6 induces ectopic eye differentiation with little extra growth (Halder et al., 1995; Dominguez et al., 2004). These results, together with the independent regulation of ey and eyg (Jang et al., 2003; Dominguez et al., 2004), indicate that eye determination and eye proliferation rely on largely separate pathways. One surprising result, though, is that the expression of eyg just along the dorsoventral boundary is sufficient to rescue the proliferation of eyg mutant discs (Dominguez et al., 2004). Because eyg is a nuclear protein, this observation suggests that some other factor/s must lie downstream of eyg to relay the dorsal/ventral organizer proliferative signal (Dominguez et al., 2004).
Chao and coworkers (2004) have identified unpaired (upd), a ligand of the JAK/STAT pathway (Hombria and Brown, 2002), as a target of eyg and acting as a long-reaching signal. Although it seems clear that upd is a key player in the Notch-eyg control of eye disc growth, that ectopic expression of eyg and Notch can induce local proliferation in a upd-independent manner suggest a more complex relationship than a linear pathway from Notch to eyg to upd (Chao et al., 2004).
upd is transcribed in a small patch of cells at the point where the Notch activation border (the dorsal/ventral boundary) intersects the posterior disc margin (Bach et al., 2003; Tsai and Sun, 2004; F. Gutierrez-Aviñon and M. Dominguez, unpublished observations). This is the exact point where retinal differentiation starts in early third instar discs, the so-called “firing point”, and a point of integration of “dorsal/ventral” Notch signals and “anterior/posterior” hh signals (Fig. 3B). What is the role of upd at the “firing point”? This answer is currently unknown, but as we discuss below, a primary role of upd signaling in the eye is the induction of cell proliferation. Therefore, it is reasonable to suggest that the “firing point” is a critical region, not only for the timing of the initiation of retinal differentiation, but also for the coordination of eye growth. It is interesting to note that, although the eyes overgrow dramatically when Notch signaling is overactivated in the eye disc, these overgrowths are not isometric; rather, the eyes grow enlarged mostly along the dorsoventral axis, revealing some internal coordinates of preferred proliferation. What these coordinates represent in molecular and cellular terms is still unknown. wg also plays an essential role in controlling cell proliferation, but this role is intertwined with its role in cell fate allocation within the eye disc and will be discussed in the following sections.
WG-DPP ANTAGONISM AND THE CONTROL OF THE ONSET OF EARLY RETINAL GENES
Although the eye selectors ey and toy and the signaling molecules wg and dpp are already expressed in first instar discs, it is only during second instar that the eye primordium is defined within the disc by the coexpression of early retinal genes (Kumar and Moses, 2001; Kenyon et al., 2003). What causes this delayed onset of early eye genes? An explanation has been put forward recently, based on the antagonism between wg and dpp, and limitations imposed by the size of the early disc (Kenyon et al., 2003). wg (Cadigan an Nusse, 1996) and dpp(a BMP-4 family member; Heberlein et al., 1993) are respectively expressed, already in first instar discs, in anterior and posterior nearly complementary domains (Cho et al., 2000). Because both genes encode secreted molecules, and taking into account the small size of the disc at this stage, it is expected that most cells receive both signals (see Fig. 3). This finding is true at least for wg (Royet and Finkelstein 1998; Cavodeassi et al., 1999; Pichaud and Casares, 2000).
The activation of early retinal genes requires dpp signaling (Wiesdorff et al., 1996; Chanut and Heberlein, 1997; Chen et al., 1999; Curtiss and Mlodzik, 2000; Kenyon et al., 2003), whereas high levels of wg block this activation (Hazelett et al., 1998). Therefore, according to the model, during early second instar, when most or all cells receive high levels of wg, eye specification cannot take place. Kenyon and coworkers (2003) suggest that it is the growth of the disc induced by Notch pathway would be responsible for pulling apart the wg and dpp expression domains (Fig. 3A); this way, posterior cells will fall out of the wg range of action. The two immediate outcomes of this low wg and high dpp signaling in the posterior region are (1) that hth, an eye repressor under wg control, is cleared, and (2) that expression of early retinal gene (such as so/Six and eya) expression starts.
Although this model accounts for most reported data, there are some findings that do not conform to this model. Thus, in eyg mutants, the disc is very small and wg gene is still expressed (Dominguez et al., 2004). In the tiny eyg mutant discs, the early retinal genes are normally turned on (Dominguez et al., 2004; Fig. 3B). In these eyg mutant discs, however, retinal differentiation does not start. The latter is not a direct consequence of the loss of eyg activity, because in clones of cells mutant for a null eyg allele, retinal differentiation proceeds normally (Dominguez et al., 2004). Thus, despite the lack of growth, the eye fate is still specified. It is the process of retinal differentiation that seems abolished.
Another complication to the model is the observation by Hazellet and coworkers (and M. Dominguez, unpublished results, Fig. 3C,D) that the coactivation of wg and Notch pathways does not result in a blockage of eye fate. Instead, these flies show a dramatic overgrowth of the eye territory (Fig. 3D). In Figure 3, we propose a variation to the model by Kenyon and coworkers (2003) that bring all these observations together. First, the growth of the eye disc enables anterior “head” (represented by the domain of hth, in blue) and posterior “eye” (represented by the domain of eya, in red) regions to form in response to wg and dpp signaling, respectively. As the disc grows, the separation of the opposing domains of hth and eya activities act as a “trigger” to promote retinal differentiation initiation from the point of highest eya-so/Six (Fig. 3A). In growth-deficient mutants, the persistent expression of hth does not prevent eya-so/Six activation but blocks retinal differentiation initiation (Fig. 3B).
The function of wg signaling in the subdivision of the Drosophila eye disc discussed here may have parallels during specification of telencephalic and eye anlagen in vertebrates (for a review on this topic, see Wilson and Houart, 2004). In the vertebrate embryo, Six3 transcription factors are expressed and required in the telencephalic and eye anlagen of the prospective forebrain (low Wnt activity); whereas various genes of the Iroquois (Irx) family of transcription factors are expressed in the prospective diencephalon (high wnt activity). Enhanced Wnt signaling suppresses Six3 genes expression and activates Irx gene expression. These changes result in the transformation of telencephalon and eye field into diencephalon. Conversely, low wnt signaling (or loss of Tcf3 function) reduces Irx genes expression and complementary expands rostrally Six3 genes expression (reviewed in Wilson and Houart, 2004; and citations therein). Thus, in each case high and low levels of Wnts, respectively, define opposing domains of transcription factors that specify distinct structures: head fate (hth expression) and eye fate (eya/so expression) in the eye disc and prospective diencephalon (Irx expression) and telencephalon and eye field (Six3 expression) in vertebrate neural plate.
EYE PRIMORDIUM CELLS TRANSITIONS: FROM PROLIFERATIVE AND UNDIFFERENTIATED TO NONPROLIFERATIVE AND COMMITTED
The random proliferation of the early eye–primordium cells becomes spatially patterned just before the overt of retinal differentiation at third instar (Wolff and Ready, 1993; Fig. 4C,D). Retinal differentiation starts in the posterior region of the eye primordium, adjacent to the “firing point”—the point where the dorsoventral boundary intersects the margin and where hedgehog (hh) and upd are expressed (see Fig. 4B). Retinal differentiation progresses anteriorly, in a wave-like manner, leaving on its wake clusters of differentiating retinal cells. The front of this differentiation wave is marked by an indentation of the disc's epithelium called the morphogenetic furrow (MF; see Heberlein and Treisman, 2000, for a review). The posterior marginal cells do not contribute to eye tissue themselves, but they express and secrete the hh signals that trigger initiation of retinal differentiation in adjacent (anterior) cells (Fig. 4B; Dominguez and Hafen, 1997; Chanut and Heberlein, 1997; see also Borod and Heberlein, 1998; Curtiss and Mlodzik, 2000; Pappu et al., 2003). These posterior margin cells also express dpp, but its expression is stronger along the lateral margins and weaker in the “firing point” region (see Fig. 4B).
As the MF moves, and at any given time, approximately 10-cell diameters ahead of the MF, cells stop proliferating and coordinately stall their cell cycles in G1 (Wolff and Ready, 1993; Fig. 3C). Among these arrested cells, regularly spaced atonal-expressing cells will be selected to become the founder (R8 photoreceptor) cell of each ommatidium, or eye unit (reviewed in Frankfort and Mardon, 2002; Voas and Rebay, 2004). R8 will successively recruit the R2, R3, R4 and R5 photoreceptors (Wolff and Ready, 1993) through reiterative activation of the EGF receptor pathway (Freeman, 1996; Dominguez et al., 1998; Kumar et al., 1998; Halfar et al., 2001; Yang and Baker, 2001). Just behind the MF, cells that are not included in the five cell ommatidial preclusters divide synchronously one last time in response to signals from the differentiating cells (Baonza et al., 2002b; Baker and Yu, 2001; Yang and Baker, 2001, 2003). These synchronized mitoses are equivocally called second mitotic wave, because no first mitotic wave happens anterior to the MF. These last mitoses will provide the extra pool of cells to produce the R1, R6, R7 neuronal cells, and the complement of non-neuronal accessory cone and pigment cells, to complete each of the approximately 18 cells that compose each ommatidium (Wolff and Ready, 1993; see also Baonza et al., 2002b). Therefore, the size of the eye is largely determined by the number of ommatidia formed (which is fairly constant and approximately 750), and this number equals the number of R8 founder cells selected from the cell population anterior to the cell cycle stop point.
The dividing cells anterior to the MF express, in addition to ey/toy, at least two other transcription factors: hth (see above) and teashirt (tsh; Pai et al., 1998; Pichaud and Casares, 2000; Bessa et al., 2002, Singh et al., 2004). tsh encodes a zinc-finger transcription factor (Fasano et al., 1991), which behaves both as a repressor and as an inducer of eye development in different regions of the disc (Pan and Rubin, 1998; Singh et al., 2002). hth, ey, and tsh are able to positively cross-regulate each other (Bessa et al., 2002). When cells stop their cell cycle ahead of the MF, they lose hth expression, keeping ey and tsh; simultaneously, these cells up-regulate early retinal genes, like eya (Bessa et al., 2002), and other genes that set cells into a preproneural state (Greenwood and Struhl, 1999; Baonza and Freeman, 2001; Bessa et al., 2002). The preproneural state precedes further eye differentiation at the MF and posterior to it.
The coexpression of hth-ey-tsh in the proliferating cells has been proposed to serve, at least, two roles: (1) it blocks precocious eye differentiation by preventing early retinal gene expression, and (2) it keeps cells proliferating (Bessa et al., 2002; Singh et al., 2002). Whether the first is causing the second is yet unknown; however, it is clear that a key control point in this proliferation-to-differentiation transition is the expression of hth. Thus, eye field cells that lose hth do not grow and de-repress early retinal genes, while ectopic expression of hth, together with ey and tsh, blocks retinal differentiation and triggers extraproliferation. The finding that hth-ey-tsh proteins interact in vitro, suggests a model in which the composition of the protein complex (i.e., presence or absence of hth) might alter their transcriptional regulatory activity (Bessa et al., 2002). Interestingly, in the vertebrate eye, Pax6 maintains retinal cell multipotency (Marquardt et al., 2001). Additionally, the products of the Meis genes, the homologues of the fly hth, positively regulate Pax6 expression during lens development (Zhang et al., 2002), so it seems that some relationships between these genes are found in both invertebrate and vertebrate eyes.
REGULATING TRANSITIONS: ROLES OF DPP, WG, AND HTH
The cell cycle and gene expression changes described above are coordinated by the action of wg and dpp pathways. wg has a role in maintaining proliferation of the undifferentiated eye disc cells. Thus, wg mutant animals have discs smaller than normal. Conversely, hyperactivation of the wg pathway induces overgrowths, in which hth, tsh, and ey are activated, and the early retinal genes (eya, so, and dac) are repressed (Lee and Treisman, 2001; Baonza and Freeman, 2002a). dpp signaling has apparently the opposite function to wg: it is responsible for the G1 arrest of cells ahead of the MF. dpp is expressed at the MF and is thought to act ahead (and anterior) of the MF as a long-range signal (Horsfield et al., 1998). It is proposed that dpp antagonizes the mitogenic role of other signals received in the anterior cells, including the upd/JAK/STAT pathway (reviewed above). Detailed genetic analysis by Bach and coworkers (2003) indicates the contrary. For example, mutations in dpp itself, its receptor or its nuclear transducer, Mad, suppress the overproliferation induced by the overactivation of JAK/STAT. Similarly, mutations in dad, a negative regulator of the dpp pathway, enhance the overproliferation phenotype (Bach et al., 2003). Therefore, more work is needed to unravel the role of these signaling pathways, and their integration, in stimulating and halting proliferation of the undifferentiated cell pool in the eye disc (Fig. 3B).
It is interesting to note that, whereas the discs of individuals lacking dpp grow to a smaller size than those of wild-type individuals, if the loss of dpp signaling is induced in clones of cells along the disc's margin, the mutant cells overgrow, sometimes inducing ectopic retina nonautonomously (reviewed in Heberlein and Treisman, 2000). These cells deficient for dpp signaling de-repress wg. Therefore, we suggest that it is the apposition of wg and dpp signals that stimulates proliferation. A similar mechanism has been shown to operate in leg discs (Theisen et al., 1996).
The regulation of hth expression by these two signals seems to be at the crossroads between cell proliferation and differentiation controls, as hth contributes to maintaining cells proliferating and undifferentiated. Whether hth is expressed or not depends on the ratio of wg and dpp signals (Figs. 2C, 3B; Pichaud and Casares, 2000; Lee and Treisman, 2001; Baonza and Freeman, 2001, 2002a). Thus, as the MF advances, dpp recruits cells from the undifferentiated pool by repressing their hth expression, while causing their withdrawal from the cell cycle. hth's ability to induce tissue overgrowth together with ey and tsh suggests that loss of hth may be directly involved in cell cycle arrest (Bessa et al., 2002).
Another transcription factor regulating cell proliferation is dMitf (Hallsson et al., 2004). Misexpression of wild type or dominant-negative dMitf forms throughout the eye disc represses eye disc cells proliferation (Hallsson et al., 2004). As mention earlier, dMitf is a member of the Myc superfamily of basic helix-loop-helix transcription factors required for correct specification of RPE in vertebrates (Nguyen and Arnheiter, 2000). Two observations are particularly relevant: First, ey/Pax6 directly interferes with dMitf's transcriptional activation ability in vitro, as is also the case for mammalian counterparts (Planque et al., 2001; Hallson et al., 2004). Although ey does not seem to play a major role in promoting cell proliferation, other Pax6 or Pax6-like molecules (such as eyg) might interact with dMitf in this process. Second is the late expression of dMitf in the peripodial cells overlaying the MF, where the dpp gene is expressed. This finding correlates with a proposed role for peripodial cells in controlling cell proliferation in the main epithelial cells (Gibson and Schubiger, 2000). Molecular information from the peripodial to the main disc epithelium might be transferred through peripodial cytoplasmic extensions (Cho et al., 2000; Gibson and Schubiger, 2000; Hallsson et al., 2004). Even so, dMitf cannot itself convey this molecular information, because it encodes a nuclear factor, but could regulate signals that might be transferred across epithelial layers. As dpp and dMitf seem to act on the same cell population and in a similar manner, it is plausible that a regulatory relationship between dMitf and dpp exits. More definitive answers on dMitf role during eye development in flies await experiments in which dMitf function is assayed in a controlled manner at specific regions of the disc and developmental time points and the isolation of dMitf loss-of-function mutations.
WG EXPRESSION SPECIFIES THE HEAD-FORMING REGION AND CONTROLS THE SIZE OF THE ADJACENT EYE-FORMING REGION
An epithelial fold separates the third instar disc margin, or prospective head capsule, from the internal main epithelium and future eye field (Fig. 3A). The central region of the disc and the lateral–anterior margin specify their identity later during development (see above). This finding is in contrast to the posterior margin identity that is fixed earlier, leading to a lineage restriction that separates the main epithelium and posterior margin during second instar (Lim and Choi, 2004; M. Dominguez, J. Bessa, and F. Casares, unpublished results). This clonal restriction seems to reflect the specialized role of the posterior margin in inducing retinal differentiation (Dominguez and Hafen, 1997; Borod and Heberlein, 1998) and also in organizing nonautonomously the planar cell polarity of the differentiated retina (Lim and Choi, 2004).
This further subdivision of the eye disc into head capsule vs. eye domains relies primarily on wg. Removal of wg signaling function during larval development leads to replacement of head cuticle by ectopic retinal tissue (Ma and Moses, 1995; Treisman and Rubin, 1995; Baonza and Freeman, 2002), resulting in flies with heads composed mostly of eye tissue. Reciprocally, forcing wg signaling causes eye tissue replacement with head cuticle (reviewed in Heberlein and Treisman, 2000; and see also Royet and Finkelstein, 1996; Lee and Treisman, 2001; Baonza and Freeman, 2002).
The multiple roles of wg are in agreement with its dynamic expression: wg expression is initially along the dorsal half of the disc (Cavodeassi et al., 1999). Subsequently, it is shifted to an anterior–dorsal and an anterior–ventral marginal domains owing to the repressive action of dpp (Royet and Finkelstein, 1996; reviewed in Heberlein and Treisman, 2000) and positive actions of hh (Dominguez and Hafen, 1997), pannier (Maurel-Zaffran and Treisman, 2000; see also Pichaud and Casares, 2000; Singh and Choi, 2003), and hth (Pichaud and Casares, 2002). Finally, during third instar and continuing through pupal stages, wg expression extends along the whole disc margin to encircle the eye (Royet and Finkelstein, 1996; Tomlinson, 2003; Wernet et al., 2003)—the factors driving this late expression of wg are unknown. Thus, during these stages, wg defines head capsule identity and, probably through its antagonism of dpp, restricts the eye field (Ma and Moses, 1995; Treisman and Rubin, 1995; see also Bessa et al., 2002; Fig. 3A), and concomitantly, it stimulates the proliferation of the cell pool from which both eye and head capsule will originate (Fig. 3B; Pichaud and Casares, 2000; and see above). Finally, and during pupal stages, wg signaling from the head is critical to pattern the peripheral retina (Tomlinson, 2003; Wernet et al., 2003).
The role of wg as a “visual” repressor extends to the development of the ocelli, a group of three simple eyes at the dorsal–medial head. wg expression is excluded from the ocelli primordium, in this case by the combined action of ocelliless and hh (Royet and Finkelstein, 1996, and citations therein). If wg is not removed from this region, the presumptive ocellar region develops as head capsule tissue instead (Royet and Finkelstein, 1996).
EVOLUTIONARY PERSPECTIVE: KEEPING PROPORTIONS AND EVOLUTION OF SHAPES
The development of the eye–antennal disc into the organs of the fly head is one of impressive consistency, with these organs bearing strikingly constant size relations (proportions). These proportions are species-specific (see Fig. 5). That these organ primordia are sequentially specified as adjacent cell populations within the same disc suggests that their growth is coordinated during development to ensure those proportions. An alternative view would be that, once specified, each organ follows an intrinsic growth pace, regulated independently in each organ (see Stern and Emlen, 1999, for a general review on organ proportions in insects). In support of the hypothesis that the sizes of the eye and head capsule are coordinated are experiments in which the size of one of them is drastically affected by alterations in the relative size of the other, without affecting the global size (Fig. 4B). These size alterations might be due to the partition of a common primordium, with a fixed final size, into head vs. eye territories; the larger one is, the smaller the other one becomes (Fig. 5). In wg mutants the eye expands at the expense of head territory. Similarly, several mutations result in reduced eyes and enlarged head (Dac mutants (M. Dominguez and F. Casares, unpublished observations). Also, eya mutant cells develop into head capsule instead of eye (Treisman and Heberlein, 1998; Bessa et al., 2002) or bigger antennae (eyg; M. Dominguez, unpublished observations), leaving the overall size of the head fairly normal. This finding does not preclude that each primordium might activate its own intrinsic growth program. Thus, mutations that act intrinsically in growth control will affect eye size without altering head size (see effects of overexpression of Delta in Dominguez and de Celis, 1998, for example).
The mechanisms of intrinsic eye disc growth control are beginning to be unveiled. In contrast, how the relative growth of the head and eye is regulated or which are the signals that mediate this coordinated control are unknown. Studies in other insects can give some insight into this problem. For example, onion flies (Delia antiqua; Fig. 5A) show a striking sexual dimorphism, with males developing bigger eyes and smaller heads than the females; therefore, in this species, sexual development controls organ proportions. The problem of proportioned growth of organs in this case, the eye and surrounding head capsule, resembles the phenomenon of exaggerated trait development in some insect species (Nijhout and Emlen, 1998; Emlen and Nijhout, 2000). In these species, it is not infrequent that in males, for a given body size, the larger the trait of one organ (for example, the cephalic horns in the horned beetle Onthophagus taurus), the smaller the physically adjacent organ (in this case, the eye; Nijhout and Emlen, 1998). This phenomenon can be explained if the two adjacent organs are competing for a limiting growth factor. Whereas in Drosophila these proportions are hardwired in its developmental program, in case of sexual dimorphisms or when exaggerate traits develop at the expense of others (organ size “tradeoff”; Emlen and Nijhout, 1999), gender or other environmental causes might fine-tune the proportion–control mechanism. In case of organ size tradeoffs, variations in the sensitivity to hormones or competition for a limiting growth factor between adjacent organs can lead to different relative sizes. Such factors might be like the Drosophila imaginal disc-specific growth factors (IDGFs; Kawamura et al., 1999), secreted by the fat body in response to the quality of the diet (Britton and Edgar, 1998; Kawamura et al., 1999). If indeed these organs were part of the same disc or disc complex, as is the case for the eye and head capsule, or the eye and the antenna in the fruit fly, we could imagine signaling mechanisms, operating like wg in the Drosophila eye disc, linking alternative tissue fates and growth. Unfortunately, there is still little information on the development of these flamboyant insects.
Apologies to all those whose work was not cited due to space limitations. Thanks to members of the Dominguez and Casares laboratories who helped us with discussions and unpublished findings, to James Castelli-Gair and John Robert Pearson for critical reading of the manuscript, and to the anonymous reviewers for helpful suggestions.