Progression of the morphogenetic furrow in the Drosophila eye: the roles of Hedgehog, Decapentaplegic and the Raf pathway

The adult retina of Drosophila is composed of a crystalline array of repeating units, ommatidia, which are themselves highly organised structures containing a precise, stereotyped pattern of photoreceptors and support cells (Ready et al., 1976). Ommatidia first appear in dorsal-ventral columns at the posterior edge of the retinal primordium. As these units mature they induce columns of undifferentiated cells located anteriorly to initiate ommatidial development. The first cell of each nascent ommatidium to differentiate as a photoreceptor is the R8 cell, which subsequently recruits other photoreceptors into the growing cluster (Tomlinson and Ready, 1987). Immediately anterior to the most recently formed column of R8 cells is a physical indentation, the morphogenetic furrow, that runs parallel to the ommatidial columns. As the eye develops, the furrow progresses anteriorly, ahead of the photoreceptors, marking the crest of a wave of retinal differentiation that sweeps across the eye. The rate of progression of this wave of differentiation determines the size of the eye, while the orderly recruitment of differentiating cells behind the wave generates the pattern of the retina.

The movement of the morphogenetic furrow, and concomitant photoreceptor differentiation, is driven by the extracellular signaling molecule Hedgehog (Hh). Hh is secreted by differentiated photoreceptors located many cell diameter lengths posterior to the furrow. Loss of Hh severely retards, or halts, furrow progression (Heberlein et al., 1993; Ma et al., 1993; Treisman and Heberlein, 1998). Conversely, ectopic activation of the Hh transduction pathway, produced either by directly expressing Hh (Heberlein et al., 1995) or activating downstream components of the pathway (Dominguez and Hafen, 1997; Ma and Moses, 1995; Pan and Rubin, 1995; Strutt et al., 1995; Wehrli and Tomlinson, 1995), triggers ectopic photoreceptor differentiation and the formation of an ectopic furrow ahead of the endogenous furrow.

Although Hh is essential for orchestrating furrow progression, the mechanisms involved have remained uncertain. Of particular interest are the results of experiments in which Hh signal transduction is blocked in clones of cells by the elimination of Smoothened (Smo), an essential component of the Hh signal transduction pathway (Alcedo et al., 1996; Chen and Struhl, 1996; van den Heuvel and Ingham, 1996). Although clones of cells lacking Smo function cannot transduce Hh, they nevertheless remain capable of forming wild-type ommatidia (Strutt and Mlodzik, 1997; DomÍnguez, 1999), presumably in response to signals induced by Hh in neighboring, wild-type tissue.

Here we show that Hh controls furrow progression as well as the initiation of photoreceptor differentiation by inducing at least two other signals. The first signal is Decapentaplegic (Dpp), which is expressed within and ahead of the morphogenetic furrow (Blackman et al., 1991). We show that Dpp acts at long range from Hh secreting cells to shift cells in front of the furrow from a naive, undifferentiated state to a discrete ‘pre-proneural’ state. This pre-proneural state is marked by the upregulation of a proneural repressor, Hairy, which together with a related repressor, Extramacrochaetae (Emc), blocks the further transition of these cells to the proneural state.

In addition, we find that Hh induces a second signal, as yet unknown, which operates at short range to shift pre-proneural cells to the proneural state. This transition is marked by the downregulation of Hairy and the expression of Atonal (Ato), a proneural activator that is responsible for initiating differentiation of the R8 photoreceptor, the founding cell of the ommatidium (Jarman et al., 1994). This second signal is transduced by activation of the Serine-Threonine kinase Raf and we show that Raf activity is both necessary and sufficient to induce pre-proneural cells to express Ato, become proneural, and differentiate as R8 photoreceptors.

Finally, we present evidence that acquisition of the pre-proneural state is essential for any cell in the retinal primordium to enter the proneural pathway and differentiate as a photoreceptor in response to Raf activation. Although Dpp signaling is sufficient to establish the pre-proneural state, it is not essential. Instead, Hh itself can induce the pre-proneural state in the absence of Dpp signaling, albeit at shorter range. Under these circumstances, the morphogenetic furrow still progresses, but at an abnormally slow rate, resulting in a rudimentary eye. These findings lead us to propose that the primary role of Dpp signaling in the eye is to accelerate the transition of undifferentiated cells to the pre-proneural state. This acceleration is responsible for the normally rapid rate of furrow progression, and hence for size of the eye.

The following null or amorphic alleles for smo, tkv, raf and DER were used: smo³ (Chen and Struhl, 1996), tkv^strII (Nellen et al., 1994), Egfr^top18A (Price et al., 1989) and raf^EA75 (Melnick et al., 1993).

The following previously described transgenes were used: arm-lacZ (inserted on 2L or 2R; Vincent et al., 1994), arm-Gal4 (Sanson et al., 1996), Actin5C>raf⁺>nuc-lacZ (Struhl and Basler, 1993), dpp-lacZ3.0 (Blackman et al., 1991), FRT 39 (Chou and Perrimon, 1992), FRT42 (Xu and Rubin, 1993), hs-flp (inserted on chromosomes 1 or 3; Struhl and Basler, 1993), and UAS-tkv^Q253D, UAS>CD2 y⁺>tkv^Q253D and UAS>CD2 y⁺>dpp (Nellen et al., 1996).

Tubulinα1> f⁺y⁺>raf*, an activated form of the raf gene, raf*, was generated by fusing the coding sequence for the N-terminal myristylation signal of Drosophila Src (MGNKCCSKRQDTMA) to that of a constitutively active form of Human Raf (22W pΔraf, a 305 amino acid N-terminal truncation of human Raf (Stanton et al., 1989; Casanova et al., 1994). The raf* gene was then inserted into a P-element transformation vector (C20NXT with the ry gene deleted), downstream from the Tubulinα1 promoter, and interrupted by an ‘FRT cassette’ containing the forked⁺and yellow⁺ genes (Zecca et al., 1995).

Clones of tkv, smo, DER and smo tkv mutant cells were generated by FRT-mediated recombination (Golic and Lindquist, 1989; Xu and Rubin, 1993). To express Raf^*, Dpp and Tkv^Q253D, or to remove Raf function, various permutations of the ‘flp-out’ technique (Struhl and Basler, 1993) were utilized as indicated below. To create clones by mitotic recombination, first or second instar larvae were heat shocked for 60 minutes at 37°C. To catalyse uniform expression from ‘flp-out’ transgenes, larvae were heat shocked for 60 minutes at 37°C. To produce rare, isolated clones by the flp-out technique, first or second instar larvae were heat shocked for 30 minutes at 34°C.

Activating Raf: hsp70-flp/+ or Y; Tubulin α1> f⁺y⁺>raf*/+. Removing Raf: raf^EA75hsp70-flp; Actin5C>raf⁺>nuc-lacZ/+. Activating Dpp: hsp70-flp/+ or Y; arm-Gal4/UAS>CD2 y⁺>dpp. Activating Tkv: hsp70-flp/+ or Y; arm-Gal4/UAS>CD2 y⁺>tkv^Q253D/+.

Removing Tkv: hsp70-flp; tkv^strIIFRT39E/arm-lacZ FRT39E. Removing Smo: hsp70-flp; smo³FRT39E/arm-lacZ FRT39E. Removing both Tkv and Smo: hsp70-flp; smo³tkv^strIIFRT39E/arm-lacZ FRT39E.

Removing DER: hsp70-flp; FRT42 Egfr^top18A/FRT42 arm-lacZ KP135 (KP135 is a P-element insertion that causes a Minute phenotype; Hart et al., 1993).

Standard protocols for immunofluoresence (Struhl and Basler, 1993) were followed using mouse αCD2 (OX34; Serotec), rabbit αβgal (Cappel), mouse αβgal (Promega), mouse α22C10 (Zipursky et al., 1984), rabbit αato (Jarman et al., 1994), rat αhairy (Howard and Struhl, 1990), rat αCi (Motzny and Holmgren, 1995), mouse αhairy (Carroll and Whyte, 1989) and mouse αelav antisera (O’Neill et al., 1994).

Hh, secreted by maturing photoreceptor cells, is normally responsible for inducing cells within and ahead of the morphogenetic furrow to initiate photoreceptor differentiation (reviewed in Treisman and Heberlein, 1998). Nevertheless, cells which lack Smoothened (Smo) function, and hence the ability to transduce Hh, can form normal ommatidia (Strutt and Mlodzik, 1997; DomÍnguez, 1999). These findings suggest that Hh can induce photoreceptor differentiation in Smo-deficient cells through the induction of other signaling molecules in neighboring wild-type tissue. As a first step towards identifying such secondary signals and analyzing their roles, we have examined the consequences of creating clones of cells homozygous for smo³, an amorphic mutation, on two early markers of retinal development, the expression of Ato and Hairy, which are expressed in adjacent dorso-ventral stripes within and anterior to the morphogenetic furrow.

Ato expression has two prominent phases in the developing eye (Dokucu et al., 1996; Sun et al., 1998; Fig. 1A,B; see also Fig. 7A). In the first phase, Ato is expressed uniformly in a narrow dorso-ventral swath of cells, which demarcates the anterior edge of the furrow. This uniform swath then breaks up into small clusters of Ato expressing cells and resolves into the second phase, a spaced pattern of single Ato expressing cells (the future R8 photoreceptor cells). The first phase of Ato expression is severely reduced or absent in clones of smo³ cells, similar to large clones that lack Hh (DomÍnguez and Hafen, 1997). However, the second phase of expression still occurs, even though it is displaced posteriorly indicating that it is delayed (Fig. 1A,B). This displacement is more severe in the middle of the clone than along the dorsal and ventral borders, producing a crescent shaped distortion of the line of spaced single cells that express Ato. We conclude that cells within smo⁻mutant clones can be induced to express Ato even though they cannot receive Hh, provided that they are located near to wild-type cells across the clone border.

Equivalent effects were observed for Hairy expression. Hairy is normally expressed at peak levels in a dorso-ventral stripe positioned immediately anterior to the Ato stripe, but is abruptly downregulated in more posteriorly situated cells (Carroll and Whyte, 1989; Brown et al., 1995; Fig. 1C,D; see also Fig. 7A). Clones of smo³ cells have only a modest effect on Hairy expression anterior to the furrow, causing a slight, but consistent, posterior displacement of the anterior edge of the stripe. However, they are associated with a pronounced failure to repress Hairy expression in some, but not all, posteriorly situated smo³ cells (Fig. 1C). As in the case of Ato expression, the exceptional mutant cells which retain the normal downregulation of Hairy are those positioned close to the lateral and posterior borders of the clones. Just within the lateral border, we typically observe a line of cells, one or two cell diameter lengths wide, where Hairy expression is repressed (e.g. Fig. 1D). Along the posterior border, the zone of mutant cells in which Hairy expression is repressed is usually wider.

We interpret these results to indicate (1) that Hh normally induces cells to express a secondary signal (or signals), which can activate Ato expression and repress Hairy expression; (2) that this signal acts non-autonomously, allowing it to move from wild-type cells where it is induced by Hh to nearby smo³ cells where it regulates Ato and Hairy expression; and (3) that the range of this signal is short, restricting its action to only one or two cells across the lateral borders of smo³ mutant clones. A somewhat greater range of action is apparent along the posterior borders of such clones, perhaps because the adjacent wild-type cells were induced by Hh to send this signal at an earlier time than those along the lateral (more anterior) borders of the clone, allowing the signal more time to accumulate to higher levels and to move deeper into mutant tissue.

To examine how the posterior displacements in Ato and Hairy regulation in smo³ clones influence subsequent ommatidial development, we examined the expression of the protein Elav, a marker of photoreceptor differentiation (Robinow and White, 1988). As shown in Fig. 2, clones of smo³ cells are capable of differentiating as photoreceptors, in agreement with previous findings (Strutt and Mlodzik, 1997; DomÍnguez, 1999). However, there is a significant delay. In wild-type tissue, Elav expression initiates immediately posterior to the morphogenetic furrow with the specification of the R8 cell and continues as other photoreceptors are recruited into the ommatidial cluster (Fig. 2A). In clones of smo³ cells, there is a clear posterior displacement in the onset of photoreceptor differentiation in mutant cells (Fig. 2A,B): photoreceptor differentiation is first seen at the posterior, and occasionally lateral, edges of the clone, correlating with the effects of neighboring wild-type tissue on Hairy and Ato expression and indicating a general delay in photoreceptor differentiation. However, as seen in more posteriorly situated clones (Fig. 2C), most or all of the smo³ tissue eventually differentiates as normally patterned ommatidia.

Thus, Hh signal transduction is not autonomously required for presumptive eye cells to express Ato, downregulate Hairy, or differentiate as photoreceptors. This is in contrast to the general requirement for Hh signaling revealed by experiments in which Hh signaling is blocked throughout the entire disc using temperature-sensitive hh mutations (Ma et al., 1993). In the latter case, loss of Hh signaling causes a rapid and complete block in photoreceptor differentiation and furrow progression. As we demonstrate below, Hh signaling appears to induce at least two secondary signals that are essential for the normal recruitment of undifferentiated cells to form the R8 photoreceptors. One of them appears to be the short-range signal that can induce Ato expression and repress Hairy in clones of smo⁻ cells. The second, Decapentaplegic (Dpp), appears to act at longer range to prime cells to receive this short range signal.

One candidate for a secondary signal, which acts downstream of Hh in the developing retina, is the TGF-β homologue Dpp. Dpp is induced by Hh just anterior to the morphogenetic furrow (Heberlein et al., 1993). Moreover, experiments in other discs have established that Dpp can act at long range from its source to mediate the organizing activity of Hh on more anteriorly situated tissue (Zecca et al., 1995; Lecuit et al., 1996; Nellen et al., 1996). However, previous studies have shown that Dpp signaling is not essential for either photoreceptor differentiation or propagation of the furrow once photoreceptor differentiation initiates at the posterior edge of the eye primordium (Burke and Basler, 1996; Wiersdorff et al., 1996; Chanut and Heberlein, 1997; Pignoni and Zipursky, 1997). These findings challenge the notion that Dpp mediates the organizing activity of Hh in front of the furrow.

To test whether Dpp has such a organizing role we performed two kinds of experiments. In the first, we ectopically expressed Dpp, or activated Thickveins (Tkv), a type I TGFβ receptor required for all known Dpp activities (Ruberte et al., 1995), anterior to the furrow. In the second, we blocked Dpp expression or Tkv activity. The results of these experiments indicate that Dpp signaling is both necessary and sufficient to upregulate Hairy expression anterior to the furrow and to maintain the normal rate of furrow progression, but that it is neither necessary nor sufficient to activate Ato expression and initiate photoreceptor differentiation in more posterior cells.

Ectopic Dpp signaling

Dpp was constitutively expressed in most cells of the eye primordium, including those ahead of the furrow, using a combination of the Flp-out and Gal4/UAS techniques (Nellen et al., 1996; see Materials and Methods). Normally, Hairy is expressed at high level in a stripe of cells positioned just anterior to the furrow, its level of expression fading in progressively more anterior cells to form a gradient (e.g. Fig. 3A,C; see also Fig. 7A). However, in eye discs in which Dpp expression is induced in most cells, Hairy is expressed at uniformly high levels in presumptive eye cells anterior to the furrow (Fig. 3B,D). In contrast, there appears to be little if any alteration in the behavior of cells within or behind the furrow: both the expression of Ato as well as the pattern of neural differentiation (assayed by staining with the monoclonal antibody MAb22C10; Zipursky et al., 1984) appear normal.

Ectopic activation of Tkv

A constitutively active form of Tkv, Tkv^Q253D, was expressed, as above, except using a modification of the Flp-out/Gal4 technique to generate isolated clones of tkv^Q253D expressing cells marked by the loss of expression of the reporter protein CD2 (Nellen et al., 1996; see Materials and Methods). We find that clones of UAS>tkv^Q253D cells express uniformly high levels of Hairy when they are located ahead of the furrow, even in regions where the level of Hairy expression is normally low or undetectable (Fig. 3E). Moreover, this high-level Hairy expression appears to be autonomous to the clone, indicating that Hairy is normally upregulated only in anterior eye cells in which the Dpp signal transduction pathway is active. Clones of UAS>tkv^Q253D cells located within or behind the furrow appear to develop normally in terms of Hairy expression and photoreceptor differentiation (Fig. 3F; data not shown), except that UAS>tkv^Q253D cells in the vicinity of the furrow are associated with a slight anterior displacement in the stripe of Hairy expression (Fig. 3F). This anterior displacement suggests that the furrow has accelerated as it progressed through the clone. These results, as well as the results obtained with ectopic expression of Dpp, suggest that Dpp signaling anterior to the furrow normally induces high levels of Hairy expression, but is not sufficient to initiate photoreceptor differentiation. They also suggest that Dpp signaling positively regulates the rate of furrow progression.

Absence of Dpp signaling

The dpp^blk mutation is associated with a deletion of cis-acting regulatory sequences that are essential for Hh-dependent transcription of dpp in the eye (Masucci et al., 1990). As a result, Dpp signaling in the eye disc is abolished or severely reduced anterior to the furrow, and the resulting eye is greatly reduced in size in both the dorsal-ventral and antero-posterior axis. We compared Hairy expression in wild-type and dpp^blk disks, using the upregulation of Cubitus interruptis (Ci), a protein that is stabilized in response to Hh signaling (reviewed in Ingham, 1998), as a marker of the position at which the furrow should normally form (Fig. 4).

In wild-type eye discs, Ci accumulates to peak levels in a dorso-ventral stripe of cells just posterior to the stripe of peak Hairy expression (Fig. 4A), consistent with our finding that Hairy expression is repressed in response to Hh signaling within the furrow (Fig. 1C,D), but is activated by Dpp signaling anterior to the furrow (Fig. 3). In contrast, the stripe of maximal Hairy expression is displaced posteriorly in dpp^blk discs relative to the stripe of maximal Ci expression (Fig. 4B). Moreover, the furrow appears to have moved only a small distance from the posterior edge of the presumptive eye primordium, even in eye discs from mature third instar larvae, consistent with the ‘small eye’ phenotype observed in the adult.

These results indicate that Dpp signaling is normally required to activate high level Hairy expression in a stripe positioned just anterior to the furrow. They also indicate that Dpp signaling is necessary to sustain the normal rate of furrow progression. Finally, they suggest that Dpp signaling influences the response of cells to peak levels of Hh signal transduction: Hairy expression is downregulated in these cells in wild-type discs, but not in dpp^blk discs.

Absence of Tkv activity

Tkv function was eliminated in clones of marked cells using Flp-mediated mitotic recombination (Golic, 1991; see Materials and Methods). As shown in Fig. 5, clones of cells homozygous for tkv^strII, a null allele (Nellen et al., 1994), are associated with a severe reduction in Hairy expression anterior to the normal position of the furrow. In addition there appears to be a posterior displacement of Hairy expression, so that the gene is upregulated in Hairy expression. The first is Dpp, which is normally required for the upregulation of Hairy anterior to the furrow. The second is a factor that can induce Hairy expression within and behind the furrow, but which does so only in the abnormal situation in which Dpp signaling is blocked. As we show below, this second factor is Hh itself.

Ato expression is also altered in tkv⁻ clones (Fig. 5C,D). In particular, the first phase of uniform, high level Ato expression is absent or greatly reduced, correlating with the displaced, posterior expression of Hairy. However, the second phase of spaced, single cell expression of Ato in the presumptive R8 cells is apparent within the mutant clones, although it is displaced posteriorly relative to spaced single cells expression of Ato in neighboring, wild-type tissue (Fig. 4C,D). The posterior displacements we observe in both Ato and Hairy expression indicate that the rate of furrow progression is retarded in mutant cells, corroborating previous findings that tkv mutant clones cause delays in subsequent steps of photoreceptor differentiation (Burke and Basler, 1996). Thus, neighboring, posteriorly situated cells where it would normally be downregulated. Based on these results, as well as those obtained using the dpp^blk allele, we infer that there are at least two factors in the eye capable of upregulating our results with both the dpp^blk mutation and tkv⁻ mutant clones show that Dpp signaling is not essential for Ato expression or for photoreceptor differentiation. However, Dpp signaling does appear to be necessary to prime cells anterior to the furrow so that they are incorporated into the furrow at the normal rate.

In the absence of Dpp signaling, Hairy upregulation and Ato expression are retained, but are displaced posteriorly so that they now occur in cells that are likely to be exposed to peak levels of Hh (Figs 4, 5). To determine whether this delayed expression is induced directly by Hh signaling or is due to another signaling molecule induced in response to Hh, we generated clones of cells lacking the activities of both Smo as well as Tkv (see Materials and Methods). We find that smo³tkv^strII clones fail to express high levels of Hairy in a cell autonomous fashion (Fig. 6A). Similarly, smo³tkv^strII clones autonomously fail to express Ato (Fig. 6B). These results indicate that in the absence of Dpp signaling, undifferentiated eye cells must receive Hh in order to upregulate Hairy and express Ato.

We note that the loss of both Hh and Dpp signal transduction not only prevents Ato expression, which is required for R8 photoreceptor differentiation, but also results in a general block in photoreceptor differentiation. First, we find that there is a complete and autonomous absence of Elav expression in smo³tkv^strII clones (Fig. 6C,D). Second, such clones fail to contribute any photoreceptors in the adult eye (G. S. and A. Tomlinson; data not shown). These findings contrast with those obtained from the analysis of clones of ato mutant cells (Jarman et al., 1994). Although cells deficient for Ato activity cannot differentiate as R8 photoreceptors, they can be recruited by wild-type R8 cells to make any or all of the remaining photoreceptors in an ommatidial cluster. Thus, presumptive eye cells must receive either Hh or Dpp in order to be able to enter the proneural pathway and differentiate as photoreceptors (see Discussion).

Our analysis indicates that Hh orchestrates Hairy and Ato expression as well as furrow progression in a manner that depends on Dpp signaling ahead of the furrow. Yet Dpp signaling is not sufficient to activate Ato or to initiate photoreceptor differentiation (Figs 3, 4). Conversely, cells devoid of Smo, and hence unable to receive Hh, can still be induced by neighboring wild-type cells to express Ato and make photoreceptors (Figs 1, 2). These findings argue for an additional signal involved in mediating the organizing activity of Hh on photoreceptor differentiation and furrow progression. The signal transduction pathway downstream of receptor tyrosine kinases involving Ras and Raf has been shown to have multiple roles during the formation of the Drosophila eye, most notably in mediating signals by the Sevenless and Epidermal Growth Factor receptors (reviewed in Wassarman et al., 1995; Freeman, 1997). Further, misexpression of activated forms of Ras, or the EGF receptor, anterior to the furrow can initiate photoreceptor differentiation (DomÍnguez et al., 1998; Hazelett et al., 1998). We therefore asked whether this signaling pathway might also be utilized in the induction of Ato.

To constitutively activate the Raf transduction pathway in the eye disc, we expressed a myristylated and truncated form of human Raf, which we call Raf*, in clones of cells using the Flp-out technique (Materials and Methods). Under conditions that cause expression of constitutively active Raf* in most cells, we observe that the anterior stripe of peak Ato expression expands several cell diameter lengths forward, to fill the region just anterior to the position of the furrow where Hairy is normally expressed at peak level (compare Fig. 7A,B). However, the anterior limit of peak Ato expression does not extend past that of Hairy expression. Examination of Hairy under the same conditions reveals that the normally high levels of expression just anterior to the furrow are diminished (Fig. 7B). This reduction in Hairy expression, alone, is not likely to account for the expansion of Ato expression, because Ato expression is not altered in eye discs lacking hairy activity (Brown et al., 1995). However the combined loss of both Hairy and the related repressor protein Emc causes ectopic Ato expression anterior to the furrow (Brown et al., 1995). In the presence of indiscriminate Raf* activity, we find that Emc expression, like that of Hairy, appears to be diminished anterior to the furrow (data not shown). Hence, we can attribute the expansion of Ato expression to the concomitant reduction in both Hairy and Emc expression.

Constitutive expression of Raf* also induces ectopic photoreceptor differentiation anterior to the furrow, as assayed by the expression of the neuronal antigen 22C10 (Fig. 7C). Similar to the expansion of Ato induced by Raf* (Fig. 7B), ectopic photoreceptor differentiation is restricted to a dorso-ventral stripe of cells just anterior to the normal position of the furrow. Not all the cells within this column differentiate as photoreceptors. Rather, there is a saltatory pattern of differentiation, and we attribute this to the process of lateral inhibition which normally restricts the number of proneural cells that differentiate as photoreceptors (Cagan and Ready, 1989; Baker et al., 1996; Baker and Yu, 1997; Spencer et al., 1998; Chen and Chien, 1999).

Similar results are obtained under conditions that create small, rare clones of cells expressing Raf* (see Materials and Methods). In this case, we observe small, isolated clusters of photoreceptors, located anterior to the endogenous furrow and surrounded by what appears to be an ectopic furrow (marked in this experiment by the expression of a dpp-lacZ reporter gene; Fig. 7D). As we observed for neural differentiation induced by widespread expression of Raf* (Fig. 7C), small clones of Raf* expressing cells initiate ectopic photoreceptor development only when they are close to the furrow.

We interpret these findings as follows. During normal development, Dpp signaling anterior to the furrow creates a pool of cells which are primed to enter the proneural pathway, but blocked from doing so by the expression of high levels of Hairy and Emc. We refer to these cells as being ‘pre-proneural’. Release from this block requires an additional signal which is induced at short range by Hh signaling and transduced by activation of the Raf pathway.

The results seen with activated Raf suggest that endogenous Raf may normally regulate Ato expression. Therefore, Ato expression was examined in clones of cells that lack Raf function. Clones of raf⁻ cells marked by the expression of a nuclear-localized form of β-galactosidase were generated using the Flp-out technique (see Materials and Methods). Raf⁻ cells do not express Ato (Fig. 7F-H). Moreover, the absence of Ato expression is cell-autonomous, indicating that Raf is normally required to transduce a signal that is essential for Ato expression.

Raf normally mediates signals from receptor tyrosine kinases (reviewed in Schlessinger and Bar-Sagi, 1994; Marshall, 1995). Two candidate receptor tyrosine kinases, which could activate Raf ahead of the furrow, are the Drosophila EGF receptor (DER) and one of the Drosophila FGF receptor homologues, Heartless (Htl). DER is expressed at high levels ahead of the furrow (Zak and Shilo, 1992), and is required for photoreceptor differentiation (Xu and Rubin, 1993). Similarly, two ligands for DER, Spitz and Vein, are active within the furrow and regulate photoreceptor differentiation (Tio et al., 1994; Tio and Moses, 1997; Spencer et al., 1998). However, we find that Ato is expressed in DER⁻ clones, indicating that DER is not essential for the Raf-dependent activation of Ato expression (Fig. 7E).

Htl is also expressed in the morphogenetic furrow (Shishido et al., 1993; data not shown). Similar to DER, however, removal of Htl has no effect on Ato expression (data not shown). It is possible that both DER and Htl can receive the Ato inducing signal, accounting for the ability of cells lacking one or the other function to initiate Ato expression. Alternatively, Raf may transduce a signal that is received by another receptor.

The organizing activity of Hh in most ectodermal derivatives of Drosophila depends on their subdivision into anterior (A) and posterior (P) compartments (reviewed in Lawrence and Struhl, 1996). This subdivision is governed by the selective activity of the transcription factor Engrailed (En) in P compartment cells, which directs them to express Hh. Conversely, the absence of En activity in A cells renders them responsive to Hh and immiscible with P cells, creating a stable, inductive interface between P and A cells. Hh signaling across this interface directly controls the development of nearby A compartment cells (Mullor et al., 1996; Strigini and Cohen, 1997; Struhl et al., 1997a,b) and induces secondary signals such as Dpp and Wg, which act at longer range to organize growth and patterning throughout the segment (Basler and Struhl, 1994; Tabata and Kornberg, 1994).

The development of the Drosophila eye is a notable exception to this general paradigm for Hh signaling. Here, Hh signaling is not linked to a fixed, compartmental segregation between Hh-sending and Hh-receiving cells. Instead, the entire eye arises from cells within the A compartment, and special mechanisms exist, first, to create a small subpopulation of Hh expressing cells at the posterior edge of the eye primordium (Pignoni and Zipursky, 1997), and second, to recruit additional cells to enter and expand the population of Hh secreting cells anteriorly across the primordium (Heberlein and Moses, 1995; Treisman and Heberlein, 1998). Recruitment depends on the addition of a novel regulatory loop between Hh reception and Hh signaling that is not found in other ectodermal derivatives.

In essence, presumptive eye cells that receive an Hh signal are induced to become Hh secreting cells, thereby creating a ‘rolling’ Hh signaling interface. This novel regulatory loop is also linked to neural differentiation. The first cells in the eye to secrete Hh appear to be the first mature photoreceptors. Hh emanating from these cells then induces anteriorly situated, undifferentiated cells to develop into Hh-secreting photoreceptors, thus linking the rolling Hh interface with the wave of retinal differentiation which progresses across the eye primordium.

Here, we show that normal progression of the Hh-signaling interface across the retinal primordium requires at least two additional signals, both induced by Hh (Fig. 8). The first signal, Dpp, appears to act at long range on anteriorly situated cells to prime them to be responsive to a second, proneural inducing signal. We term such primed cells as ‘pre-proneural’, a state which correlates with the upregulation of Hairy. The second signal, as yet unknown, appears to act at short range and to induce pre-proneural cells to enter the proneural pathway. Transition to the proneural state depends on activation of the Raf signal transduction pathway and correlates with the upregulation of Ato, a proneural activator. Finally, we find that in the absence of Dpp signaling, undifferentiated cells can still be induced by Hh to adopt the pre-proneural state. However, this transition occurs at shorter range and appears to be less efficient, reducing the rate of furrow progression and the size of the eye. Thus, we infer that Dpp signaling is not required for progression of the furrow, per se, but rather for the normal rate of progression.

In our interpretation of Hh signaling and furrow progression, we propose that undifferentiated cells anterior to the furrow are first primed to become ‘pre-proneural’, and that acquisition of this distinct cell state is a prerequisite for their further progression to the proneural state, a transition that involves the activation of Raf and consequent upregulation of Ato. It is therefore relevant to ask what properties define pre-proneural cells and distinguish them from the other cells in the eye.

Although we have used the upregulation of Hairy to mark the pre-proneural state, acquisition of this state also correlates with other regulatory changes that occur anterior to the furrow, such as the concomitant upregulation of Daughterless (Da), a proneural activator (Brown et al., 1995, 1996; S. G., unpublished findings), the loss of Cyclin A expression, and the entry of cells into G1 arrest (Thomas et al., 1994). We note that Hairy and Da, as well as Extramacrochaetae (Emc), a related basic helix-loop-helix transcription factor expressed anterior to the furrow, are likely to be determinants of the pre-proneural state. Hairy and Emc are transcriptional repressors that actively block the expression or activity of proneural activators including Ato and Da. Thus, acquisition of the pre-proneural state appears to involve the upregulation of proneural gene functions such as Da, as well as proneural repressors like Hairy which, together with Emc, holds these proneural activities in check.

This view is supported by the observation that the simultaneous loss of Hairy and Emc activity leads to the precocious differentiation of photoreceptors several cell diameter lengths anterior to the furrow (Brown et al., 1995). The ectopic differentiation thus revealed appears to coincide with that of the normal domain of peak Hairy and Emc expression. It is also located in the same position as the ectopic stripe of Ato expression caused by indiscriminate activation of Raf (e.g. Fig. 7). However, photoreceptor differentiation, or Ato expression, does not extend further anteriorly into the zone of mitotically active, undifferentiated cells in either case. Thus, pre-proneural cells can also be distinguished from more anteriorly situated, undifferentiated cells by their latent potential to activate Ato expression and differentiate as photoreceptors following either the loss of proneural repressors, or the activation of Raf.

We note that another attribute of the pre-proneural state is that it appears to be a general prerequisite for photoreceptor differentiation. In particular, we observe that cells which are devoid of both Smo and Tkv activity, and hence unable to make the transition to the pre-proneural state, appear unable to be recruited into developing ommatidia to make any of the photoreceptors. This result is in contrast to the finding that cells which retain either Smo or Tkv activity can be recruited to form any one, or all, of the eight photoreceptors (Burke and Basler, 1996; Strutt and Mlodzik, 1997; DomÍnguez, 1999; the present work). It also contrasts with previous findings showing that cells which lack Ato activity can form any photoreceptor except for R8 (Jarman et al., 1994).

We interpret the absolute failure of smo tkv mutant cells to differentiate as photoreceptors as evidence for a two-step mechanism of photoreceptor differentiation in which cells are first induced by either Hh or Dpp to become pre-proneural, and then require activation of Raf to become proneural (Fig. 8). We further propose that this mechanism is general, beginning with recruitment of the R8 photoreceptor and applying to each subsequent round of Raf activation, which adds the remaining photoreceptors that compose the mature ommatidium.

The finding that smo mutant cells can express Ato and form normally patterned ommatidia provided that they are near wild-type cells indicates that Hh can specify photoreceptor differentiation indirectly, through the induction of other signaling molecules. However, we have shown that Dpp, which is induced by Hh, does not have this capacity. Hence, we infer that some other signal induced by Hh, which we refer to as factor X (Fig. 8), is normally responsible for initiating the proneural state in pre-proneural cells.

Although we have not identified factor X, we have obtained evidence that it operates through the induction of the Raf transduction pathway. First, we find that ectopic activation of the Raf pathway can cause pre-proneural cells to ectopically express Ato and differentiate precociously as photoreceptors. Second, absence of Raf activity autonomously blocks the expression of Ato in these same cells. Our data also suggest that factor X has a relatively short range of action. For example, the ability of wild-type cells to block Hairy expression and induce Ato expression in neighboring smo mutant tissue appears tightly restricted, extending over only a few cell diameter (Fig. 1).

The most likely candidates for factor X are Spitz and Vein, ligands of the Drosophila EGF receptor (DER), which appear to be active within, or just posterior to, the furrow (Tio et al., 1994; Tio and Moses, 1997; Spencer et al., 1998). However, we have found that clones of cells that are devoid of DER activity nevertheless express Ato within the furrow with little if any posterior displacement relative to neighboring wild-type cells (Fig. 7E). We have obtained similar results for the FGF receptor Htl, which like DER is also expressed in the eye (Shishido et al., 1993; data not shown). This is in contrast to clones of cells devoid of Raf activity in which we have never observed Ato (e.g. Fig. 7F,G). These results suggest that if DER and Htl transduce factor X, they do so in a redundant fashion with each other, or with other receptors. Similarly, they also suggest that neither Spitz nor Vein correspond to factor X.

We envisage pre-proneural cells as metastable, having a latent proneural capacity that is actively held in check by proneural repressors such as Hairy and Emc. How does activation of Raf precipitate the transition to the proneural state? Because the simultaneous loss of both Hairy and Emc activities causes a similar expansion of Ato expression to that resulting from the expression of activated Raf, we suggest that Raf activation may normally induce transition to the proneural state by blocking the expression or activity of these repressors. Consistent with this possibility, Hairy contains potential phosphorylation sites for MAPK, a kinase downstream of Raf in the signaling pathway (Marshall, 1994). We note that Daughterless expression is also upregulated in the furrow and is necessary to maintain Ato expression (Brown et al., 1996). Moreover, Daughterless, like Hairy, contains phosphorylation sites for MAPK, raising the possibility that Raf activity may directly potentiate proneural activators at the same time that it downregulates the activities of their repressors. Similar events may also occur in mammalian neural differentiation, as NGF-induced differentiation of the mammalian neuronal cell line PC12 is mediated by the phosphorylation of HES-1, a Hairy related protein (Strom et al., 1997).

When Dpp signaling is blocked by eliminating the signal itself (as in dpp^blk mutants; Fig. 4), or by eliminating Tkv, an essential component of the Dpp signal transduction pathway (Fig. 5), we find that undifferentiated cells can still be induced (1) to adopt the pre-proneural state (marked by Hairy expression), (2) to become proneural (marked by Ato expression) and (3) to differentiate as photoreceptors (marked by Elav and 22C10 staining and confirmed by the presence of mutant photoreceptors in the adult). Further, in this abnormal context, all three outputs depend on the direct action of Hh on the responding cells, as indicated by the cell-autonomous absence of Hairy and Ato expression in smo tkv mutant cells as well as the complete failure of these double mutant cells to differentiate into photoreceptors. Thus, Hh signaling, without any help from Dpp, can organize the normal progression of cell states observed during normal eye development.

Although Hh signaling can drive furrow progression independently of Dpp, the rate at which this progression occurs is abnormally slow. This is evident both in the marked posterior displacement of Hairy and Ato expression in tkv mutant clones, and in the severely retarded movement of the furrow in dpp^blk mutant discs. As a consequence, the mutant tissue forms too few ommatidia in the antero-posterior axis, contributing to the ‘small eye’ phenotype of dpp^blk mutant flies.

How might Dpp signaling regulate the rate of furrow progression? Dpp signaling could accelerate the conversion of undifferentiated cells into pre-proneural cells because it moves further anteriorly than Hh or because it is more potent than Hh in inducing this transition in cell state. Dpp signaling might also accelerate the transition of pre-proneural cells to proneural cells, even though it is not sufficient to induce this transition in the absence of Hh signaling or Raf activation. For example, Dpp signaling could enhance the ability of cells to respond to the short-range signal which induces Raf activity.

One way to view the small eye phenotype of dpp^blk eyes is that it represents an evolutionarily more primitive eye, dependent only on Hh signaling. From this perspective, the addition of Hh-induced Dpp signaling provides a significant enhancement of the organizing capacity of Hh through acceleration of the rate of furrow progression.

The situation in the eye may thus be analogous to that of Hh signaling in the limbs. Although Hh signaling can operate at short range in the limbs to govern patterning in A compartment cells close to the A/P compartment boundary (Jiang and Struhl, 1995; Mullor et al., 1997; Strigini and Cohen, 1997), the elaboration of limbs from the body wall appears to depend on the induction of Dpp and Wg as secondary signals (Campbell and Tomlinson, 1995). When these outputs are blocked, limbs do not form, or are rudimentary. The situation in the limb-bearing segments contrasts with that of the limb-less abdominal segments where Hh acts directly, rather than through Dpp and Wg, to organize most aspects of cell pattern in the A compartment (Struhl et al., 1997a,b). We suggest that the abdominal segments, like the eyes of dpp^blk flies, represent the evolutionarily more primitive condition, in which Hh signaling operates on its own. To organize larger, more complex structures such as the wild-type eye or the limbs, Hh signaling might need to operate through the induction of additional signaling molecules, such as Dpp and Wg.

We thank Atsuko Adachi for generating transformant lines of the Tubulinα1> f⁺y⁺>raf* transgene, L. Zipursky, A. Jarmen, K. Howard, R. Holmgren, S. Carroll and G. Rubin for antisera, and A. Tomlinson, G. Campbell and L. Jaffe-Greenwood for advice and discussion. S.G. was supported by the College of Physicians and Surgeons, Columbia University. G.S. is an Investigator of the Howard Hughes Medical Institute.