Broad-complex, but not Ecdysone receptor, is required for progression of the morphogenetic furrow in the Drosophila eye

Hormones play essential roles in the development of many animals, particularly during postembryonic developmental changes, such as amphibian and insect metamorphosis, and mammalian puberty. While DNA-binding nuclear receptors have been characterized for many steroid hormones, our understanding of how steroid hormone signaling directs tissues to differentiate remains rudimentary.

In Drosophila the ecdysteroids co-ordinate and trigger the developmental events associated with molting and metamorphosis. Ecdysone is released from the ring gland in pulses during development, and converted to 20-hydroxyecdysone (20-HE) in the fat body and target tissues (Riddiford, 1993). Pulses of ecdysteroids at the end of larval life and during pupal development direct many larval tissues to die (Jiang et al., 1997), others to remodel (Levine et al., 1995), and imaginal tissues to undergo proliferation, morphogenesis and differentiation (Fristrom and Fristrom, 1993).

A heterodimeric nuclear receptor complex for 20-HE is composed of the Ecdysone receptor (EcR) and Ultraspiracle (Usp); the EcR/Usp dimer can bind 20-HE-responsive regulatory DNA sequences, repressing transcription in the absence of hormone, and activating it in the presence of 20-HE (Cherbas et al., 1991; Koelle et al., 1991; Thomas et al., 1993; Yao et al., 1993). Although null EcR or usp mutants die as embryos or young larvae (Bender et al., 1997; Perrimon et al., 1985), the essential roles of these genes at metamorphosis have been demonstrated by rescuing mutants through early lethal phases with inducible transgenes; both usp and EcR third instar mutants fail to execute the normal program of metamorphosis (Hall and Thummel, 1998; Li and Bender, 2000). Usp, an RXR homolog, is thought to be able to dimerize with multiple steroid receptors (Oro et al., 1990; Sutherland et al., 1995; Zelhof et al., 1995a).

Ashburner and colleagues proposed that ecdysone signaling directly activates a set of early puffs or genes. These early genes were proposed to encode factors that transduce the ecdysteroid signal and impose a temporal sequence of response by activating other, late, genes as well as regulating their own transcription (Ashburner et al., 1974). In support of this model, the early gene Broad complex (BR-C; br – FlyBase) was found to be required for the expression of late genes as well as for metamorphic events in tissues other than the salivary glands (Belyaeva et al., 1980; Belyaeva et al., 1981; Stewart et al., 1972). The early genes at the E74, E75 and BR-C loci all encode transcription factors, and have mutant phenotypes consistent with their roles as transducers of ecdysone signals (Burtis et al., 1990; Buszczak et al., 1999; DiBello et al., 1991; Fletcher et al., 1995; Fletcher and Thummel, 1995; Kiss et al., 1988; Segraves and Hogness, 1990).

The BR-C plays a key role in directing the appropriate, stage-specific responses to hormone signaling at pupariation, and may provide competence to cells to respond to ecdysone signals (Karim et al., 1993). BR-C encodes a family of transcription factors that all possess one of four possible alternative pairs of zinc fingers (Z1-Z4), and a BTB-POZ protein interaction domain (Bardwell and Treisman, 1994; Bayer et al., 1996a; DiBello et al., 1991; Zollman et al., 1994). BR-C null mutants (non-pupariating; npr1, that lack all BR-C protein isoforms) are unable to pupariate, and die after a prolonged third instar, whereas mutants lacking isoform subgroups die as pupae (Kiss et al., 1988; Stewart et al., 1972). In mid and late third instar, activation of BR-C is less dependent on EcR and Usp function than are E74 and E75 (Bender et al., 1997; Hall and Thummel, 1998). In addition, BR-C function is required for maximal induction of E74 and E75 (Belyaeva et al., 1981; Karim et al., 1993; Zhimulev et al., 1982), whereas BR-C is not reciprocally regulated by E74 (Fletcher and Thummel, 1995).

There is evidence to support a novel hormone response pathway in the mid-third instar. This includes the moderate transcriptional induction of EcR, E74B and BR-C in a usp-independent manner (Andres et al., 1993; Hall and Thummel, 1998; Huet et al., 1993; Thummel, 1996; von Kalm et al., 1994). The DHR78 steroid receptor (Hr87 – FlyBase) may act in this hierarchy; it is widely expressed at this time, and binds to many early puffs (Fisk and Thummel, 1995; Fisk and Thummel, 1998; Zelhof et al., 1995b). DHR78 mutants fail to exhibit the characteristic mid-third instar pattern of hormone-dependent gene activity. Ectopic expression of DHR78 is not detrimental, suggesting that its activity may be regulated by a ligand (Fisk and Thummel, 1998).

One dramatic consequence of ecdysteroid signaling at metamorphosis is the transformation of the imaginal discs into the epidermis and peripheral nervous system of the adult fly. During pupal development, discs evert, elongate and terminally differentiate in an ecdysteroid-dependent manner (Fristrom and Fristrom, 1993). EcR, usp and BR-C mutants all show defects in disc eversion and elongation (Fristrom et al., 1981; Hall and Thummel, 1998; Kiss et al., 1988; Li and Bender, 2000).

Prior to overt morphogenesis, imaginal discs undergo patterning and specification of neuronal elements (Cohen, 1993). For example, in the third instar eye disc, retinal specification begins at the posterior margin and expands anteriorly, with the addition of new rows of photoreceptor clusters every two hours (Wolff and Ready, 1993). Associated with the anterior boundary of retinal patterning are a synchronized cell cycle arrest in G1 (Thomas et al., 1994; Wolff and Ready, 1991), relaxation of heterochromatic gene silencing (Lu et al., 1998), and a coordinated apical-basal contraction of cells that produces the indentation known as the morphogenetic furrow (Ready et al., 1976). The expression of Hedgehog (Hh) and possibly Decapentaplegic (Dpp) signaling proteins posterior to and in the furrow, respectively, drive the furrow anteriorly at a controlled pace (Curtiss and Mlodzik, 2000; Heberlein and Treisman, 2000).

Such early neuronal differentiation in imaginal discs is also under the control of ecdysteroids. Reduction of ecdysteroid titer in the Drosophila eye disc in vivo with the ecd^ts mutation results in irreversible furrow arrest and ommatidial disarray (Brennan et al., 1998). 20-HE also sustains furrow progression in vitro (Li and Meinertzhagen, 1995). The travelling furrow is associated with the localized expression of a reporter of ecdysteroid-dependent gene activity and of Broad-complex proteins (Brennan et al., 1998). In the Manduca eye disc, minimal levels of either ecdysone or 20-HE are required to sustain furrow progression; below the threshold level, the furrow reversibly stops during diapause (Champlin and Truman, 1998b). In the Drosophila wing disc, the sensory precursor cells in the wing margin undergo differentiation and neurite outgrowth in response to 20-HE (Schubiger and Truman, 2000). Accompanying the onset of differentiation in imaginal discs is the relaxation of heterochromatic gene silencing in imaginal tissues (Lu et al., 1998). In the eye disc, this relaxation commences at the morphogenetic furrow, and in wing, leg and eye imaginal discs, this relaxation can be induced in vitro by 20-HE (Lu et al., 1998).

Recent evidence suggests that the molecular hierarchies that transduce ecdysteroid signals in discs at these early stages are distinct from those operating in larval tissues. In the wing disc, usp represses BR-C during the mid third instar (Schubiger and Truman, 2000); however, in the whole animal, no upregulation of BR-C is seen at any time (Hall and Thummel, 1998). This suggests that genetic interactions in imaginal discs may be masked in studies of whole animals by effects in larval tissues, which predominate by mass. The shared dependence among imaginal discs of neural differentiation and heterochromatic derepression on ecdysteroid signaling suggests that the eye disc may be a good model for steroid hormone signaling in differentiating tissues.

In this study we report whole disc and mosaic mutant analysis to investigate the requirements for EcR, DHR78 and BR-C during late third instar eye development. We show that neither EcR nor DHR78 is required for normal furrow progression or ommatidial assembly. In contrast, BR-C-encoded alternative zinc finger isoforms are differentially required in these processes. We also find evidence that the pattern of BR-C expression is affected by Hh signaling.

Clones were generated using EcR^M554fs, a null allele (Bender et al., 1997) by gamma-irradiation (1000R ¹³⁷Cs) at 24-48 hours after egg-laying. Disc clones were negatively marked with both arm-lacZ (at 51A) and EcR antibody stains, to confirm proximal mitotic recombination and to control for perdurance of EcR protein. Adult clones were negatively marked with a p(w⁺) transgene at 47A. DHR78² clones were also generated by gamma-irradiation, and marked with arm-lacZ for disc clones, or a p(w⁺) transgene at 70ºC for adult clones. Although DHR78² is a null allele with a stop codon in the ligand-binding domain (Fisk and Thummel, 1998), it produces non-functional protein that is detected by anti-DHR78 antibody. Therefore, clones could not be confirmed by loss of anti-DHR78 staining; however, in 12 clones examined, no ommatidial disruption was seen. y w; EcR^M554fs/EcR^V559fs; hs-EcRB2^30.1/+ larvae were rescued through the molt to the third instar by repeated heat shocks and then maintained at 18°C for 4 more days. EcR protein is undetectable 12 hours after heat shock (Li and Bender, 2000). Hemizygous y npr1³w, y br⁵w, y rbp⁵ and y 2Bc¹ males were identified by their yellow mouthhooks and dissected. BR-C subfunction mutations used: npr1³ is null for all BR-C proteins, rbp⁵ disrupts Z1 isoforms, br⁵ is null for Z2 isoforms and 2Bc¹ is null for the Z3 isoforms (Bayer et al., 1997; Belyaeva et al., 1980; DiBello et al., 1991; Emery et al., 1994; Kiss et al., 1988; Stewart et al., 1972). BR-C, smo Mad, and Pka clones were generated by hsFlp-induced FRT recombination (Xu and Rubin, 1993). FRT18 arm-lacZ; MKRS hsFlp/TM6 males were crossed to y npr1³w FRT18/FM7, y br⁵w FRT18/FM7, y rbp⁵FRT18/FM7 and y 2Bc¹FRT18/FM7 females. smo²Mad^B1 FRT40/CyO and PKA^h2 FRT40/CyO males were crossed to w hsFlp; arm-lacZ FRT40/CyO females. Recombination was induced by 1 hour heat shocks 24-48 hours after egg-laying.

Eye discs were fixed and stained as previously described (Brennan et al., 1998). Antibodies used: rabbit anti-β-galactosidase 1:2500 (CR7001RP2; Cortex Biochem); mouse anti-β-galactosidase 1:200 (Z378A, Promega); mouse anti-EcR 11D9.6 1:100 (Koelle et al., 1991); rat anti-Elav 1:150 (Bier et al., 1988); mouse anti-BR-C core 25E9 1:100, anti-BR-C Z1 3C11 1:5 and mouse anti-BR-C Z3 9A7 1:5 (Emery et al., 1994); rabbit anti-Atonal 1:1500 (Jarman et al., 1994); and mouse anti-Cyclin B 1:100 (Knoblich and Lehner, 1993). All secondary antibodies were from Jackson Laboratories and included: HRP-goat anti-rat IgG, rhodamine-donkey anti-rat, FITC donkey anti-mouse, Cy5-donkey anti-rat, Cy5-donkey anti-mouse, FITC-goat anti-mouse, TRITC-goat anti-mouse, Cy5-goat anti-rabbit, FITC-goat anti-rabbit, and LRSC-goat anti-rabbit. In situ hybridizations were performed as previously described (Mlodzik et al., 1990; Tautz and Pfeifle, 1989). Digoxigenin-labelled antisense RNA probes specific for Z1 and Z2 zinc finger-containing BR-C isoforms were transcribed from EcoRI-linearized pSP64-Z1 and SalI-linearized pSP65-Z2 (Bayer et al., 1996a).

We examined mosaic clones and entire eye imaginal discs lacking EcR function. EcR clones at the posterior margin of the eye disc and in the center of the eye disc were normal or showed very minor defects (Fig. 1A-F). This suggests that EcR is not required for the initiation of retinal differentiation at the posterior margin, for the anterior progression of the furrow, or for proper ommatidial assembly. The mosaic discs were stained with EcR antibody (Fig. 1B,E); lack of staining in the clones demonstrated that perdurance of EcR protein was not responsible for the lack of phenotype. We note that the EcR antigen did not appear to be nuclear, as one might expect. In these same specimens, the anti-Elav antibody was capable of showing the nuclear localization of Elav – thus it appears that the fixation and detergent conditions in this experiment do allow full penetration of the IgG primary and secondary antibodies. As details of the EcR stain differ from the anti-β-galactosidase stain, it appears that this stain is not simply bleed-through from the other channel. Thus, the experiment does show a cytoplasmic location for EcR in the eye disc. We have observed some cytoplasmic localization of this antigen before, in the embryo (Koelle et al., 1991), so this is not unprecedented.

Although six EcR clones both posterior and anterior (not shown) to the furrow were examined, no clones spanning the furrow were recovered. To confirm that EcR is not required for furrow progression, and to test the possibility that EcR may be non-autonomously required for early events in eye development, we examined eye discs from EcR homozygotes that had been rescued through the molt to the third instar by a hs-EcR construct. Such animals show delayed wandering behavior and pupariation failure, including the lack of anterior spiracle eversion and larval cuticle hardening, and the persistence of larval organs (Li and Bender, 2000). However, in these discs, anterior progression of retinal differentiation proceeded normally. Moreover, discs were seen in which the furrow had traversed the entire eye field, reaching the antennal boundary (Fig. 1G). This represents a late stage of eye disc development that is normally only reached following pupariation, and suggests that furrow progression is uncoupled from metamorphic requirements for EcR function.

Initiation of furrow progression at the posterior margin occurs prior to salivary chromosome puff stage 1 (Brennan et al., 1998), suggesting that the ecdysteroid control of furrow progression might be mediated by the alternative mid-third instar hormone transduction hierarchy. To test whether DHR78, which has been proposed to be the critical hormone receptor at this early stage (Fisk and Thummel, 1998; Zelhof et al., 1995b), mediates the ecdysteroid regulation of furrow progression, we examined DHR78 mutant clones. However, they displayed normal furrow progression and ommatidial assembly (Fig. 1H,I).

Despite this demonstrated lack of requirement for known or candidate ecdysteroid receptors in transducing the hormone requirement for furrow function, previous evidence suggested that the early ecdysone response gene Broad-complex (BR-C) plays a role in this process. Broad-complex proteins are expressed at the furrow in an ecd-dependent manner, and discs null for all BR-C function display defects in furrow progression and ommatidial organization (Brennan et al., 1998).

We characterized the expression pattern in the eye disc of the different zinc finger-containing BR-C isoforms using antibody and mRNA probes (Fig. 2). An antibody that recognizes all isoforms (‘core antibody’) stained cells both in and flanking the furrow, with a peak of expression just posterior to the furrow (Fig. 2A,B). BR-C is highly expressed in the nuclei of cells in the peripodial membrane (Fig. 2C). BR-C proteins appeared to be expressed in cells until they differentiated; immediately posterior to the furrow, Elav-expressing differentiating photoreceptors express BR-C, whereas more posteriorly, expression is stronger in cone cells (Fig. 2D). Staining with an antibody that detects only Z1-containing isoforms of BR-C was restricted to posterior to the furrow (Fig. 2E), and a Z3-specific antibody stained the entire disc at very low levels, with slight upregulation in photoreceptors posterior to the furrow (Fig. 2F). Although no Z2-specific antibody is available, we reasoned that the staining anterior to the furrow seen with the core antibody might be represented by Z2-containing isoforms. This was confirmed by in situ hybridization using probes specific for the Z1 and Z2 zinc fingers (Fig. 2G,H). Z2 mRNA was detected anterior to the furrow, followed by Z1 posterior to the furrow. A switch in BR-C isoforms expression from Z2 forms to Z1 forms around the time of pupariation in imaginal discs has been described (Emery et al., 1994). Our results show that this switch from larval to prepupal forms is precocious and asynchronous in the eye disc, and is associated with the morphogenetic furrow as it traverses the disc. Although levels of BR-C expression increase dramatically during the last few hours of the third instar, corresponding to the late larval ecdysone pulse, BR-C expression still remains highest near the furrow (see time series in Fig. 2I-L).

Bar is a dominant mutation causing premature arrest of the furrow, which results in the deep anterior nick in the adult eye (compare wild type in Fig. 3A with +/Bar in Fig. 3B; Kojima et al., 1991). As Bar has the dominant effect of stopping the furrow early, one might expect loss-of-function mutations at other loci that normally act to promote furrow progression to be genetic enhancers of Bar and loss-of-function mutations in genes that normally antagonize the furrow to act as genetic suppressors of Bar. Thus, we chose to examine genetic interactions between BR-C sub loci and Bar.

We found that mutants defective for different BR-C subfunctions displayed unexpected heterogeneity in their genetic interactions with Bar, suggesting that the role of the BR-C in the regulation of furrow function might be complex. BR-C has several recessive lethal complementation groups that correspond to mutations that remove the function of all or individual zinc finger-containing isoforms subgroups. npr1 mutations lack all function, whereas rbp, br and 2Bc mutant groups correspond to the loss of Z1-, Z2-, and Z3-containing isoforms, respectively (see Table in Fig. 3; Bayer et al., 1997; Crossgrove et al., 1996; DiBello et al., 1991; Emery et al., 1994; Liu and Restifo, 1998; Sandstrom et al., 1997). Both npr1/Bar and br/Bar eyes were significantly smaller than +/Bar (Fig. 3C,D), indicating a dominant enhancement of the Bar furrow-stop phenotype, consistent with the earlier reports that the BR-C was required for furrow progression (Brennan et al., 1998). However, br/Bar eyes were smaller than npr1/Bar, suggesting that the BR-C might encode isoforms that act antagonistically during furrow progression, so that the effect of losing isoforms that positively regulate furrow progression is more severe than losing all isoforms. This idea is supported by the observation that rbp/Bar and 2Bc/Bar eyes are larger than +/Bar, suppressing the phenotype, and possibly representing furrow-antagonistic functions of rbp- or br-encoded BR-C isoforms (Fig. 3E,F).

Hemizygous males of all BR-C mutant groups survived through the third instar, and the eye discs of these males displayed defects consistent with the genetic interactions with Bar. npr1/Y discs showed ommatidial disorganization, and signs of furrow failure, including mature ommatidial clusters at the furrow (Fig. 3G, Brennan et al., 1998). br/Y discs showed a much more dramatic failure of furrow progression, as well as ommatidial disorganization (Fig. 3H). rbp/Y and 2Bc/Y discs did not show any defects (Fig. 3I,J). We note that while rbp does interact strongly with Bar it shows no eye disc defect alone – it may be that rbp is a redundant function.

Disc clones lacking individual BR-C subfunctions were generated to further characterize the mutant phenotypes (Fig. 4). Both npr1 and br clones displayed defects in ommatidial organization, including the wrong number of photoreceptors in clusters (Fig. 4A,B). rbp and 2Bc clones appeared normal (Fig. 4C,D).

To investigate the stages at which BR-C function is required for normal third instar eye development, we examined mosaic clones spanning the morphogenetic furrow (Fig. 5). In contrast to br clones, which were frequently associated with some slowing of furrow progression (Fig. 5B), none of over fifteen npr1 furrow-spanning clones was associated with any retardation of the furrow (Fig. 5A). This is consistent with the hemiygous disc phenotypes that showed a more severe furrow effect in br than npr1 mutants (Fig. 3). Neither npr1 (not shown) nor br (Fig. 5C) clones at the posterior margin of the disc showed any visible defects in furrow initiation which is consistent with the lack of requirement for ecd function for furrow initiation (Brennan et al., 1998).

Both npr1 and br clones were associated with ommatidial disarray, and were indistinguishable in this regard. Consistent with this, both npr1 and br clones were defective in the specification of the founding R8 photoreceptors, as shown by Atonal expression (Fig. 5D-I). atonal is the proneural gene for photoreceptor cells, and is first expressed in a broad equivalence group of cells anterior to the furrow, and subsequently restricted to expression in the R8 founder photoreceptors (Jarman et al., 1994). A complex network of inductive and inhibitory signals mainly involving the Notch pathway specifies the R8 cells and ensures their correct spacing (Baker et al., 1996; Cagan and Ready, 1989). In both npr1 and br clones, clusters of three Atonal-positive R8 cells were seen. This suggests a failure in the lateral inhibition mechanisms that pattern these founder cells.

The cell cycle is influenced by steroid hormones, including the insect ecdysteroids (Champlin and Truman, 1998a). Cell cycle control in the Drosophila eye disc is essential for the orderly specification of the retinal array; the synchronous cell cycle arrest in G1 in the furrow is necessary for proper cell-cell communication (Thomas et al., 1994; Wolff and Ready, 1993). Dpp, produced from cells in the furrow, is thought to regulate cell cycle synchronization in cells anterior to the furrow (Horsfield et al., 1998; Pignoni and Zipursky, 1997). Cyclin B is expressed in cells at the G2/M transition (Edgar and Lehner, 1996). Normally, Cyclin B is expressed in unsynchronized cells far anterior to the furrow, turned off just anterior to and in the furrow and reactivated in a tight band of cells posterior to the furrow (Thomas et al., 1994; Wolff and Ready, 1991). We assessed Cyclin B expression in br clones. In a long narrow br clone that extends across the furrow, there was a delay in cessation of Cyclin B expression anterior to the furrow (Fig. 5J-L). However, close examination revealed that this was a non-autonomous defect; the greatest delay occurred just adjacent to the clone, in a region that is directly anterior to the section of the mutant clone that crosses the furrow. We stained one other br clone for Cyclin B that fell in this region of the eye disc and it showed an indistinguishable phenotype. This suggests that the delay in cell cycle synchronization anterior to the furrow is a secondary defect resulting from delays in events at the furrow, possibly including delayed production of Dpp.

br mutations lack the function of Z2 isoforms of the Broad-complex, which are expressed in and anterior to the furrow in the eye disc. This corresponds to the zone in which the earliest defects in clones are noted: defects in R8 photoreceptor patterning. To confirm the hypothesis that lack of Z2 isoform expression anterior to the furrow is the cause of the defects in br and npr1 clones, expression of various BR-C isoforms was assessed in clones. As expected, in npr1 clones, expression of all BR-C proteins was eliminated (data not shown). More surprising was the finding that br clones were unreactive to an antibody that recognizes all BR-C proteins (Fig. 6A). This suggests that the expression of Z2 isoforms anterior to the furrow is required for the subsequent expression of Z1 isoforms posterior to the furrow. Evidence of such positive autoregulation between Z2 and Z1 expression has been described (Bayer et al., 1996b; Emery et al., 1994; Karim et al., 1993). rbp clones also lacked expression of Z1 isoforms, as expected (Fig. 6C), yet have no defects in retinal specification (compare Fig. 6B,D), suggesting that the defects in br clones are due to the lack of Z2 expression, and not to the lack of Z1. Although 2Bc clones do show Z3 protein (data not shown), this protein is presumably nonfunctional, as the 2Bc¹ allele has been shown to be a genetic null (Emery et al., 1994).

Although BR-C expression is dependent on highly diffusible ecdysteroids, the pattern of BR-C expression in the eye disc is spatially restricted. To test the possibility that local signals that control the rate of furrow progression also influence the spatial domain of BR-C expression, we examined clones in which cells were unable to respond to the Hh and Dpp signals that drive furrow progression, or in which the Hh signaling pathway is constitutively activated (Curtiss and Mlodzik, 2000). smoothened Mothers against dpp (smo Mad) double mutant cells, which cannot transduce Hh or Dpp signals (Alcedo et al., 1996; van den Heuvel and Ingham, 1996; Wiersdorff et al., 1996), still expressed BR-C proteins near the furrow, showing that reception of these signals is not strictly required for BR-C expression (Fig. 7A-H). Aberrant nuclear migration patterns in smo Mad cells distorted the appearance of the BR-C staining pattern: basal optical sections showed that homozygous mutant cells resembled their heterozygous neighbors in BR-C expression levels (Fig. 7D). Neural differentiation was blocked in smo Mad clones (Fig. 7B,F); BR-C expression persisted in such clones far posterior to the furrow, suggesting a link between differentiation and cessation of BR-C expression (Fig. 7G,H). Constitutive activation of the Hh signaling pathway in Pka mutant eye disc cells generates ectopic Mad-dependent circular furrows radiating outwards (Pan and Rubin, 1995; Strutt et al., 1995; Wiersdorff et al., 1996). A Pka clone examined just posterior to the endogenous furrow showed ectopic expression of Z1 isoforms anterior to the endogenous Z1 domain, surrounding the clone (Fig. 7I,J). This suggests that, although expression of BR-C proteins at the furrow does not require Hh or Dpp signaling, the timing of the switch from Z2 to Z1 isoforms is regulated, directly or indirectly, by these signals.

To assess the requirements for BR-C, EcR and DHR78 for the terminal differentiation that occurs during pupal development, we examined white-marked clones in the adult eye. br clones at the anterior margin of the adult eye showed a small nick and subtle disordering of the ommatidial array, defects that probably result from the furrow-associated defects described above (Fig. 8A). In contrast, npr1 clones showed a scar, indicative of a requirement for non-br-encoded BR-C functions in later stages of maturation (Fig. 8B). Such defects often represent failure of differentiation, and subsequent death of the affected photoreceptors. EcR clones also produced scars (Fig. 8C). This suggests that BR-C might function downstream of EcR during later stages of eye development, or that the two genes function in parallel, regulating maturation events. Either model would be consistent with the demonstrated requirements for 20-HE for later events in eye morphogenesis (Li and Meinertzhagen, 1997). Homozygous loss-of-function clones of the repressor dSin3A also leave retinal scars (Neufeld et al., 1998), suggesting that the repressive functions of EcR may be important for proper morphogenesis. DHR78 clones in the adult eye showed no defects (Fig. 8D).

We have shown that the ecdysteroid requirement for furrow progression and orderly ommatidial assembly in Drosophila eye development does not depend on the activity of two known steroid hormone receptors that are required for many of the events of metamorphosis, EcR and DHR78. Mutant EcR or DHR78 clones show no defects in ommatidial specification, and eye discs from larvae lacking detectable EcR protein display normal furrow progression, although EcR is required for later events in eye development.

In contrast, we found that BR-C proteins, whose expression is lost in discs in which the ecdysteroid titer is reduced in vivo (Brennan et al., 1998), play important roles in third instar eye imaginal disc development. We have shown that BR-C function is not required for initiation of retinal patterning at the posterior margin, but is required to maintain furrow progression, for the appropriate specification of founding R8 photoreceptors in the furrow, for proper ommatidial assembly and for later maturation events. These requirements for BR-C are differentially met by distinct isoform-encoding genetic subfunctions of the gene. The furrow-promoting activity of BR-C is supplied by the br subfunction, which encodes Z2 isoforms. A furrow-repressing activity appears to be encoded by other BR-C subfunctions, as npr1 (BR-C null) discs and clones show less severe impairment of furrow progression than br. We were not able to identify the furrow-antagonizing genetic subfunction of npr1; the lack of furrow acceleration in rbp (Z1-encoding) or 2Bc (Z3-encoding) discs and clones may be due to redundancy with each other or other ecdysone-response genes, or may suggest there are other BR-C functions not represented by br, rbp or 2Bc. br and npr1 disc tissue was similarly impaired in proper R8 specification and subsequent ommatidial organization. Both npr and br tissue in the eye lack BR-C expression, as detected with the core antibody; the reason for the more severe furrow phenotype of br than npr is not known.

We found that a switch from Z2-containing to Z1-containing BR-C isoform expression occurs at the morphogenetic furrow. This isoform switch normally occurs in imaginal discs at pupariation, suggesting that the furrow may mark a general boundary between larval and prepupal patterns of gene expression. We also found evidence that expression of Z2 isoforms anterior to the furrow may be required for subsequent Z1 expression posterior to the furrow; br clones, whose primary defect is failure of Z2 isoform expression, also failed to express Z1 isoforms. The primary failure in br and npr1 discs appears to occur in or anterior to the furrow, corresponding to the zone of Z2 expression, and is manifested as failure to properly specify R8 cells. Our result showing ectopic Z1 isoform expression near a Pka clone suggests that the switch from Z2 to Z1 isoforms may be influenced by furrow-associated expression of Hh and Dpp.

We have found that EcR does not mediate the ecdysteroid requirement for morphogenetic furrow progression. The lack of a requirement for EcR in furrow progression is surprising for several reasons. EcR is expressed in the eye imaginal disc during the time of furrow progression (Fig. 1), and is required at this time in other tissues (Li and Bender, 2000). We had previously found that a heptamer of Ecdysone Response Element (EcRE) sequences was able to drive expression of a reporter in the furrow (Brennan et al., 1998). Evidence suggesting that only the EcR/Usp dimer is able to activate transcription from EcREs includes that EcRE in Schneider cells is bound by a single complex that is destroyed by anti-EcR antibodies, EcR/Usp binds EcRE in the presence or absence of hormone, and transfection of EcR confers 20-HE responsiveness to cells carrying an EcRE reporter (Koelle et al., 1991; Yao et al., 1992). The lack of EcR requirement for furrow progression suggests either that a contribution by EcR to furrow regulation is masked by redundant contributions by another ecdysteroid receptor, or that an unidentified receptor that can also activate transcription from EcREs exclusively mediates the hormone regulation of furrow progression. Redundancy seems unlikely because loss of EcR function has strong phenotypes in almost all tissues where it is expected to function, including at the larval molts and at pupariation (Li and Bender, 2000). One candidate steroid receptor to mediate the hormonal regulation of furrow progression is DHR96; it is expressed during the mid-late third instar, and is able to bind EcREs in gel-shift assays (Fisk and Thummel, 1995).

The EcR clone phenotypes in both discs and adult eyes are distinct from those of usp, which are reported to produce furrow acceleration in the disc and abnormal rhabdomeres, but not scars, in the eye (Oro et al., 1992; Zelhof et al., 1995a). These results, along with earlier findings of distinct embryonic and early metamorphic phenotypes (Bender et al., 1997; Hall and Thummel, 1998; Li and Bender, 2000; Perrimon et al., 1985), suggest that EcR and Usp each have functions independent of the other.

Does the receptor that transduces the ecdysteroid requirement for furrow progression include Usp? usp retinal clones are reported to show variable degrees of furrow acceleration that correlate to some degree with stage of development; clones spanning furrows near the posterior margin show a more marked degree of furrow advancement than those examined at later stages (Zelhof et al., 1997). The ecdysteroid titer increases dramatically during the time that the furrow traverses the eye disc; it is possible that Usp represses furrow progression at the low titers prevailing near the time of furrow initiation, but that this repression is relieved by increasing hormone concentrations over the next several hours. This would be somewhat analogous with the situation in the wing disc, where Usp repression of neural development is relieved by 20-HE (Schubiger and Truman, 2000); however, Usp repression of furrow progression at low to moderate hormone levels could only be partial, because furrow progression does occur.

We previously reported that BR-C expression in the eye disc is dependent on ecd function, suggesting that it is regulated by ecdysteroid titer (Brennan et al., 1998). While BR-C expression in the eye disc is unlikely to be regulated by EcR, Usp may play a role. BR-C and usp disc clone phenotypes are opposite in that the former shows some furrow retardation while the latter shows a slight acceleration (Zelhof et al., 1997). Possible Usp repression of BR-C, or Usp regulation of BR-C isoform ratio could explain these differences. BR-C and usp eye clones also resemble each other in that both show ommatidial disarray. This similarity underscores the importance of orderly progression of the furrow for subsequent events, and may reflect an essential hormonal input that regulates timing of furrow progression (Brennan and Moses, 2000).

Local events appear to regulate BR-C expression at a number of levels. Two regulatory inputs into the switch from Z2 to Z1 expression at the furrow have been suggested: (1) the lack of Z1 expression in br clones suggests that prior expression of Z2 isoforms anterior to the furrow is required for subsequent expression of Z1 isoforms; (2) the ectopic Z1 expression seen surrounding a PKA clone suggests that Hh-dependent signaling at the furrow also promotes the switch to Z1 isoforms. Cessation of BR-C expression in eye disc cells correlates with cell-fate determination; photoreceptors diminish expression levels earlier than cone cells, and failure to differentiate in smo Mad cells is associated with prolonged expression of BR-C proteins. BR-C proteins are thought to confer competence to respond to ecdysteroid signals (Karim et al., 1993); in the asynchronously developing eye disc, it may be essential to spatially restrict such competence.

How might BR-C proteins exert their effects on eye development? Although ecdysteroids regulate ommatidial patterning at many stages of eye development in Manduca (Champlin and Truman, 1998b), our evidence suggests that the critical stage at which early eye development in Drosophila is regulated by Broad-complex is just anterior to and in the furrow. This corresponds to the site of Z2 expression and the zone at which the earliest defects are detected, in R8 specification. No transcriptional targets of BR-C have been identified in the eye; the aberrant specification of R8 founders suggests that targets may include elements of the Notch signaling pathway. BR-C regulation of such targets is likely to be complex, with different isoforms acting antagonistically.

We thank the Iowa Developmental Studies Hybridoma Bank, Greg Guild, David Hogness, Andrew Jarman, Patrick O’Farrell, David Strutt, Carl Thummel, Jessica Treisman and Laurie von Kalm for antibodies and flies; Lucy Cherbas, Nora Ghbeish, Margrit Schubiger, Carl Thummel, Chih Cheng Tsai and Kevin White for sharing results before publication; and Laurie von Kalm and Kathryn Anderson for comments on the manuscript. This research was supported by an NSF graduate fellowship to C. A. B. and NIH grants (GM53681) to M. B. and (EY09299) to K. M.