Barbu: an E(spl) m4/mα-related gene that antagonizes Notch signaling and is required for the establishment of ommatidial polarity

Notch signaling has been shown to regulate a wide variety of cell fate decisions during Drosophila development. These developmental functions of Notch can be grouped into three broad categories, namely lateral inhibition processes, lineage decisions and inductive processes at boundaries of cell populations as well as between individual cells in wing and eye imaginal discs (reviewed in Bray, 1998). Lateral inhibition processes, which have been studied most extensively, involve Notch signaling among developmentally equivalent groups of cells and lead to the selection of individual cells within a group as specific precursors (reviewed in Simpson, 1997). For example, Notch-mediated lateral inhibition processes within proneural cell clusters in the embryonic neuroectoderm and imaginal discs result in the selection of single cells as neuroblasts and sensory organ precursors, respectively. In the eye imaginal discs, analogous processes result in the selection of R8 photoreceptor precursors (Baker et al., 1996). In the somatic mesoderm, Notch signaling serves to select single cells within groups of equivalent mesodermal cells (‘promuscular clusters’) as muscle progenitors and, in the cardiogenic mesoderm, related processes contribute to the selection of pericardial cell progenitors and presumably also the progenitors of cardioblasts (Corbin et al., 1991; Hartenstein et al., 1992; Bate et al., 1993; Carmena et al., 1995). In contrast to lateral inhibition, Notch activity during lineage decisions and boundary formation involves signaling between cells with different properties, which results in biased responses. During asymmetric cell divisions, this bias is established by the unequal segregation of the Numb protein into only one of the two daughter cells, which in turn leads to inhibition of Notch signaling in the Numb-containing sibling (reviewed in Knoblich, 1997). Well-studied examples for the role of Notch and numb in asymmetric cell fate decisions include the specification of sibling neuronal cell fates during CNS development (Spana and Doe, 1996; Skeath and Doe, 1998), specification of neuron, glia, shaft and socket cells during successive divisions of sensory organ precursor cells (Frise et al., 1996; Guo et al., 1996), and the differential specification of the daughter cells of muscle progenitors as founders of two different body wall muscles (Ruiz-Gomez and Bate, 1997; Carmena et al., 1998b; Park et al., 1998).

The Notch protein acts as a transmembrane receptor that receives signals from neighboring cells via its ligands Delta and Serrate (reviewed in Kimble and Simpson, 1997; Greenwald, 1998). Ligand-binding appears to trigger a proteolytic processing event, which depends on the activity of Presenilin and allows the intracellular domain of Notch to translocate into the nucleus (Kidd et al., 1998; Lecourtois and Schweisguth, 1998; Struhl and Adachi, 1998; Struhl and Greenwald, 1999; Ye et al., 1999). Most of the known responses to Notch require its association with the Su(H) protein, which confers DNA-binding activity to the protein complex (Fortini and Artavanis-Tsakonas, 1994; Lecourtois and Schweisguth, 1995; Bailey and Posakony, 1995). To date, only a few direct target genes of the Notch signaling cascade have been identified. The best-studied examples are the bHLH encoding genes of the Enhancer of split (E(spl)) complex, which are turned on in the cells within proneural clusters in which the Notch pathway is active (Jennings et al., 1994). The E(spl) bHLH proteins function as transcriptional repressors and execute lateral inhibition by repressing proneural gene expression in these cells (Jennings et al., 1994; Oellers et al., 1994; Heitzler et al., 1996a). Thus, proneural gene expression is maintained only in the one particular cell within each proneural cluster in which the Notch pathway is not active, thereby defining it as a presumptive neuroblast or sensory organ precursor. Another known transcriptional target is vestigial (vg), which is activated by Notch/Su(H) at the dorsoventral boundary of the wing disc (Kim et al., 1996).

As in many other signaling cascades, negative regulators play important roles in modulating the activity of the Notch signaling pathway. Fringe is an example of an extracellular protein that specifically inhibits Notch from being activated by the Serrate ligand, while it potentiates Notch activation by Delta (reviewed in Irvine and Vogt, 1997). Intracellularly, inhibitors of Notch signaling have been found to act at various levels of the cascade. For example, inhibiting effects from Wingless (Wg) appear to be mediated mainly by the Wg effector protein Dishevelled (Dsh), which can bind to the intracellular domain of Notch (Axelrod et al., 1996). Similarly, the inhibitory function of Numb during asymmetric cell divisions appears to involve interactions between Numb and intracellular Notch (Guo et al., 1996). The Hairless (H) protein seems to execute its antagonistic activity in the nucleus, presumably by binding directly to Su(H) and inhibiting activation of Notch target genes (Brou et al., 1994; Bang et al., 1995; Maier et al., 1997). In the wing disc, nubbin, which encodes a POU-domain transcription factor, may negatively regulate Notch target genes as a transcriptional repressor (Neumann and Cohen, 1998). The phenotypes of gain-of- function alleles of the genes Bearded (Brd), E(spl) m4 and E(spl) mα indicate that their gene products, small proteins with unknown biochemical and intracellular functions, can also antagonize Notch signals, but the lack of loss-of-function phenotypes has hampered the analysis of their role during normal development (Leviten and Posakony, 1996; Leviten et al., 1997; Apidianakis et al., 1999).

Here we present an analysis of a novel gene, Barbu (Bbu), and show that its gene product can antagonize Notch- dependent processes. We show that ectopically expressed Barbu can interfere with lateral inhibition processes in the neuroectoderm and mesoderm during early embryogenesis, as well as during sensory organ formation in imaginal discs. The observed transcriptional activation of Bbu by activated Notch and the presence of Su(H) consensus binding sites in its genomic upstream sequences suggest that Bbu itself may be a transcriptional target of the Notch signaling cascade. In addition to its expression in the embryonic ectoderm and mesoderm, Bbu is also expressed on either side of the morphogenetic furrow in eye imaginal discs. Loss-of-function alleles of Bbu cause lethality and adult escapers homozygous for a partial loss-of-function allele display disruptions in the polarity of the ommatidia and bristles of the notum, which may be due to abnormally increased Notch activity during sense organ and photoreceptor cell specification. These data and the structural similarities among the small proteins encoded by Barbu, m4 and mα from the E(spl) complex, and Bearded suggest that members of this protein family have related and partially redundant roles in a negative feed-back loop during Notch signaling processes.

Drosophila strains were raised on standard cornmeal-yeast agar medium at 25°C or at room temperature. The null allele Su(H)⁸ was as described by Schweisguth and Posakony (1992). The deficiencies uncovering the Bbu and frizzled loci (Df(3L)D-5rev12 and Df(3L)fz- GF3b) are described in Lindsley and Zimm (1992). The enhancer trap line EP(3)0487 was obtained from the BDGP and is described in the BDGP database.

The Bbu alleles, Bbu⁸, Bbu⁹⁹ and Bbu¹⁰⁵, were isolated in an F² screen using the P-element insertion line EP(3)0487. Dysgenic males of the genotype yw; EP(3)0487/Δ(2-3),Sb were crossed to w/w;TM3Sb/TM6Hu females. Non-Sb F² progeny with white eyes were collected and crossed with Df(3L)D-5rev12/TM2Ubx to determine the lethality. Each candidate was retested over Df(3L)fz- GF3b /TM6B to exclude any mutations affecting the frizzled locus (70E), and over the original viable EP(3)0487 line to test for any second site mutation. All three alleles of Bbu were balanced over TM3Sb p[hg-lacZ] to identify and examine homozygous embryos.

The GAL4 lines used in the overexpression of Bbu or Notch^intra (a gift from G. Struhl) were pnr^MD237-GAL4 (Calleja et al., 1996; Heitzler et al., 1996b), GMR-GAL4 (Freeman, 1996), twist-GAL4 (a gift from M. Akam), 24B-GAL4 (Brand and Perrimon, 1993) and Kr- GAL4 (from P. Gergen through M. Leptin).

For mosaic analysis, clones of Bbu allele were induced by the FLP/FRT method (Xu and Rubin, 1993). The Bbu⁹⁹ allele was recombined onto a chromosome carrying the FRT cassette at position 80B (yw; P{w⁺=πMyc}75C, P{ry⁺=neoFRT}80B). To generate Bbu⁹⁹ mutant clones in the eye in a Minute background, yw; Bbu⁹⁹, πM75, FRT80/TM6B females were crossed with yw, P{ry⁺=ey-FLP1}; P{w⁺=armLacZ}69B, M(3)67C, P{ry⁺=neoFRT}80B/TM6B males (ey-FLP1 was a generous gift from B. Dickson through J. Treisman). Adult flies with clones in the eye were identified by using the Tubby marker.

The 1 kb Bbu cDNA was used to screen an EMBL4 Canton-S genomic library (Clontech Laboratories, Inc.), using standard molecular biology methods (Sambrook et al., 1989). The identification of three λ phage clones was confirmed by Southern blot hybridization and restriction mapping. The two EcoRI fragments 2 kb and 3.5 kb long, around the Bbu unit were sequenced by the Biotechnology Center of the Utah University. Database searches using the BLAST program (Altschul et al., 1990) and secondary structure analysis of the Bbu protein were done with the MacVector 6.0.1 software package.

Flies were first dehydrated through a graded series of ethanol. Subsequent critical point drying and scanning electron microsopy were done at the Analytical Imaging Facility at the Albert Einstein College of Medicine as described by Nguyen et al. (1997). Adult eyes were fixed and sectioned as described by Tomlinson and Ready (1987).

The full-length Bbu cDNA was derived from the EcoRI-SalI fragment of pGAD-Bbu (see below) and subcloned into EcoRI-XhoI sites of the pUAST vector (Brand and Perrimon, 1993). The EcoRI-XhoI Bbu fragment was generated by polymerase chain reaction (20 cycles in the presence of high fidelity Taq (Roche Molecular Biochemicals, Mannheim) from the Bbu cDNA using primers spanning the initiation and termination sites. The resulting UAS constructs were introduced into yw⁻embryos to generate transgenic flies by standard injection methods (Rubin and Spradling, 1982). Several independent transformants were collected and mapped, and two lines for each UAS constructs were made homozygous to be used in the experiments.

In situ hybridizations were done for embryos as described in Azpiazu and Frasch (1993), except that hybridizations were carried out at 55°C with DIG-labeled T7 RNA transcripts of the 1 kb cDNA insert (Roche Molecular Biochemicals, Mannheim). For fluorescent in situ hybridizations, the following modifications were made. Instead of the regular anti-DIG antibody coupled to alkaline phosphatase, a sheep anti-DIG antibody (Roche Molecular Biochemicals, Mannheim) was used. A secondary donkey anti-sheep biotinylated antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) was used in conjunction with the VectaStain ABC components (Vector Laboratories, Burlingame, CA). Finally Renaissance Tyramide Signal Amplification (NEN, Boston MA) was used to perform the color reaction with fluorescein. Double stainings and immunostainings were performed as described in Knirr et al. (1999). For in situ hybridizations of imaginal discs, a 20 minute postfixation with 4% formaldehyde, 0.1% Triton X-100, 0.1% sodium deoxycholate in PBS was performed after the initial fixation. The antibodies used in this study have been described previously: rabbit anti-β-galactosidase antibody (1:3000; Cappel), rat anti-Lethal of scute antibody (1:800; Carmena et al., 1995), mouse anti-Achaete antibody (1:5; Skeath and Carroll, 1991), mouse anti-Atonal antibody (1:5000; a gift from J. Treisman and Y.-N. Jan), rabbit anti-Tinman antibody (1:1500; Xu et al., 1998), rabbit anti-MEF2 antibody (1:1500; Bour et al., 1995), rabbit anti-Slouch antibody (1:10; Dohrmann et al., 1990), mouse anti-Ladybird antibody (1:3; Jagla et al., 1998), rabbit anti-Eve antibody (1:3000; Frasch et al., 1987), guinea pig anti-Eve antibody (1:300; Kosman et al., 1998), rabbit anti-Zfh-1 antibody (1:3000; gift from R. Lehmann). A Zeiss Axiophot and a Leica confocal microscope were used to analyze the stainings. Cross-sections of whole-mount embryos were done as described by Azpiazu and Frasch (1993).

For analyzing the β-galactosidase staining in A101, which carries an insertion in neuralized (Boulianne et al., 1991), wild-type or Bbu- overexpressing prepupae were selected from the cross between UAS- Bbu/UAS-Bbu; A101/TM6Tb and pnr-GAL4/TM6Tb by the absence of the dominant Tubby marker and the wing imaginal discs were dissected and treated as described by Xu and Rubin (1993).

An 800 bp PCR fragment (oligos: TCGTGGATCCCACGGGAAT and CCTGTCGACGTTTAGGCCTGA) encoding amino acids 1-158 was amplified with Bbu cDNA as a template and subcloned into the pET30c expression vector (Novagen Inc.). The bacterially induced protein was purified on a Ni-NTA column following standard conditions suggested by the manufacturer (Qiagen Inc.). After dialysis, Bbu fusion protein was used to immunize a guinea pig following the protocol of Covance Research Product Inc. Anti-Bbu sera were affinity purified with Bbu recombinant proteins covalently coupled to a CN-Br column (Frasch et al., 1987) and subsequently the antibodies were preabsorbed with fixed 0-2 hour old embryos.

Full-length tinman was cloned in-frame to the GAL4 DNA-binding domain encoded in the pGBT9 vector (Stratagene) and used as bait to screen a 3-12 hour embryonic cDNA library fused to the GAL4- activating domain in the pGADNOT vector (a gift from R. Mann). Approximately 750,000 colonies were screened as described by Bartel et al. (1993). The nucleotide sequences of positive clones were determined by double-strand sequencing as described in Sambrook et al. (1989). Three clones were identified that were homologous to E(spl) m4 and mα, and called Bbu cDNAs.

In vitro binding assays were performed essentially as described previously (Shen et al., 1997). Briefly, the expression of Bbu/GST- fusion protein or GST were obtained with the pGEX-5X1 vector (Pharmacia) and expressed following the Pharmacia protocol. The GST or GST-fusion proteins were then bound to glutathion-Agarose beads (Sigma) in PBS and then washed five times in 0.1% Tween- 20/PBS. The beads were stored at 4°C and used as a 50% suspension. Both ³⁵S-labeled Tinman or Luciferase proteins were expressed using the TNT-coupled lysate system (Promega). For the in vitro binding assay, 30 μl of the 50% suspension of beads containing approx. 3 μg of GST-Bbu protein were incubated at 4°C for 30 minutes with 15 μl lysate containing ³⁵S-labeled proteins. To the protein-bead mixtures, 150 μl 0.5% NP-40/PBS were added and following an incubation at 4°C for 30 minutes, the mixtures were washed five times 30 minutes in 0.5% Tween-20/PBS. The protein- bead mixtures were loaded on a 10% SDS-PAGE gel and then analyzed by autoradiography.

In a yeast two-hybrid screen for gene products interacting with the homeodomain protein Tinman (see Materials and Methods; Azpiazu and Frasch, 1993; Bodmer, 1993), we isolated three cDNA clones encoding a transcript with a dynamic pattern of expression in the early mesoderm and ectoderm during embryogenesis. Sequence analysis of these cDNA clones revealed a long ORF inframe with the yeast GAL4-activating domain. A BLAST database search with this sequence identified an EST clone, LD05688, which contains overlapping sequences that extended further towards the 5′ end. Analysis of the combined sequences from this gene, which we named Barbu (Bbu), revealed a coding region of 476 nucleotides (ntd.)/158 amino acids (aa.), a 59 nucleotide 5′-untranslated region (UTR) and a 418 nucleotide 3′-UTR containing a polyadenylation signal (AATAA) at position 869 (relative to the A of the start codon; see GenBank accession no. AF132987). The deduced amino acid sequence displayed significant homology with two other Drosophila proteins, M4 and Mα of the Enhancer of split Complex (E(spl)-C) (Klämbt et al., 1989; Wurmbach et al., 1999). These small proteins were predicted to possess an α-helix with basic amphipatic structure in their N-terminal region, which appears also to be present in Barbu (Fig. 1A; Leviten et al., 1997; Wurmbach et al., 1999; see Materials and Methods). Sequence comparisons among these three proteins show that, in addition to the conserved domain in the N-terminal region, their C-terminal region (aa. 116-158) defines a second domain with significant homology (49% identities and 56% similarities; Fig. 1A). In addition to the sequence conservation found among the coding regions of E(spl) m4/mα and Bbu, we identified several highly conserved motifs in the 3′-UTR sequence of Bbu, two ‘Brd boxes’ (+593 and +895), two ‘GY boxes’ (+825 and +881) and one ‘K box’ (+775) (Fig. 1B,C; Leviten et al., 1997; Lai and Posakony, 1997; Lai et al., 1998; Wurmbach et al., 1999). These motifs were initially identified in the 3′-UTR of transcripts from several other genes, including genes from the E(spl)-C, some proneural genes and Bearded, and have been proposed to confer transcript instability (Lai and Posakony, 1997; Leviten et al., 1997; Lai et al., 1998).

In preparation for a genetic analysis of Bbu, we isolated three overlapping λ phage clones covering ∼25 kb of the Bbu locus from a genomic library (Fig. 2A). A search for other transcription units around the Bbu locus localized the Bearded gene (Leviten et al., 1997) in close vicinity, which maps downstream of Bbu and is transcribed in the same direction (Fig. 2A). Although Leviten et al. (1997) reported that E(spl) M4 shares a weak homology with Bearded, our sequence analysis failed to detect any significant sequence similarities between Barbu and Bearded, except in the 3′-UTRs of their transcripts (see above). Based upon the map position of Bearded, Bbu is located at 71A1-2 on the third chromosome. A BLAST search of the BDGP data base containing P-element flanking sequences identified an enhancer trap insertion, EP(3)0487, whose insertion site corresponded to a position at 128 bp upstream of the first nucleotide of the EST cDNA and 192 bp upstream of the start of the Bbu ORF (Fig. 2A). Because the EP(3)0487 line is homozygous viable and displays no visible defects, we performed a P-element excision screen to delete Bbu encoding sequences. 20 lethal or semilethal excision derivatives failed to complement Df(3L)D-5rev12, which uncovers the Bbu locus. A number of these mutations were shown to affect the Bbu region by Southern blot analysis and by sequencing of DNA isolated from several of the obtained alleles. Three mutations, Bbu⁸, Bbu⁹⁹ and Bbu¹⁰⁵, were found to affect exclusively the Bbu unit. Bbu⁹⁹ has a large deletion removing only the Bbu unit (Fig. 2B), a result confirmed by in situ hybridization which failed to detect any Bbu transcripts in homozygous embryos for this mutation. The analysis of the Bbu⁸ and Bbu¹⁰⁵ mutations showed that both affect the promoter region of Bbu. Bbu⁸ has a 1.7 kb deletion and Bbu¹⁰⁵ has additions and exchanges of several nucleotides, which have occurred just upstream of the putative TATA box (at +72 bp with respect to the original P insertion; Fig. 2B). While Bbu⁹⁹ is embryonic lethal without producing any gross abnormalities, both Bbu⁸ and Bbu¹⁰⁵ display pupal lethality of homozygous animals (see below), and in situ hybridizations detected reduced levels of Bbu transcripts in homozygous embryos (data not shown).

The temporal and spatial pattern of Bbu mRNA was determined by northern blot and whole-mount in situ hybridizations. A single 1.3 kb Bbu transcript appears at 2-4 hours of development with a peak expression at 4-8 hours after fertilization (Fig. 2C). Bbu mRNA expression decreases dramatically following late embryonic stages until its expression becomes barely detectable after the first instar larval stage (Fig. 2C). Given the length of Bbu mRNA and taking account for a poly(A) tail, it is possible that the 5′-UTR extends further upstream than our longest cDNA.

Whole-mount in situ hybridization using the Bbu cDNA insert as a probe first detected Bbu expression in the anterior- most and in central regions of early blastoderm embryos (Fig. 3A). At the end of the blastoderm stage, Bbu is uniformly expressed, but expression is excluded from the ventralmost part of the embryo, which corresponds to the presumptive mesoderm domain (Fig. 3B,K). Because the zinc finger protein Snail is known to act as a repressor within the mesodermal territory (Kosman et al., 1991; Leptin, 1991), we tested whether Bbu expression is negatively regulated by snail. Wild-type and mutant embryos were double stained to visualize Snail protein and Bbu transcript. In wild-type blastoderm embryos, the expression patterns of the two genes are exactly complementary (Fig. 3L) and, in snail mutant embryos of the same stage, Bbu RNA expands throughout the entire ventral region (Fig. 3M; compare with K). These data indicate that Bbu expression is normally repressed by Snail in the presumptive mesoderm. By contrast, shortly after gastrulation and until the end of the germband elongation, Bbu expression is seen throughout the mesoderm (Fig. 3C). In order to confirm the mesodermal expression of Bbu and test for co-expression with Tinman, we performed fluorescent double stainings with Tinman antibodies and Bbu antisense RNA probes. These experiments showed that Bbu and Tin co-localize during stages 8 and 9 in the entire mesoderm (data not shown) and during stage 10 in most of the dorsal mesodermal cells (Fig. 3D). During late germband elongation, when the specification of mesodermal tissues and muscle progenitors initiates, Bbu expression assumes a segmental pattern in the ventral mesoderm (Fig. 3E,F). At the same time, Bbu expression in the ectoderm becomes patchy and expression levels increase in the tracheal pit territories (Fig. 3G). Bbu mRNA expression in the ectodermal layer was further studied using confocal microscopy. During germband elongation and at the beginning of neuroblast segregation, Bbu mRNA is excluded from one or two row(s) of ventral ectodermal and mesectodermal cells in each segment (Fig. 3H). Slightly later, rapid and complex changes in the Bbu expression pattern occur, during which Bbu expression becomes transiently excluded from segmental, bilaterally symmetric patches of cells (see below). Furthermore, at the beginning of germband retraction when most neuroblasts have segregated, Bbu becomes more weakly expressed in the ventral region of the ectoderm as compared to its dorsal expression (Fig. 3J). During germband retraction, Bbu expression rapidly decays and, after dorsal closure, is only seen in the pair of lymph glands close to the anteriormost portion of the dorsal vessel (Fig. 3I).

Immunocytochemical stainings of embryos with antibodies against bacterially expressed Bbu demonstrated that the spatial and temporal expression of Bbu protein mimics the pattern observed for Bbu mRNA (Fig. 4A-C, compare with Fig. 3B,E,H). Moreover, counterstaining with propidium iodide showed that the Bbu protein is predominantly localized to the cytoplasm and excluded from the nuclei (Fig. 4A,A′,C,C′). In the ectoderm, the bulk of Bbu protein appears to be present in the apical compartment of the cytoplasm (Fig. 4C,C′). We also detect some punctate signals within the nuclei, but because of the presence of unspecific signals (e.g., Fig. 4A, bottom part) we cannot be certain at present whether these reflect endogenous Bbu protein.

Since the above observations indicated that the Bbu expression pattern might be related to the process of neuroblast segregation, it was pertinent to examine the relationship of its ectodermal distribution with neurogenic processes more carefully. We double labeled embryos to compare the Bbu mRNA pattern with those of two proneural gene products, Achaete (Ac) and Lethal of scute (L’sc) (Alonso and Cabrera, 1988; Skeath and Carroll, 1991; Carmena et al., 1995). These stainings showed that, in the neuroectoderm, Bbu expression is excluded from the proneural clusters and neuroblasts in which these proneural proteins are expressed (Fig. 4D,E, arrows). In clusters from which single cells have been selected to maintain proneural gene expression, all cells except for the presumptive neuroblast express Barbu mRNA (Fig. 4D, arrowhead). These and the above observations raised the possibility that the Notch signaling pathway, while repressing proneural genes, may activate Bbu in the presumptive non-neuroblast cells. To establish whether active Notch controls Bbu expression, the Gal4/UAS system (Brand and Perrimon, 1993) was used to generate ectopic expression of activated Notch in defined areas along the anteroposterior (AP) axis. The Kr-Gal4/UAS-N^intra combination of constructs served to misexpress the intracellular domain of Notch in the central region of the embryo, which will then translocate to the nucleus and, together with Suppressor of Hairless protein (Su(H)), activate target genes (Struhl and Adachi, 1998; Lecourtois and Schweisguth, 1998). Indeed, we detect overexpression of Bbu mRNA in the Krüppel domain of these embryos, where Bbu mRNA invades most of the empty areas that are normally occupied by proneural clusters (Fig. 4G, compare with Fig. 4F). Lethal of scute protein is largely absent in the same region, thus demonstrating that the activated Notch signaling pathway has opposite effects on the expression of Bbu and proneural genes. The presence of three putative Su(H) binding sites in the presumed Bbu promoter sequence (see GenBank accession no. AF175684) indicates that activation of Bbu by Notch is possibly direct. One of these sequences ([−791]CGTGGG- AAA) matches exactly the previously reported consensus sequence for Su(H) sites [(C/T)GTG(G/A)GAA(C/A)] (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995), while the other two (at −80 and −118 with respect to +1 of the cDNA, respectively) differ in one position (A instead of T in position 3). Thus, the control of Bbu expression with respect to its dynamic ‘filling in’ of proneural clusters and exclusion from prospective neuroblasts seems to be a direct result of its activation by the Notch signaling pathway in the neuroectoderm.

Previous studies have shown that embryos mutant for neurogenic genes have expanded domains of muscle precursor cells (Corbin et al., 1991; Bate et al., 1993). Since the expression of Bbu suggested a possible relationship of Bbu regulation and function with Notch signaling, we examined whether mesodermally expressed Bbu may have a role during these mesodermal lateral inhibition processes. We overexpressed Bbu with the UAS-GAL4 system exclusively in the mesodermal germ layer just prior to and during this specification step, using a combination of two drivers, twist- and 24B-GAL4 (Brand and Perrimon, 1993; Baylies et al., 1995). The effects of Bbu overexpression were monitored with various markers for specific muscle and heart progenitors, including Even-skipped (Eve), MEF2, Ladybird (Lb), Slouch (Slou or S59) and Lethal of scute (L’sc) (Frasch et al., 1987; Bour et al., 1995; Carmena et al., 1995; Jagla et al., 1998; Knirr et al., 1999).

In stage 11 wild-type embryos, each hemisegment contains a pair of Eve-expressing pericardial precursors (pc) and a single Eve-positive dorsal muscle founder (dmf), which gives rise to muscle 1 (Fig. 5A; Frasch et al., 1987; Carmena et al., 1998b). By contrast, upon overexpression of Bbu in the mesoderm, the number of Eve-expressing mesodermal cells is significantly increased at the same stage (Fig. 5B). Carmena et al. (1998b) have previously shown that Notch signaling acts to restrict the mesodermal expression of both eve and the proneural gene lethal of scute (l’sc) to a single progenitor within a ‘promuscular’ cluster of eve- and l’sc-expressing cells. Double staining of stage 11 embryos with antibodies against Eve and L’sc showed that, upon Bbu overexpression in the mesoderm, there is a notable de-repression of both genes within these clusters. Whereas in a wild-type background, Eve and L’sc are restricted to a single cell, the progenitor P15 (Fig. 5C; Carmena et al., 1998b), ectopic Bbu expression causes continued expression of the two gene products in several cells of the corresponding clusters (Fig. 5D). The increased number of mesodermal Eve cells persists until late stages of embryogenesis. Closer inspection of late stage embryos suggests that Bbu overexpression produces supernumerary muscle 1 founder cells and, as a consequence, larger or duplicated muscle 1 syncytia (Fig. 5F, green arrowhead). Simultaneous staining for Zfh-1, which is expressed in pericardial cells and adult muscle progenitors at this stage, indicates that the Eve/Zfh-1 double-positive pericardial cells are not affected (Fig. 5F, yellow arrowhead; Lai et al., 1991; Su et al., 1999). However, the number of adult muscle precursors is significantly increased (Fig. 5F, red arrowhead). Similar increases in the number of muscle progenitors are observed in embryos with reduced Notch signaling activities, although these embryos additionally lack Eve pericardial cells (Hartenstein et al., 1992; Zaffran et al., 1995; Carmena et al., 1998b; Ruiz-Gomez et al., 1997).

Overexpression of Bbu affects also the development of lateral and ventral muscle progenitor muscles. For example, the number of Slouch (S59)-expressing progenitors in the mesodermal clusters I and II is increased (Fig. 5H compare to 5G). Stainings for Lb, which is a marker for muscle 8 (Jagla et al., 1998), show that overexpression of Bbu causes increases in the number of progenitors and syncytial nuclei of muscle 8 (Fig. 5J, compare with 5I; and data not shown).

Mesodermal overexpression of Bbu also causes increases in the number of cardioblasts. Whereas stage 16 embryos have a double row of MEF2-expressing cardioblasts along the dorsal midline (Fig. 5K), Bbu overexpression causes the formation of many supernumerary cardioblasts in the dorsal vessel (heart; Fig. 5L). This effect resembles mild neurogenic phenotypes in cardioblast development (Hartenstein et al., 1992). Together, these results suggest that overexpression of Bbu in the mesoderm antagonizes Notch signaling during lateral inhibition processes but probably not during asymmetric cell divisions in which Notch/numb-dependent activities specify mesodermal cell fates (Corbin et al., 1991; Bate et al., 1993; Baker and Schubiger, 1996; Ruiz-Gomez and Bate, 1997; Carmena et al., 1998b; Park et al., 1998).

To further examine the interactions between Notch signaling and Bbu activity, we ectopically expressed Bbu in wing imaginal discs during sensory organ development. The spatial distribution of the thoracic sensory bristles is regulated by Notch signaling, which delimits the expression of proneural genes to defined numbers of sensory organ precursor cells (for review, see Simpson, 1997). As shown in Fig. 6B,E, overexpression of Bbu protein using pnr-GAL4 as a driver leads to an excess of dorsocentral macrochaetes and microchaetes in the notum, as well as additional macrochaetes in the dorsal region of the head (compare with Fig. 6A,D). Even more dramatic increases in the number of bristles are observed when a UAS-Bbu Δ3′-UTR transgene is used in combination with pnr-GAL4 (Fig. 6C,F compare with 6B,E), which is consistent with the previously proposed role of the above described 3′-UTR motifs in lowering mRNA accumulation post-transcriptionally (Lai and Posakony, 1997; Lai et al., 1998). Duplication of bristles is associated with an increase or duplication of the socket structures (Fig. 6E).

To show that the observed effects on bristle patterning are due to excessive sensory organ precursor formation, we monitored lacZ reporter gene expression of the line A101, which carries an enhancer trap insertion in the neuralized gene and expresses lacZ in all sensory organ precursors (Fig. 6G; Boulianne et al., 1991). Upon overexpression of Bbu transgenes via pnr-GAL4, we indeed observe segregation of an excess number of A101-β-gal- expressing sensory cell precursors in the wing disc areas with pnr activity, which will give rise to anterior and posterior scutellar macrochaetes (Fig. 6H,I; Calleja et al., 1996; Heitzler et al., 1996b). Since similar effects can be observed in the absence of positively acting components of the Notch pathway, these results indicate that Bbu can act as an antagonist of Notch signaling during lateral inhibition processes in imaginal disc development (Hartenstein and Posakony, 1989; Heitzler and Simpson, 1991).

Although the presumed null allele Bbu⁹⁹ confers embryonic lethality, we have not yet been able to detect any specific abnormalities in ectodermally or mesodermally derived tissues in homozygous embryos for this mutation. However, the phenotypic analysis of the hypomorphic alleles Bbu⁸ and Bbu¹⁰⁵ proved to be informative. Both Bbu⁸ and Bbu¹⁰⁵ confer pupal lethality and pharate adults homozygous for these mutations fail to undergo head eversion (Fig. 7A). To test for genetic interactions between Bbu and components of the Notch pathway, we examined whether lowering the dosage of Su(H) can modify this phenotype. As shown in Fig. 7B, pharate adults of the genotype Su(H)⁸/+;Bbu⁸/Bbu⁸, although still lethal, display almost completely everted head structures. This phenotypic suppression is reciprocal, since lowering the dose of Bbu partially rescues the phenotype of homozygous Su(H)⁸ mutants, which are larval lethal with small eye discs (Schweisguth and Posakony, 1992). Su(H)⁸/Su(H)⁸;Bbu⁸/+ animals develop into the pupal stage with pharate adults displaying partially everted heads (Fig. 7C). Thus, reduced levels of Notch/Su(H) signaling can partially rescue the pupal phenotype caused by reduced levels of Bbu expression.

Bbu¹⁰⁵ gives rise to a small number of escapers when the flies are grown at low density. As shown in Fig. 7E, the bristles on the notum of these adults, particularly in the anterior portion, show defects in their polarity (compare with Fig. 7D). Since both planar polarity and lateral inhibition involve the Notch signaling pathway (for reviews see Adler, 1992; Shulman et al., 1998; Simpson, 1997), our results suggest that Bbu may act to adjust the levels of Notch signaling during these processes.

During third larval instar, Barbu expression occurs in eye- antennal imaginal discs and is restricted to narrow bands of cells on either side of the morphogenetic furrow from which most cells enter the neuronal pathway (Fig. 8A; see Wolff and Ready, 1991). Notch signaling has been shown to participate in the control of the expression of the proneural protein Atonal (Ato) which is essential for cell fate specification of the first neuronal cells (photoreceptors 8, R8; Baker et al., 1996; Nagel and Preiss, 1999). In Bbu¹⁰⁵, the pattern of Ato expression appears irregular compared with the wild-type pattern and, occasionally, Ato-stained R8 cells are missing (Fig. 8B). In addition, Ato levels anterior to the morphogenetic furrow appear increased relative to the levels in the R8 cells. Despite these abnormalities, no obvious defects are detected in the expression of Elav protein, showing that, in most ommatidia, eight neuronal photoreceptor cells are formed (data not shown). Adult escapers that are homozygous for Bbu¹⁰⁵ display rough eye phenotypes. Scanning electron microscopic analysis reveals that the roughened appearance in the eyes of Bbu¹⁰⁵ escapers is due to the failure to form a regular array of hexagonal corneal lenses, which is accompanied by an irregular distribution of interommatidial bristles (Fig. 8D,D′ compare with C,C′). This phenotype suggests that the Bbu¹⁰⁵ mutation causes polarity defects in the ommatidia, which was further confirmed by the analysis of tangential sections through the eyes. As shown in Fig. 8H, the degree of rotation of the ommatidia from homozygous Bbu¹⁰⁵ flies appears random since some ommatidia have not rotated at all, while others have rotated only 45° or have continued the rotation beyond the normal 90° (compare with Fig. 8G). Similar phenotypes have been reported previously with null mutations in frizzled and in flies in which Notch activity was altered during photoreceptor development (Zheng et al., 1995; Fanto and Mlodzik, 1999; Cooper and Bray, 1999). Delta/Notch signaling between the presumptive R3 and R4 photoreceptors was shown to determine the identities of these two photoreceptor subtypes which, in turn, determines the chirality and rotational direction of the ommatidium. While forced increase or decrease of Notch activity frequently results in duplications of R4 or R3, respectively (Fanto and Mlodzik, 1999; Cooper and Bray, 1999), both of these photoreceptor subtypes appear to be present in the majority of the ommatidia in Bbu¹⁰⁵ mutants and the overall composition of photoreceptors is normal (Fig. 8H). Thus, it appears that the primary defect in Bbu¹⁰⁵ is a randomization of the binary R3/R4 cell fate decision, which leads to random chiralities and uncoordinated rotations of the ommatidia (see Discussion).

To examine the consequences of complete loss of Bbu activity for eye development, we generated clones of cells that were mutant for the null allele Bbu⁹⁹ in adult flies. For this purpose, the FRT/FLP system was used in combination with a chromosome carrying ey-FLP, which drives the FLP recombinase specifically in the eye (Xu and Rubin, 1993; B. Dickson, personal communication). Because we were unable to use the white eye color marker to identify the clones in this experiment, we used a 3L FRT chromosome arm carrying a Minute mutation (Morata and Ripoll, 1975), which caused slow growth and recessive lethality of cells that are heterozygous or homozygous for Minute, respectively. Therefore, Bbu mutant clones produced by this combination covered more than 90% of the eye surface. Scanning electron micrography revealed severe roughening as well as a size reduction of the ventral half of the eyes containing these large Bbu⁹⁹ clones (Fig. 8E,E′ compare with C,C′). As shown in the tangential sections in Fig. 8I, the defects include aberrant rotation of some ommatidia from mutant clones, which is similar to the phenotype observed in homozygous Bbu¹⁰⁵ flies (Fig. 8I compare to H). However, many ommatidia from Bbu⁹⁹ clones display stronger defects and show a significant reduction of the number of photoreceptor cells to four, five or six cells per ommatidium in the plane of sectioning (Fig. 8I). Thus, in addition to ommatidial polarity, complete loss of Bbu activity appears to disrupt photoreceptor recruitment.

For further analysis of Bbu function during eye development, we used the UAS-GAL4 system to overexpress Bbu during this process. A GMR-GAL4 line (Freeman, 1996) was used to drive Bbu Δ3′-UTR within the eye. Flies carrying this combination of constructs have roughened eyes, which is particularly evident in posterior portions of the eyes (Fig. 8F). This differential effect may be due to higher expression levels in posterior areas, as indicated by the increased w⁺ activity in posterior areas of the eyes in flies from the particular GMR- GAL4 driver line used for this experiment. A high magnification view shows that the ommatidia lack a regular organization and frequently display defects in the corneal lenses, which appear to contain extra sockets, suggesting transformations of photoreceptors to non-neuronal fates (Fig. 8F′). In agreement with this interpretation, tangential sections of such eyes with ectopic Bbu expression show many ommatidia with less than the normal number of photoreceptors. In addition, ommatidia with a full complement of photoreceptors show defects in their chirality (Fig. 8J). Thus, overexpression of Bbu affects not only the chirality, but frequently also the specification of neuronal cells of the ommatidia.

Barbu is the third member of an emerging family of genes that also includes the non-bHLH genes m4 and mα of the Enhancer of split complex. Bearded, which maps directly downstream of Barbu, encodes a protein with a low degree of sequence similarity to the N-terminal half of M4 and Mα and may be a more distant member of this gene family. While the primary structure of these small proteins does not give us any clues to their biological function, the genetic data presented here provide strong support for the notion that the proteins of this family act as antagonists of Notch signaling activity.

The clearest evidence for a function of Bbu in downregulating Notch activity is derived from ectopic expression experiments. Overexpression of Bbu, either in the mesoderm or wing imaginal discs, causes phenotypes similar to the ones observed for mutations that reduce the activity of Notch or other genes required for Notch signaling. The overexpression phenotype in eye discs is also compatible with decreased Notch activity, since Notch has been shown to be active in multiple cell fate decisions during development of both photoreceptor and non-neuronal cells, which in turn influences ommatidial polarity. Thus, Bbu can antagonize Notch in multiple developmental contexts and different germ layers. The events that can be antagonized by Bbu most efficiently involve lateral inhibition, during which Notch activity prevents most cells of a cluster to adopt a fate as muscle progenitor, neuron or sensory organ precursor, respectively. In both mesoderm and notum, however, the phenotypes upon Bbu overexpression are significantly milder than those observed for null mutations in Notch signaling components (Corbin et al., 1991; Heitzler and Simpson, 1991; Bate et al., 1993; Schweisguth and Posakony, 1992). Although substitution of the 3′-UTR, which contains conserved sequences that were previously found to confer mRNA instability (Lai and Posakony, 1997; Lai et al., 1998), with a 3′-UTR from SV40 results in enhanced phenotypes, it still does not completely abolish Notch activity. Since our GAL4 drivers can provide transcript levels that are in the range of endogeneous mRNAs, the observed weaker phenotypes could either be due to a very short half-life of the Bbu polypeptide or to the intrinsic inability of Bbu to fully inhibit Notch signaling. Of note, the Bbu overexpression phenotypes in the notum are very similar, qualitatively and quantitatively, to those observed upon ectopic expression of E(spl) m4 and mα (Apidianakis et al., 1999) and to gain-of-function phenotypes of Bearded (Leviten and Posakony, 1996). Therefore, it appears that, apart from their sequence similarities, the proteins of this family share some functional properties. In addition to their effects on lateral inhibition, they also appear to have some, albeit more limited, capacity to interfere with Notch activity during asymmetric cell fate specifications. In particular, strong gain-of-function alleles of Brd cause occasional transformations of hair and socket cell fates into neuron and sheath cell fates (Leviten and Posakony, 1996). This phenotype would be expected if Notch activity is inhibited during the first asymmetric division of the sensory organ precursor. By contrast, Apidianakis et al. (1999) did not detect any effects of ectopic m4/mα expression on Notch- dependent steps during asymmetric cell specification in the adult sensory organs and our studies with overexpression of Bbu did not reveal a reduction of pericardial cells, which were previously shown to be formed in a Notch-dependent event during asymmetric division of a progenitor cell (Ruiz-Gomez and Bate, 1997; Carmena et al., 1998b). Therefore, the overall ability of these proteins (at least of Bbu, M4, and Mα) to interfere with asymmetric cell specification via Notch appears to be significantly lower than that of Numb, a Notch antagonist that has an endogeneous role in blocking Notch activity during asymmetric cell fate decisions (Rhyu et al., 1994; Ruiz-Gomez and Bate, 1997; Carmena et al., 1998b; Park et al., 1998).

Bbu is the first member of this gene family to show a loss- of-function phenotype. While Bbu loss-of-function alleles cause lethality and developmental defects, null mutants for Brd (Leviten and Posakony, 1996) and E(spl) m4 (Apidianakis et al., 1999) are homozygously viable and do not exhibit any detectable abnormalities. It is likely that the wild-type phenotypes of Brd and m4 null mutants are due to functional redundancy among genes of this family. Partial functional redundancy may also explain our failure to detect any phenotypes in Bbu null mutant embryos. Nevertheless, the observed pupal and adult phenotypes, as well as the genetic interactions of Bbu loss-of-function mutations with Su(H) mutations, are compatible with a role of Bbu in the downregulation of Notch signaling activity. Prominent features that are shared with Notch gain-of-function phenotypes are polarity defects and disrupted cell specification during eye development. Similar to heat-shock induction of N^intra, reduction of Bbu activity results in increased levels of atonal expression anterior to the morphogenetic furrow and, in some instances, loss of ato expression in the prospective photoreceptor R8 (Fig. 8B; Baker et al., 1996; Baker and Yu, 1997). We do not observe any overt R3 to R4 transformations in the mis-oriented ommatidia upon reduction of Bbu activity, although such cell fate transformations were observed upon ectopic Notch activation and thought to be a cause for polarity defects (Cooper and Bray, 1999; Fanto and Mlodzik, 1999). Rather, we observe randomly oriented ommatidia with apparently normally specified photoreceptors, which is very similar to the defects reported for mutations in frizzled and its downstream effector gene dishevelled (Zheng et al., 1995; Boutros et al., 1998). Recent models for the origin of the ommatidial polarity proposed the occurrence of reciprocal Notch/Delta signaling between the prospective R3 and R4 cells. This mutual interaction is thought to be biased by increased Frizzled/Dishevelled activities in R3, which downregulates Notch activity and upregulates the production of Delta in this cell (Cooper and Bray, 1999; Fanto and Mlodzik, 1999). The observed Bbu mutant phenotype in ommatidia suggests that, upon the reduction of Bbu levels, this bias is removed and the decisions between R3 and R4 fates within each ommatidum become randomized. Thus, we propose that, in the normal situation, Bbu is required together with activated Frizzled to antagonize Notch activity in the prospective R3 photoreceptor. While we prefer a model in which Bbu acts in parallel with Frizzled, perhaps by potentiating or stabilizing its negative effect on Notch, we cannot rule out the alternative that Bbu acts in the Frizzled pathway upstream of Notch. Because Notch has been shown to participate in the specification of essentially all cell types of the eye disc (Cagan and Ready, 1989), the reduced numbers of photoreceptors in Bbu null mutant clones may also be due to increased activities of the Notch pathway during different steps of eye development. Taken together, the combined data from overexpression and loss-of-function experiments make a strong case for a role of Bbu in downregulating Notch activity. However, by no means do they exclude the possibility that Bbu (and by extension, E(spl) m4, mα, and Brd) acts in additional pathways that do not involve Notch signaling.

At present, we do not have any information about the molecular role of Bbu and other members of this protein family. The primary sequence strongly indicates that Bbu has an intracellular function. In this regard, it is intriguing that we had isolated Bbu as a result of its protein-protein interactions with the homeodomain protein Tinman. Although we do not have formal proof that this interaction is relevant in vivo, several lines of evidence support its specificity, including the fact that three independent clones of Bbu cDNAs were isolated in the yeast two-hybrid screen and the observation that GST- Bbu fusion protein specifically associates with radiolabeled Tinman protein in in vitro binding assays (S. Z. and M. F., unpublished observations). The observed co-expression of Bbu and Tinman in the early mesoderm would also favor the possibility of interactions of the two proteins in vivo. If this interaction were to occur, it would imply a nuclear function of Bbu, presumably as a cofactor of Tin in the mesoderm and additional transcription factors in other tissues. In this hypothesis, Bbu would interfere with transcriptional outputs of the Notch signaling cascade that require tissue-specific transcription factors such as Tinman. For example, it is possible that activated Notch/Su(H) complexes need to bind to enhancer sequences of certain target genes together with Tinman to activate their expression in the mesoderm, and that the binding of Bbu to Tin interferes with this cooperative activity. Indeed, tin-dependent specification events in heart and somatic muscle development also involve Notch signaling (for reviews see Frasch, 1999a,b). While the predominantly cytoplasmic localization of Bbu protein seems to argue against an interaction with Tin in vivo, it is still possible that the nuclei contain low levels of Bbu protein that cannot be detected with antibody stainings above background levels. This situation could be similar to that of the intracellular domain of Notch, which also cannot detected with certainty in the nuclei of wild- type embryos by using antibodies, although more sophisticated experiments have demonstrated that it does function within the nuclei (Lecourtois and Schweisguth, 1998; Struhl and Adachi, 1998).

The expression pattern of Bbu is related, but not identical, to the patterns of E(spl) m4 and mα expression. All three genes are expressed near the morphogenetic furrow in eye imaginal discs, but the expression domains of Bbu are broader and only Bbu is prominently expressed posterior to the morphogenetic furrow (Fig. 8A; Wurmbach et al., 1999). These differences could explain why the wild-type activities of m4 and mα cannot (at least not fully) substitute for loss of function of Bbu during eye development. In the wing disc, Bbu is expressed at significantly lower levels than in the eye disc, and this expression does not correspond to discrete proneural clusters, as observed for m4, mα and Brd (S. Z. and M. F., unpublished results). In addition, in early embryos during cellular blastoderm stages and gastrulation, when m4 and mα are expressed only in mesoectodermal cells, Bbu is expressed in all cells of the embryo except those of the presumptive mesoderm. Furthermore, during early neurogenesis, Bbu expression is initially excluded from proneural clusters while, during the process of lateral inhibition, it becomes re-activated in all cells except for the presumptive neuroblasts. A similar expression profile appears to occur during lateral inhibition in the mesoderm. Thus, the Notch-dependent activation of Bbu in proneural clusters is similar to the previously reported responses of E(spl) m4 and mα to Notch, with a major difference that, prior to this event, Bbu is expressed in the entire neuroectoderm and becomes transiently excluded from proneural clusters. As in the mesoderm, the transient repression of Bbu in proneural clusters may be controlled by genes of the snail family such as escargot, which is expressed in these clusters (Whiteley et al., 1992).

What might be the biological role of Bbu and the related proteins of this family? The transcriptional activation of its gene by the Notch pathway and its activity as an antagonist of Notch signaling would suggest that Bbu is involved in a feed- back inhibition loop during Notch signaling. Since during embryonic and imaginal development, Notch-dependent regulatory events frequently occur in a rapid sequence, it is important for Notch signaling activity from the first event to be extinguished before the next one initiates. For example, during early embryogenesis, some of the cells from the initially formed proneural or promuscular clusters in which Notch has been active appear to become incorporated into newly formed clusters shortly afterwards (Hartenstein and Campos-Ortega, 1984; Carmena et al., 1998a). In order to allow neuroblast and muscle progenitor determination during the subsequent rounds of segregation, the prior termination of Notch signaling activity from the previous event would seem to be necessary. In turn, a short half-life of the Bbu mRNA, which appears to be dictated by 3′-UTR sequences that are also found in the mRNAs of m4, mα and Brd, and of the Bbu protein, may be essential for effective initiation of a new round of Notch signaling. Because of the presumed functional redundancy of genes within this family, simultaneous removal of the activity of several of them will be required to test this model of feed-back inhibition and the functional range of these genes during Notch-dependent, as well as perhaps Notch-independent, developmental processes.

We thank A. Carmena, S. Carroll, B. Dickson, C. Jagla, Y.-N. Jan, R. Lehmann, M. Leptin, C. Maurel-Zaffran, H. Nguyen, G. Struhl and J. Treisman for generous gifts of antibodies and fly strains and R. Mann for the pGAD library. We gratefully appreciate the help of Corinne Maurel-Zaffran during the tangential sections of the adult eyes and thank Hanh Nguyen for comments on the manuscript. We thank the Berkeley Drosophila Genome project and the Bloomington and Umea stock centers for information and fly strains used in this study. This work was funded by a grant from the American Heart Foundation to M. F. and a fellowship from the ‘Association Française contre les Myopathies’ to S. Z.