The Drosophila gene Bearded encodes a novel small protein and shares 3′ UTR sequence motifs with multiple Enhancer of split Complex genes

Distributed in a stereotyped pattern on the body surface of adult Drosophila is a diverse assortment of external sensory organs that together comprise most of the peripheral nervous system (PNS) of the fly. Each organ is composed of a specific set of differentiated cell types, which includes one or more sensory neurons and a small number of non-neuronal accessory cells that produce the characteristic stimulus-receiving structures of the organ. For example, the mechanosensory bristles that occur in large numbers over most of the cuticular surface each contain a single bipolar neuron and three other cells (the thecogen, trichogen and tormogen) that construct, respectively, a cap over the dendrite of the neuron, the bristle shaft, and the socket around the base of the shaft. The four cells that comprise each bristle derive, via a fixed lineage, from a single sensory organ precursor (SOP) cell (Hartenstein and Posakony, 1989).

The development of these sensory bristles commences during the late larval and early pupal stages within undifferentiated ectodermal sheets (the imaginal discs and histoblast nests) that will ultimately give rise to the cuticle of the adult fly. First, small groups of cells called proneural clusters express two proneural genes, achaete (ac) and scute (sc), which encode transcriptional activators of the basic helix-loop-helix (bHLH) class (Alonso and Cabrera, 1988; Cabrera and Alonso, 1991; Cubas et al., 1991; Skeath and Carroll, 1991; Van Doren et al., 1992). The activity of the ac and sc proteins confers on the cells of the proneural cluster the potential to adopt the SOP cell fate. However, inhibitory cell-cell interactions within each cluster, mediated by the neurogenic genes, then restrict the expression of the SOP fate to a single cell, leaving the inhibited cells to adopt the ordinary epidermal fate (Dietrich and Campos-Ortega, 1984; Hartenstein and Posakony, 1990; Parks and Muskavitch, 1993; Schweisguth, 1995; Schweisguth and Posakony, 1992, 1994; Tata and Hartley, 1995). Once the SOP cell has been stably specified, the four constituent cells of the bristle organ are generated by three successive asymmetric cell divisions (Hartenstein and Posakony, 1989). Cell-cell interactions requiring neurogenic gene function are required for the asymmetry of daughter cell fates at each division (Posakony, 1994).

Several genes of the neurogenic group encode proteins that are known to function as part of the Notch cell-cell signalling and intracellular transduction pathway. Two transmembrane proteins, the products of Notch (N) and Delta (Dl), act as receptor and ligand, respectively, in this system (Fehon et al., 1990; Heitzler and Simpson, 1991; Struhl et al., 1993). Inter-action of their extracellular domains results in the activation of Suppressor of Hairless [Su(H)], a sequence-specific DNA-binding protein, by means of a direct protein-protein interac-tion with the intracellular domain of the N receptor (Fortini and Artavanis-Tsakonas, 1994; Hsieh et al., 1996; Jarriault et al., 1995; Tamura et al., 1995). Su(H) then transduces the signal in part by functioning as a direct transcriptional activator of most of the genes of the Enhancer of split Complex [E(spl)-C] (Bailey and Posakony, 1995; Furukawa et al., 1995; Lecourtois and Schweisguth, 1995). This complex includes seven tran-scription units that encode bHLH transcriptional repressor proteins, as well as E(spl)m4, the product of which is a novel 152-amino acid (aa) protein of unknown function (Delidakis and Artavanis-Tsakonas, 1992; Klämbt et al., 1989; Knust et al., 1992).

A number of other known genes exhibit loss-of-function phenotypes similar to those of N pathway genes, but at present the specific roles of their protein products remain unknown. These include mastermind (mam) (Smoller et al., 1990), neu-ralized (neu) (Price et al., 1993) and kuzbanian (kuz) (Rooke et al., 1996). Recently we reported that gain-of-function mutations at a new locus, Bearded (Brd), cause phenotypic effects on adult sensory organ development that closely mimic those resulting from loss-of-function mutations in N pathway genes (Leviten and Posakony, 1996). These include the com-mitment of supernumerary proneural cluster cells to the SOP fate, and the appearance of additional sheath cell/neuron pairs at the expense of the socket and shaft cells in the sensory organ lineage. Moreover, we found that these dominant phenotypic effects of Brd gain-of-function alleles are sensitive to the dosage of two neurogenic genes, N and neu, and of Hairless (H), a negative regulator of N pathway activity (Bailey and Posakony, 1995; Bang et al., 1995; Bang and Posakony, 1992; Brou et al., 1994; Schweisguth and Posakony, 1994; Vässin et al., 1985). In an earlier study, we showed that both Brd and the genes of the E(spl)-C are directly activated in the proneural clusters of the wing imaginal disc by protein complexes that include ac and sc, and that the same genes are expressed in the vicinity of the morphogenetic furrow of the eye imaginal disc under ac/sc-independent control (Bailey and Posakony, 1995; Singson et al., 1994).

Here we present a molecular analysis of the structure and expression of both wild-type and mutant Brd genes. We have determined that Brd gain-of-function alleles are associated with an insertion of a retrotransposon of the blood family into the 3′ untranslated region (UTR) of the gene. We find that the wild-type Brd gene encodes a novel small protein of only 81 aa that is predicted to include a highly basic amphipathic α-helix. We show that the 3′ UTRs of Brd and E(spl)m4 exhibit an unusually high degree of sequence identity, that the Brd and E(spl)m4 proteins are also related, and that the E(spl)m4 protein is likewise expected to form a basic amphipathic α-helix, suggesting that the two proteins have related bio-chemical activities. Finally, we report that Brd shares with several transcription units of the E(spl)-C, including E(spl)m4, two novel heptanucleotide sequence motifs in their 3′ UTRs, suggesting that these genes are subject to a previously un-recognized mode of post-transcriptional regulation. These observations strengthen the functional association between Brd and the N signalling pathway.

Flies were grown on standard yeast-cornmeal-molasses-agar medium. Canton S flies were utilized as the wild-type stock for in situ hybrid-ization experiments. Brd alleles and additional stocks are described in Leviten and Posakony (1996).

In situ hybridization to larval imaginal discs and pupal nota was performed using digoxygenin-UTP-labeled RNA probes essentially as described by Tautz and Pfeifle (1989), with the modifications of Jiang et al. (1991) and Schweisguth and Posakony (1992). Double-labeling experiments involving the enhancer trap marker A101 were performed as described previously (Romani et al., 1989).

Brd cDNA clones were isolated from a 4-8 hour embryonic cDNA library carried in the pNB40 plasmid vector (Brown and Kafatos, 1988). Library DNA was transformed into Epicurean SCS1 super-competent cells (Stratagene) as described by the manufacturer, and colony blotting and hybridization was performed as described (Maniatis et al., 1982). Five independent cDNA clones were isolated from approximately 1×10⁴ colonies screened.

Drosophila genomic DNA used in Southern blot analysis and in PCR reactions was isolated as described in Ellis et al. (1990).

To isolate a P element-tagged DNA fragment from the Brd^P1 chro-mosome, genomic DNA from this strain was digested with EcoRI and electrophoresed on a 0.8% agarose gel. Fragments between 6.0 and 9.0 kb were gel purified using Geneclean (Bio101) and ligated to EcoRI-digested and dephosphorylated λZAP phagemid vector (Strat-agene). Phage were plated and DNA transferred to nylon filters (Gene Screen) according to the Stratagene protocol. Filters were then probed with cloned P element sequences. Positive clones were subjected to plasmid rescue according to the λZAP protocol and characterized by restriction mapping and Southern blot analysis.

Two phage clones carrying wild-type genomic DNA from the Brd locus were isolated from two separate libraries by standard screening techniques (Maniatis et al., 1982). λBrd1 was isolated from the EMBL3 ‘D’ library obtained from M. Muskavitch. This phage contains approx. 15 kb of genomic DNA extending from just 5′ of the Brd transcription unit to nearly 14 kb downstream. λBrd2 contains approx. 13 kb of genomic DNA, including approx. 8 kb upstream of the Brd transcription unit. This phage was isolated from a λFixII library constructed by Stratagene.

Poly(A)⁺ RNA preparations from carefully staged embryos were the same as those described by Ellis et al. (1990) and Margolis et al. (1994). RNA from larvae and pupae was prepared as described by Ellis et al. (1990). Poly(A)⁺ RNA was isolated using the polyATract kit (Promega) according to the manufacturer’s instructions.

Two oligonucleotide primers were used to amplify by PCR a product corresponding to bp 1-400 of the Brd cDNA sequence from genomic DNA prepared from various Brd mutant stocks. The genomic DNA was purified from homozygous mutant adults when possible, and in other cases from flies transheterozygous for a particular mutant allele and the Brd^Prv1 chromosome, which is completely deleted for the Brd gene. The 5′ primer corresponds to bp 1-18 and has an attached EcoRI site at its 5′ end; the 3′ primer extends from bp 383-400, with a synthetic HindIII site at its 5′ end. Amplified fragments were cloned into pBluescript KS (Stratagene), and sequenced using the Sequenase 2.0 kit (USB). For each mutant chromosome analyzed, at least two clones from independent amplification reactions were isolated and sequenced.

Total RNA was isolated from Brd¹ and w¹¹¹⁸ late third-instar larvae as described above. First-strand cDNA was synthesized from 50-100 μg of total RNA using the oligonucleotide JK1 (C₁₃AAGCTTT₁₅; provided by J. Kavaler) and first-strand synthesis components and protocol from the Riboclone cDNA Synthesis System (Promega). A portion of this cDNA product was PCR amplified using a 5′ primer corresponding to Brd coding sequence (bp 149-171 of Brd cDNA; AAGGCCATGAAGAAGAAGATCTTCAA) and a 3′ primer JK2 (ATCAGGTGACGTCGAC₁₃; provided by J. Kavaler). A small fraction of this first amplification reaction was then reamplified using JK2 and a second Brd primer (bp 206-223; AGAAGATCTC-AGCAACTGAAGAACCAC) positioned downstream in the Brd cDNA sequence from the first Brd primer.

A 550-bp SalI restriction fragment, extending from the 5′ UTR of Brd to a site approximately 180 bp within the blood element inserted in the Brd 3′ UTR (see Fig. 1), was filled in using Klenow poly-merase, and linkered with phosphorylated XbaI linkers. The resulting XbaI fragment was cloned into the unique XbaI site of the CaSpeR-Hsp70 P element transformation vector (Bang and Posakony, 1992). DNA of the resulting P[w⁺, Hs-Brd] transposon clone was injected into w¹¹¹⁸ recipient embryos in combination with a Δ2-3 helper plasmid to provide transposase activity. Six independent transformant lines were established.

Crawling third-instar larvae and 0-, 6-and 16-hour APF pupae were placed in plastic Petri dishes containing a moistened paper towel and exposed to a 4-hour heat-shock regimen (2× 60 minutes in a 37°C air incubator, separated by 120 minutes at 25°C).

Adult flies were anesthetized and dissected in PBS. Dissected cuticle was placed in 1:1 (v/v) lactic acid:Hoyer’s medium mixture and incubated in a 55°C oven for 60 minutes to dissolve any associated tissue. The cuticle was then mounted on a slide, covered with a drop of Hoyer’s, and incubated at 55°C to harden.

The Brd^P1 allele was recovered in a P element mutagenesis screen designed to obtain phenotypic revertants of the original dominant Brd¹ allele, and displays a partial suppression of the Brd¹ bristle multiplication phenotype (Leviten and Posakony, 1996). Southern blot analysis of genomic DNA and in situ hybridization to salivary gland polytene chromosomes (data not shown) indicated that the Brd^P1 chromosome carries a single P element insertion at cytological position 71A1-2, the known site of the Brd locus (Leviten and Posakony, 1996). Exposure of the Brd^P1 chromosome to P transposase activity (Robertson et al., 1988) yielded a high frequency of derivative chromosomes that generate various modified Brd phenotypes, indicating that the partial suppression of the Brd¹ phenotype exhibited by Brd^P1 is most likely due to a P transposon insertion near the gene.

From a Brd^P1 genomic DNA mini-library, we isolated an EcoRI fragment of 7.0 kilobases (kb) that was found to contain a 1.5-kb defective P element (Fig. 1). Southern blot analysis showed that the P element is flanked in this fragment by 1.5 kb and 0.5 kb of unique genomic sequences and that the remainder of the clone consists of additional, non-P-element repetitive DNA.

Subsequent Southern blot and DNA sequence analysis revealed the following features of the cloned genomic region in wild-type and Brd mutant DNAs (Fig. 1). Firstly, all of the chromosomes derived by mobilization of the 1.5-kb P element in Brd^P1 display alterations in restriction fragment sizes in this region. Secondly, as expected, the parental Brd¹ chromosome from which Brd^P1 was derived lacks the P element entirely. Finally, Brd¹ and most of its derivatives do contain the non-P-element repetitive DNA, but these sequences are not present in this region in wild-type chromosomes. As described in more detail below, we have found that this repetitive DNA corre-sponds to a transposable element of the blood family. We suggest that the insertion of this element is the direct cause of the original Brd¹ mutation.

The two regions of single-copy genomic DNA in the 7.0-kb Brd^P1 clone were isolated and used separately as probes on northern blots containing poly(A)⁺ mRNA from both embryos and pupae. The 0.5-kb fragment that lies between the P element and the non-P repetitive DNA (Fig. 1) detects a single tran-script of about 600 nucleotides (nt) at both developmental stages in wild-type RNA (Fig. 2), while the transcript in the Brd mutants is both shorter (approximately 500 nt) and strik-ingly more abundant (Fig. 2A,B). This is consistent with the gain-of-function nature of the original Brd mutant alleles, and more specifically with the hypothesis that these mutations are hypermorphic (Leviten and Posakony, 1996).

Fig. 2C shows the temporal pattern of Brd transcript accu-mulation in poly(A)⁺ RNA of wild-type embryos. We observe only a very low level of Brd RNA in preblastoderm embryos, including a small amount of presumably maternal transcript at 0-2 hours. In 4-to 6-and 6-to 8-hour embryos, the abundance of the transcript is sharply higher; interestingly, this is the period during which neuroblast segregation, a N-regulated process, takes place. By the next time interval (8-12 hours), Brd transcript levels have declined sharply, and remain low throughout the rest of embryogenesis. This finding indicates that, at least in the embryo, wild-type Brd RNA is likely to be short-lived. Overall, it is clear that embryonic levels of Brd transcript are subject to strong temporal modulation.

Several cDNA clones (see below) representing the wild-type Brd mRNA were isolated from a 4-to 8-hour embryo library, and used to analyze the spatial pattern of Brd transcript accu-mulation. Whole-mount in situ hybridization to imaginal discs from late third-instar larvae reveals that the Brd transcript accu-mulates specifically in a proneural cluster pattern in both wild-type and Brd mutant imaginal discs (Fig. 3A,B; see also Singson et al., 1994). Consistent with the northern blot data (see Fig. 2), Brd RNA is significantly more abundant in the mutant discs, as reflected by a much stronger in situ hybridiz-ation signal (Fig. 3B). These observations provide further support for the interpretation that the gain-of-function alleles of Brd are hypermorphs (Leviten and Posakony, 1996), since Brd is apparently overexpressed, and not spatially misex-pressed, in these mutants. In addition, the finding that Brd is normally expressed in wild-type proneural clusters is consis-tent with the hypothesis that the gene has a wild-type function in adult sensory organ development (see Leviten and Posakony, 1996).

In the wing imaginal disc, Brd transcript is detectable at the positions of all developing adult external sensory organs, analogous to the expression of the proneural genes ac and sc. We have shown previously that Brd bristle phenotypes are dependent on ac/sc function (Leviten and Posakony, 1996), that Brd transcript accumulation within wing disc proneural clusters likewise requires ac/sc activity (Singson et al., 1994), and that the Brd promoter is directly activated in proneural clusters by bHLH protein complexes that include ac and sc (Singson et al., 1994). However, in contrast to ac and sc, there is generally no clear elevation of Brd transcript in presumptive SOPs.

To examine the spatial and temporal relationship of Brd expression to SOP specification, we performed double-labeling experiments with the SOP-specific enhancer trap marker A101. The results presented here are documented in Brd mutant discs, in which the in situ hybridization signal is stronger and more clusters are clearly represented than in wild-type discs.

Double-labeled third-instar wing discs demonstrate clearly that Brd transcript is localized to sites of SOP development (Fig. 3C). These experiments also reveal that Brd expression, like ac/sc expression, precedes SOP specification as marked by A101. For example, in the anterior scutellar (aSC) macrochaete proneural cluster, uniform Brd expression can clearly be seen prior to detectable A101 staining (Fig. 3D). Subsequently, a single A101-positive cell is observed among the Brd-positive cells (Fig. 3E). In many clusters there is clearly a period in which the multiplied SOPs express both Brd and A101. This is most evident in regions in which a cross-sectional view of the disc epithelium can be seen, and the A101-positive nucleus of an SOP cell is clearly surrounded by cytoplasmic Brd tran-script (Fig. 3F). Finally, Brd expression in A101-positive cells appears to become diminished at some positions in late third-instar and early pupal discs. For example, in the A101-positive SOP cells at the posterior scutellar (pSC) macrochaete position, Brd expression is much weaker than in the aSC region at the same time (Fig. 3E). Other positions, including the anterior postalar (aPA) macrochaete site, also exhibit unde-tectable levels of Brd transcript while expressing A101 strongly (data not shown). The pSC and aPA are two of the first macrochaete SOPs to be determined (Huang et al., 1991), suggesting that the decreased Brd RNA levels observed there represent the turnover of transcript and not simply a relatively lower level of Brd expression throughout the development of these particular clusters.

In summary, these results suggest a general pattern for the progression of Brd expression within imaginal disc proneural clusters. Initially, Brd transcript is present at roughly equal levels throughout the cluster, and this expression precedes that of the early SOP marker A101. Subsequently, Brd and A101 are coexpressed within the SOP cell (with Brd tran-script remaining in the non-SOP cells as well), and then Brd transcript levels become diminished in both the SOPs and the surrounding cluster cells.

We began the detailed charac-terization of wild-type and mutant Brd transcription units by isolating five independent cDNA clones from a plasmid library representing poly(A)⁺ RNA of wild-type 4-to 8-hour embryos. All contained inserts of approximately 550 bp, con-sistent with the size of the normal Brd transcript detected on northern blots (see Fig. 2). In addition to their identical sizes, two other criteria suggest that these cDNAs are full-length: two of the clones were found to start at the same nucleotide, and analysis of upstream genomic DNA sequence reveals a consensus TATA box (TATTTAA) 30 bp upstream of the first base of the cDNAs (Fig. 4) (Singson et al., 1994). In addition, the 3′ ends of the two clones analyzed are identical in sequence, each containing a single consensus polyadenyla-tion signal (AATAAA) 18 bp 5′ of the poly(A) tail, suggesting that the complete 3′ end of the Brd transcript is also represented within these cDNAs (Fig. 4).

One cDNA clone was sequenced in its entirety, revealing the presence of a single major open reading frame extending from the first ATG triplet at position +71 to a TGA stop codon at +314, flanked by 70 bp of 5′ untranslated region (UTR) and 233 bp of 3′ UTR (Fig. 4). The sequence of the entire Brd tran-scription unit from a wild-type genomic DNA clone (data not shown) showed that the genomic and cDNA sequences are completely colinear, indicating that there are no introns in the gene.

The exact structure of the Brd mutant transcription unit was similarly determined by DNA sequence analysis of the Brd^P1 genomic DNA clone (see above and Fig. 1). The P element insertion in the Brd^P1 allele lies 125 bp upstream of the putative transcription start site (Fig. 4), suggesting that it could partially suppress the Brd phenotype by interfering with the transcrip-tional regulation of the gene. The non-P repetitive element associated with Brd gain-of-function alleles (see above) was found to lie at position +399, in the 3′ UTR of the gene. Upstream of the point of insertion of this element, the mutant genomic DNA is identical to the wild-type cDNA sequence, with the exception of a 1-bp silent polymorphism in the coding region and another 1-bp polymorphism in the 3′ UTR (Fig. 4). The sequence of the 5′ terminus of this repetitive element (Fig. 4) identified it as a member of the blood family of retrotrans-posons, first recognized in association with the blood allele of the white locus (w^bl) (Bingham and Chapman, 1986). The blood element in Brd mutants is oriented such that a polyadenylation signal is located just 3 bp inside the LTR proximal to the Brd transcription start (Fig. 4). Efficient use of this blood polyadenylation signal could account for the trun-cation of the transcript in the Brd mutants (see Fig. 2A,B). To confirm this interpretation, we cloned the 3′ ends of the Brd transcript from both wild-type and Brd¹ larvae using RT-PCR (see Materials and methods), and determined their sequence. We found that the Brd¹ transcript does include blood element LTR sequences at its 3′ terminus, and is polyadenylated 15 nt downstream of the blood polyadenylation signal (Fig. 4). This generates a mutant transcript that is approximately 100 nt shorter than the wild-type larval transcript, which itself terminates at exactly the same position as indicated by the embryonic cDNA clones (Fig. 4). These data are entirely consistent with the relative sizes of the wild-type and mutant RNAs as observed on northern blots (Fig. 2).

Examination of the sequence of the wild-type Brd gene revealed that a 9-bp sequence motif, CAGCTTTAA, is directly repeated three times in the 3′ UTR (Fig. 4). We refer to this motif hereafter as the ‘Brd box’. The blood transposon present in the gain-of-function alleles is inserted between the first and second of these elements, and causes the two most 3′ Brd boxes to be excluded from the mutant transcript (Figs 1, 4; see Discussion).

The Brd open reading frame encodes a predicted protein of only 81 aa (9.2 kDa) (Fig. 4). While the protein overall is slightly basic (predicted isoelectric point 8.65), it has a dis-tinctly basic N-terminal half and a relatively acidic C-terminal half. The N-terminal region contains a high proportion of residues potentially subject to phosphorylation, including eight threonine and three tyrosine residues in the first 24 aa. In addition, there is a clear casein kinase II consensus phospho-rylation site (Kemp and Pearson, 1990) near the Brd C terminus (Fig. 4).

The primary distinguishing structural feature of the Brd protein is an 18-aa region (residues 26-43; Fig. 4) that is strongly predicted to form a basic amphipathic α-helix. A helical wheel plot of this region (Fig. 5A) reveals the alignment of six positively charged lysine residues on one face of the helix, with largely hydrophobic residues on the opposing face.

Cloning and sequence analysis of the Brd protein coding region from three Brd point mutants (Leviten and Posakony, 1996) confirmed our identification of the Brd gene. Two of these mutants, Brd^1rv1 and Brd^1rv2, exhibit identical 1-bp changes (relative to an isogenic parental chromosome) that generate stop codons in the middle of the protein just C-terminal to the Brd helical region (Q47 to stop, CAA to TAA). These mutations completely revert the dominant Brd pheno-types to wild type. The third mutant, Brd^1rv3, which is also a complete revertant, exhibits three nucleotide changes, creating two non-conservative amino acid changes in the N-terminal region of the protein (I11 to T, ATC to ACC; Q19 to E, CAG to GAA). Although the impact of an additional threonine residue in the already threonine-rich N terminus is uncertain, the substitution of an acidic glutamate residue adjacent to the basic helical region might very well affect charge-dependent interactions involving this region of the protein.

Although the predicted Brd protein (Fig. 4) is not strongly related to any known protein in the GenBank database, we were interested to find that Brd shares weak sequence similarity with the product of the m4 transcription unit of the E(spl)-C (Klämbt et al., 1989). Alignment of the Brd and E(spl)m4 sequences reveals two regions of limited similarity, one in the N-terminal portion of both proteins, and a second within acidic portions of both proteins (Fig. 6A). Since the region of highest identity overlaps the Brd amphipathic helix, we investigated whether the corresponding region of E(spl)m4 would likewise be expected to form a such a structure. Indeed, as shown in Fig. 5B, residues 24-38 of m4 are predicted to adopt an α-helical conformation with clear basic amphipathic character, suggest-ing that the Brd and E(spl)m4 proteins may have related bio-chemical functions.

Even more striking evidence for the re-latedness of the Brd and E(spl)m4 genes was provided by the finding that they exhibit an unusual degree of nucleotide sequence identity in their 3′ UTRs. Fig. 6B shows an alignment of a continuous 40-nt region in which 31 nt are shared between Brd and m4. This region includes a 14/14 identity containing the third (most 3′) Brd box in Brd, as well as an adjacent 7/7 identity that we refer to as the ‘GY box’ (see Fig. 4 and below). A search of the GenBank database with the 23-nt sequence from Brd that contains both of these identi-ties (Fig. 6B) reveals that E(spl)m4 is unique among known genes in its degree of relatedness to Brd in this region.

The observation that the 3′ UTRs of Brd and E(spl)m4 share two extended sequence identities, the Brd box and the GY box, prompted us to search more systematically for these motifs in genes of the E(spl)-C. The results are presented in Fig. 7. We find that both sequence elements are present in the 3′ UTRs of most genes of the Complex, sometimes in multiple copies.

The Brd box is most often represented by the 7-nt core sequence AGCTTTA (Fig. 7A), although in two cases [E(spl)m4-B1 and E(spl)m5-B3], the full 9-nt element CAGCTTTAA (repeated three times in Brd) is found. In both of these cases, the Brd box motif is part of a longer sequence identity with Brd [the 14/14 identity noted above that includes Brd-B3 and E(spl)m4-B1, and a 12/12 identity that includes Brd-B2 and E(spl)m5-B3]. Overall, of the nine tran-scription units in the E(spl)-C for which appropriate sequence data are available [seven bHLH repressor-encoding genes plus E(spl)m4 and groucho (gro)], the Brd box occurs in the 3′ UTRs of six of them (Fig. 7A). The exceptions are E(spl)mβ, E(spl)m8 and gro.

The 7-nt GY box (GTCTTCC) is also present in the 3′ UTRs of multiple E(spl)-C genes (Fig. 7A). As with the Brd box, GY boxes are also frequently part of longer sequence identities between genes, such as the 11/11 block that includes E(spl)m3-G1 and E(spl)m5-G3, or the 10/10 identity that contains Brd-G1 and E(spl)m3-G1. No gene was identified that contains a GY box without an accompanying Brd box.

Fig. 7B illustrates the relative positions of Brd and GY boxes in the 3′ UTRs of Brd and the E(spl)-C genes. Overall, no consis-tent positional relationship can be discerned between the Brd and GY boxes in the different genes, or between these elements and the stop codon, polyadenylation signal or polyadenylation site.

To investigate the effects of general and/or overexpression of Brd, we constructed a fusion gene in which the Brd protein coding sequences are expressed under the control of the Hsp70 heat-shock promoter. This chimeric gene is carried in the P element vector Casper-Hsp70 (Bang and Posakony, 1992). The design of the construct was influenced by the possibility that the high levels of transcript responsible for the Brd mutant effects require the presence of blood element sequences (e.g. for LTR-dependent truncation of the Brd transcript). Accordingly, a 550-bp SalI genomic DNA fragment from the Brd^P1 allele, extending from the Brd 5′ UTR to a position approximately 180 bp inside the blood element (see Fig. 1), was used to create the Hs-Brd transgene. Five independent transformant lines carrying this construct were analyzed.

Heat-shock pulses applied during the late larval and early pupal stages (see Materials and methods) generated bristle multiplication and bristle loss phenotypes resembling those of Brd gain-of-function mutations. The bristle multiplication effects were not as severe as those observed for the Brd¹ and Brd³ mutations, even in fly lines carrying four copies of the Hs-Brd construct. As expected, pulses of late third-instar larvae affected the earlier-developing macrochaetes, causing bristle duplications at some macrochaete positions, and a more reliable and general macrochaete loss phenotype (data not shown). Heat pulses applied during early pupal development (0, 6 and 16 hours APF) resulted in increases in microchaete number and density on the head and notum, comparable to those characteristic of weak Brd mutant genotypes (data not shown). In addition, significant microchaete loss was observed, and in some positions sockets without shafts were found, along with microchaetes with severely reduced shafts. These pheno-types have all been observed previously in Brd gain-of-function mutants. Finally, no general effects on epidermal cell or wing vein development were observed.

The P[Hs-Brd] transformant lines were exposed to P trans-posase activity in order to mobilize the transgene construct and potentially reposition it adjacent to strong enhancers in the genome, which would confer more stable, high levels of expression on the Brd gene. This approach has generated two lines of flies that exhibit strong bristle defects. One of these, Brd^HSJ1, displays strong dominant macrochaete and microchaete multiplication phenotypes similar to Brd³ heterozygotes (Fig. 8). Heterozygotes of this line exhibit microchaete tufting primarily in the anterior half of the notum, with a nearly wild-type microchaete pattern posteriorly. Homozygotes exhibit stronger and more uniform notum microchaete multiplication (Fig. 8), strong head and notum macrochaete multiplication (Fig. 8), and severely roughened eyes. The dominant bristle phenotypes of Brd^HSJ1 are almost certainly due to overexpression of the Brd gene. The recessive eye roughening is potentially caused by a mutational effect of the transposon insertion on an endogenous gene, although gain-of-function alleles of Brd are capable of producing similar eye roughening defects (M. W. Leviten and J. W. Posakony, unpublished).

For comparison, we also extensively investigated the ability of four copies of a Hs-Brd construct containing a wild-type Brd 3′ UTR to phenocopy Brd gain-of-function phenotypes. We found that this construct is incapable of conferring any dominant mutant phenotypes under a variety of heat-shock regimens. Moreover, we failed to detect any dominant Brd-like effects in over 5,000 progeny flies in which this construct was mobilized [using the Δ2-3 system (Robertson et al., 1988)] in an effort to position it near strong enhancers in the genome. These results are consistent with the hypothesis that the blood transposon-mediated truncation of the mutant Brd transcript (and the consequent loss of two Brd boxes and one GY box) is important to its ability to interfere with cell fate decisions in the adult PNS.

Several criteria indicate conclusively that we have identified the previously unknown transcription unit that corresponds to the Brd locus. Firstly, particular mutant alleles of Brd are asso-ciated specifically with insertions of transposable elements in or near this transcription unit; namely, a blood transposon in the 3′ UTR in the case of the dominant gain-of-function alleles, and a defective P element upstream of the transcription start site in the case of the partial revertant allele Brd^P1. Secondly, by comparison to wild-type or gain-of-function alleles, EMS-induced mutations that completely revert the Brd dominant phenotype consist of specific base substitutions in the open reading frame of the gene we have identified as Brd. Finally, overexpression of only the identified Brd gene under the control of a heat-shock promoter in transgenic flies partially phenocopies the dominant effects of the gain-of-function mutations.

We have investigated in detail the molecular lesions associated with the dominant gain-of-function mutations of Brd described in our earlier report (Leviten and Posakony, 1996). These alleles cause, with varying degrees of severity, sensory organ multiplication and loss phenotypes very similar to those resulting from loss-of-function mutations in the genes of the N signalling pathway (for a review see Posakony, 1994). Our analysis indicates that the likely origin of these Brd dominant phenotypes is the insertion of a retrotransposon of the blood family into the 3′ UTR of the gene. Due to the efficient usage of a polyadenylation signal present in the LTR of the transpo-son, transcripts from the mutant Brd gene are shorter than wild-type and lack the most 3′ 132 bases of the normal 3′ UTR sequence. In contrast, the predicted protein product of the gain-of-function alleles is unchanged from the wild type.

We have also found that steady-state levels of Brd transcript are several-fold higher in the gain-of-function mutants. Perhaps the simplest interpretation of this effect is that the truncated Brd transcript in these mutants has a significantly greater half-life, possibly due to the loss of instability-confer-ring elements in the distal 3′ UTR of the normal mRNA (see below). However, it is also possible that enhanced transcrip-tional activity of Brd caused by the LTRs or other sequences of the blood transposon contributes to the higher RNA levels we observe in the mutants.

Our previous genetic analysis (Leviten and Posakony, 1996) indicated that the Brd gain-of-function mutants almost certainly represent either neomorphic or strong hypermorphic alleles of the gene. Since, as described here, the predicted Brd protein in the mutants is the same as the wild type, and the spatial pattern of transcript accumulation in late third-instar imaginal discs (i.e. in proneural clusters) is also apparently unchanged, we favor the latter interpretation; namely, that the Brd dominant effects we have described result from hyperac-tivity of the normal Brd protein product. Nevertheless, it is important to note that we cannot presently rule out the possi-bility that a higher steady-state level of Brd RNA (particularly one of greater stability) at one stage might result in its persis-tence into a stage (or a cell type) from which it might normally be absent, thus yielding what is formally a neomorphic mis-expression effect.

Consistent with the specific effects of Brd gain-of-function mutations on adult sensory organ development, and the lack of wing vein or wing blade defects common to many other bristle mutants, Brd is expressed specifically in the proneural clusters of the wing and other imaginal discs as a direct target of tran-scriptional activation by ac and sc (Singson et al., 1994). We have shown here that Brd transcript accumulation precedes the appearance of the early SOP marker A101, and that following the onset of A101 expression, Brd expression is diminished in at least some SOPs, as has been reported for ac/sc expression (Cubas et al., 1991; Skeath and Carroll, 1991). However, in contrast to ac, sc and the ac/sc target gene scabrous (sca), we have not observed consistent elevation of Brd transcript levels within presumptive SOPs in either wild-type or Brd mutant wing discs. Similarly, in situ hybridization fails to reveal in SOPs elevated levels of transcript from several of the genes of the E(spl)-C that are also direct targets of ac/sc regulation (Singson et al., 1994). This difference in SOP expression of proneural proteins on the one hand, and of certain of their direct regulatory targets [Brd and E(spl)-C genes] on the other, suggests that some form of transcriptional or post-transcrip-tional repression acts in the latter case. Some insight is afforded by the observation that when a lacZ reporter gene is expressed under the control of the Brd promoter, higher levels of β-galac-tosidase activity are often detected in SOPs (Singson et al., 1994). Similar results have been obtained for reporter genes driven by certain E(spl)-C promoter fragments (Bailey and Posakony, 1995; Kramatschek and Campos-Ortega, 1994; Singson et al., 1994). Thus, in assays that specifically reflect transcription-level regulation, these genes’ promoters are found to be responsive as expected to the higher levels of proneural protein activity in the SOP.

We suggest, then, that regulatory mechanisms exist to reduce transcript accumulation from certain ac/sc activation targets in the SOP. While it is presumably advantageous to upregulate SOP expression of ac/sc target genes that promote the SOP fate, or are required for sending an inhibitory signal (e.g. sca; Baker et al., 1990; Mlodzik et al., 1990), the genes of the E(spl)-C act oppositely by antagonizing proneural activity, so high levels of these proteins in the presumptive SOP are undesirable. Indeed, their inappropriate expression in SOPs extinguishes this fate (Bang et al., 1995; Nakao and Campos-Ortega, 1996; Tata and Hartley, 1995). It seems reasonable that negative post-transcrip-tional modes of regulation, in particular, could play an essential role in limiting E(spl)-C transcript and protein accumulation in the SOP. It is not yet clear whether a similar rationale applies to Brd, because the mode of action of the Brd dominant mutants in generating supernumerary SOPs is unknown.

A key result of our molecular analysis of the Brd gene is the finding that it is related to the m4 transcription unit of the E(spl)-C. Though at present the function of neither gene is understood (see below), we have identified clear similarities in both the primary and secondary structures of their predicted protein products, and even more extensive (and unique) nucleotide sequence identities in their 3′ UTRs. Besides allowing us to recognize the shared GY box motif in the 3′ UTRs of Brd and the genes of the E(spl)-C, this discovery provides strong support for the hypothesis that Brd and E(spl)m4 have related functions during development.

The presence of a predicted basic amphipathic α-helix in the deduced protein products of both Brd and E(spl)m4 does not immediately suggest the nature of their biochemical activities. Helical domains of this type are associated with a variety of functions, including protein-protein interaction (e.g. binding of calmodulin) and interaction with lipid membranes (Segrest et al., 1990). It is certainly plausible that small proteins like Brd and E(spl)m4, lacking other apparent functional domains, might well exert their effects primarily via interaction with other proteins. We are currently pursuing this possibility in sys-tematic screens for interacting protein partners, as well as by investigating the possible in vivo significance of the predicted interaction with calmodulin.

The available genetic data are unfortunately insufficient to give a clear indication of whether Brd and E(spl)m4 exert similar regulatory effects in vivo. Neither gain-nor loss-of-function point mutations in the E(spl)m4 gene have been reported, and accordingly its specific function within the proneural cluster (if any) has yet to be determined. Since the other genes of the E(spl)-C function to antagonize proneural activity, E(spl)m4 might be expected to act similarly. In contrast, the simplest deduction from the Brd gain-of-function phenotypes (Leviten and Posakony, 1996) is that Brd normally acts to promote commitment to neural cell fates. Thus, even if the Brd and E(spl)m4 proteins rely on similar biochemical mechanisms of action (as suggested by their structural simi-larities), it is a distinct possibility that they have antagonistic, rather than cognate, regulatory effects. For example, both proteins might interact with a common protein or proteins, but have opposite effects on the activity of these factors.

In light of the structural evidence that their protein products may have related (though not necessarily cognate) biochemi-cal functions, it is of course noteworthy that Brd and E(spl)m4 are co-expressed in imaginal disc proneural clusters (Bailey and Posakony, 1995; Singson et al., 1994). Both genes are subject to direct transcriptional activation by the proneural proteins (Bailey and Posakony, 1995; Singson et al., 1994), while E(spl)m4 (Bailey and Posakony, 1995), but apparently not Brd (A. Bailey, A. Singson and J. W. Posakony, unpub-lished observations), is also directly activated by Su(H) in response to N receptor activity. Thus, as N-mediated signalling in the proneural cluster proceeds, the expression of E(spl)m4 would be expected to increase relative to that of Brd. As we learn more about the in vivo functions of the two genes, this distinction in their transcriptional regulation is likely to take on added significance.

Our analysis of the molecular structure and expression of both wild-type and mutant Brd genes provides, we believe, further support for the hypothesis that Brd normally plays a role in adult sensory organ development in Drosophila, and more specifically in the cell-cell signalling events mediated by the N pathway (Leviten and Posakony, 1996). Previously, we had established that Brd is expressed in the proneural clusters of the wing imaginal disc under the direct control of the proneural transcriptional activators ac and sc, and that it is also expressed near the morphogenetic furrow of the eye disc (another site of N-mediated cell-cell interaction) under distinct, ac/sc-indepen-dent, control (Singson et al., 1994). The detailed similarity at the cellular level between Brd gain-of-function phenotypes and those caused by loss-of-function mutations in the genes of the N pathway, combined with the sensitivity of Brd dominant effects to the dosage of both N and H, constitute the strongest circumstantial evidence for a functional connection between Brd and the N pathway (Leviten and Posakony, 1996). Three findings described in the present paper appear to strengthen this association. First, we have presented molecular evidence that the gain-of-function alleles of Brd are hypermorphic, indicat-ing that their phenotypic effects are likely to rely on the hyper-activity of a normal Brd protein expressed in its normal setting. Second, the temporal dynamics of Brd transcript accumulation in wing disc proneural clusters, clearly preceding the specifi-cation of the SOP, are consistent with a role in the cell-cell sig-nalling events that lead to this cell fate choice. Finally, the clear structural relatedness of Brd and a N pathway-regulated gene, E(spl)m4, as well as the sharing of the Brd box and GY box motifs with m4 and several other N-regulated transcription units in the E(spl)-C (see below), are especially compelling.

We believe that one of the most important results of the present study is the identification of two novel heptanucleotide sequence motifs, the Brd box and the GY box, in the 3′ UTRs of Brd and the genes of the E(spl)-C. These elements, notable for the length of the strictly conserved sequence, are sometimes present in multiple copies in a single gene, and are often part of even more extensive sequence identities (up to 14/14) between particular pairs of genes. The existence of these motifs strongly suggests that Brd and most of the E(spl)-C genes are subject to a shared, and previously unrecognized, mode of regulation. Given their location and their preferential orienta-tion, we currently favor the working model that the Brd and GY elements function as part of RNA transcripts to mediate some type of post-transcriptional control. However, it is also possible that they function instead in DNA, in an unusual mode of transcriptional regulation that relies on oriented sequence elements located in the 3′ UTR.

If the Brd and GY motifs do signify a common mode of post-transcriptional control for Brd and the E(spl)-C, there are several possibilities for the nature of this regulation, including RNA stability modulation, control of translation and RNA localization. We have described here our observation that the transposon-mediated elimination of two Brd boxes and a GY box from the Brd transcript is associated with a substantial increase in its steady-state level. This finding is consistent with the interpretation that either or both of these sequence elements normally function to confer instability on transcripts that include them. Indeed, we have recently found using reporter transgenes that specific mutation of the three Brd boxes in the Brd 3′ UTR results in a substantial increase in steady-state levels of both RNA and protein products in vivo (E. C. Lai and J. W. Posakony, unpublished observations).

At present it is perhaps simplest to suggest that the Brd and GY boxes represent sites of interaction of regulatory proteins that bind single-stranded RNA. If so, it is worth noting that the unusual conserved length of these motifs (7 nt) implies an unusual degree of sequence specificity for such RNA-binding activities. An alternative possibility, suggested by the precedent of the lin-14 and lin-4 genes of C. elegans (Ha et al., 1996; Lee et al., 1993), is that the Brd and GY elements bind complementary RNA molecules (encoded elsewhere), which are capable of forming short RNA duplexes that function in post-transcriptional regulation. Such RNA duplexes could themselves interact with regulatory proteins, either sequence-specifically or by sequence-non-specific recognition of the double-stranded structure.

Clearly, much remains to be learned about the functions of the Brd box and GY box sequence motifs. Nevertheless, it seems likely that understanding in detail the shared mode(s) of Brd and E(spl)-C regulation implied by the existence of these elements will be an important step in assembling an overall picture of the complex control of adult sensory organ devel-opment in Drosophila.

We would like to express our thanks to the following investigators for providing stocks and materials: Hugo Bellen, Nick Brown, Hilary Ellis, Josh Kavaler and Marc Muskavitch. We are very grateful to Dennis Spann for preparing staged embryo RNAs, to Mark Borowsky for his assistance with embryo injections, and to Jenee Wagner for her careful and patient work in preparing the figures. Adina Bailey, Ruth Bodner, Rick Firtel, Deborah Lycan and Dave Nellesen provided very helpful critical comments on the manuscript. This work was supported by an NIH grant (GM46993) to J. W. P. and by a graduate fellowship from the Lucille P. Markey Charitable Trust to E. C. L.