The bereft Gene, a Potential Target of the Neural Selector Gene cut, Contributes to Bristle Morphogenesis

Corresponding author: Rolf Bodmer, University of Michigan, 830 N. University, Ann Arbor, MI 48109-1048., rolf{at}umich.edu (E-mail)

Communicating editor: A. J. LOPEZ

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The neural selector gene cut, a homeobox transcription factor, is required for the specification of the correct identity of external (bristle-type) sensory organs in Drosophila. Targets of cut function, however, have not been described. Here, we study bereft (bft) mutants, which exhibit loss or malformation of a majority of the interommatidial bristles of the eye and cause defects in other external sensory organs. These mutants were generated by excising a P element located at chromosomal location 33AB, the enhancer trap line E8-2-46, indicating that a gene near the insertion site is responsible for this phenotype. Similar to the transcripts of the gene nearest to the insertion, reporter gene expression of E8-2-46 coincides with Cut in the support cells of external sensory organs, which secrete the bristle shaft and socket. Although bft transcripts do not obviously code for a protein product, its expression is abolished in bft deletion mutants, and the integrity of the bft locus is required for (interommatidial) bristle morphogenesis. This suggests that disruption of the bft gene is the cause of the observed bristle phenotype. We also sought to determine what factors regulate the expression of bft and the enhancer trap line. The correct specification of individual external sensory organ cells involves not only cut, but also the lineage genes numb and tramtrack. We demonstrate that mutations of these three genes affect the expression levels at the bft locus. Furthermore, cut overexpression is sufficient to induce ectopic bft expression in the PNS and in nonneuronal epidermis. On the basis of these results, we propose that bft acts downstream of cut and tramtrack to implement correct bristle morphogenesis.

STUDIES of cell fate specification in the peripheral nervous system (PNS) of Drosophila have focused primarily on two processes: the process by which a sensillum precursor assumes a particular fate and the process by which the sensillum precursor divides to produce the cells comprising the sensory organ (for recent reviews see VERVOORT et al. 1997 ; LU et al. 1998 ; BRUNET and GHYSEN 1999 ; JAN and JAN 2000 ). Each sensory organ forms specialized structures to transduce distinct stimulus modalities; accordingly, the identity of the organ affects the fate and function of every cell within it. These organs are composed of at least four different cells, and their fates are distinguished from each other via the asymmetric divisions of their sensillum precursor. The PNS sensory organs are categorized on the basis of sensillum structure and sensory neuron morphology: type I sense organs are innervated by monopolar, cilium-containing dendrites, whereas the type II sense neurons extend multiple dendrites and are thought to be touch receptors (GHYSEN et al. 1986 ; BODMER and JAN 1987 ). Type I organs are further classified as external sensory (es) organs, which secrete cuticular structures from the larval epidermis, and as internal chordotonal (ch) organs, which form internal attachments to the larval cuticle. es organs serve as mechano- and chemoreceptors, whereas ch organs function as proprioceptors (DAMBLY-CHAUDIERE and GHYSEN 1986 ; GHYSEN et al. 1986 ; FIELD and MATHESON 1998 ).

The genetic distinction between es and ch organs is under the control of the homeobox gene cut. cut is expressed in the es sensory organ precursors and their progeny and is required to correctly specify their identity (BODMER et al. 1987 ; BLOCHLINGER et al. 1988 , BLOCHLINGER et al. 1990 , BLOCHLINGER et al. 1991 ). cut acts similarly to homeotic selector genes: when cut function is removed, es organs are directed to assume the ch fate, whereas ectopic expression of cut in ch organ lineages causes transformation of ch organs into es organs. Although from this perspective cut behaves as an "activator" of es organ fate, evidence from in vitro experiments using mammalian cut homologs suggests that it may act to transcriptionally repress target genes (reviewed in NEPVEU 2001 ). Recent genetic experiments in flies also support the idea that cut suppresses non-es organ-derived cell fates (JARMAN and AHMED 1998 ; BREWSTER et al. 2001 ). While these data suggest cut may act as a transcriptional repressor, our understanding of the mechanism by which cut specifies cell fates remains limited. Thus, the identification of cut targets would aid in elucidating how it regulates sensory organ cell fates.

In the Drosophila embryo, a simple bristle-type es organ is composed of a neuron, a glial-like cell (thecogen), and two external support cells, the shaft-forming trichogen cell and the socket-forming tormogen cell. These cells are generated from a single ectodermal precursor through asymmetric divisions (LAWRENCE 1966 ; BODMER et al. 1989 ; HARTENSTEIN and POSAKONY 1989 ; BREWSTER and BODMER 1995 ; GHO and SCHWEISGUTH 1998 ; GHO et al. 1999 ), involving the segregation of Numb, a membrane-associated protein, to one daughter cell of the dividing precursor, but not the other (RHYU et al. 1994 ). The daughter cell that receives Numb protein, the pIIb cell, ultimately produces the neuron and thecogen cell, whereas the pIIa cell is the precursor to the external support cells (KNOBLICH et al. 1995 ; GHO and SCHWEISGUTH 1998 ). The asymmetric inheritance of Numb within the sensory organ lineages [and those of the central nervous system (CNS)] is necessary and sufficient to distinguish between alternative cell daughter fates (UEMURA et al. 1989 ; RHYU et al. 1994 ; SPANA et al. 1995 ). Numb exerts its function by inhibiting signal transduction of the transmembrane protein encoded by Notch (HARTENSTEIN and POSAKONY 1990 ; GUO et al. 1996 ).

ttk, a lineage gene encoding a zinc-finger protein (HARRISON and TRAVERS 1990 ), appears to act downstream of numb to implement sensory organ cell fates (GUO et al. 1995 ). ttk mutant embryos exhibit a phenotype opposite that of numb, in that pIIa is transformed into pIIb, resulting in excess neurons and glia. Furthermore, overexpression of ttk results in a phenotype similar to that observed in numb mutants, namely the sensilla lack neurons and glia, consisting entirely of support cells. ttk acts epistatically to numb, since embryos doubly mutant for both genes exhibit a ttk phenotype. Consistent with this result, Ttk protein, which normally is excluded from the neurons, exhibits ectopic neural expression in numb mutants, whereas the distribution of Numb protein appears unaffected in ttk mutants (GUO et al. 1995 ). Thus, ttk is likely to promote cell-type-specific gene expression in the daughter cells produced from asymmetric divisions of sensory organ precursors.

Lineage genes and selector genes clearly must regulate different aspects of sensory organ formation: the lineage genes direct the asymmetric divisions of the sensory organ precursors, but they do not appear to take part in specifying the identity of the sensory organ itself. The lineage genes are required and expressed in es as well as ch organs to distinguish the daughter cells from each other (UEMURA et al. 1989 ; RHYU et al. 1994 ; GUO et al. 1995 ). By contrast, the selector gene cut is expressed only in those sensory organs it specifies, i.e., in es organs (BODMER et al. 1987 ; BLOCHLINGER et al. 1988 , BLOCHLINGER et al. 1990 , BLOCHLINGER et al. 1991 ). For appropriate organogenesis of the sensillum structures to take place, organ identity and lineage information must ultimately be integrated within individual cells of a sensory organ. Thus, a cell needs to acquire at least two pieces of information: for example, (a) support cell information (provided by lineage genes) and (b) es organ-type information (provided by selector genes).

In an effort to identify and characterize genes that might integrate information from cut and ttk, we cloned a gene, bereft (bft), that is expressed in es, but not in ch support cells. Analysis of cDNA, reverse transcribed, and genomic sequence of the bft locus does not suggest an obvious protein-coding region. Thus, bft either encodes a very small protein or may act as an RNA. Analysis of flies with deletions of the bft locus, together with the es support cell-specific expression pattern, suggest that bft function is required for correct morphogenesis of the cuticular structure forming support cells, in particular those of the interommatidial bristles of the eye. Moreover, bft expression in es organs is reduced in cut and ttk mutants, and cut and ttk interact genetically with bft. These data are consistent with the idea that bft is a target for cut and ttk in the implementation of es organ-specific structures.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks:
numb¹ (UEMURA et al. 1989 ), ttk^702/7 (SALZBERG et al. 1994 ), ct^c145 and ct^db7 (BODMER et al. 1987 ), Df(2L)esc10, Df(2L)escP2-0, Df(2L)escP3-0, and Df(2L)prd1.7 were kept over PlacZ- or Py⁺-marked balancers.

For cut overexpression studies, the UAS-Gal4 system was used (BRAND and PERRIMON 1993 ). A 3.4-kb cDNA of the cut locus (rb10; BLOCHLINGER et al. 1988 ) was subcloned into pUAST and transgenic flies were generated (M. PARK and R. BODMER, unpublished results). These flies were then crossed to Gal4 lines that were expressed in the striped "hairy" pattern or throughout the embryonic ectoderm (h-Gal4 [1L3], 69B-Gal4; BRAND and PERRIMON 1993 ).

The E8-2-46 enhancer trap line has a PlacW insertion (BIER et al. 1989 ) at cytological location 33A-B. The insertion was mobilized using the 2-3 transposase source (ROBERTSON et al. 1988 ) and excisions were identified by loss of white expression according to standard procedures (e.g., BIER et al. 1989 ).

Immunocytochemical staining and in situ hybridization:
Antibody staining procedures were essentially as previously described (BODMER and JAN 1987 ). The following primary antibodies were used: mouse anti-Cut (monoclonal antibody 2B10 used at 1:20; K. BLOCHLINGER, unpublished data), 22C10 and 21A6 (1:100 and 1:10, respectively, mouse monoclonal antibodies; ZIPURSKY et al. 1984 ), and rabbit anti-ß-galactosidase (1:2000; Cappel). Embryos were incubated with the appropriate HRP-conjugated (Bio-Rad, Richmond, CA) or biotinylated (Vectastain) secondary antibodies (1:200). For fluorescent confocal microscopy Cy3- or FITC-conjugated secondary antibodies (Jackson Laboratories, West Grove, PA) were used, and embryos were mounted in VectaShield (Vector Laboratories, Burlingame, CA). Fluorescent embryo staining was analyzed using a Zeiss LSM510 confocal microscope and the images were further processed using Adobe Photoshop.

Digoxygenin-labeled riboprobes derived from the bft locus were made using the Boehringer Mannheim (Indianapolis) Genius kit, as described by the manufacturer. The protocol for combining antibody staining with in situ hybridization is modified from PATEL 1996 . Briefly, embryos were Chlorox dechorionated and fixed in heptane and PEM-FA [3 ml PEM and 2 ml 10% methanol-free formaldehyde, EM grade from Polysciences (Warrington, PA)] in a 1:1 proportion for 20–30 min while shaking at 200 rpm. The embryos were devitellinized in heptane and prewarmed 100% methanol (1:1). Finally, the embryos were stored in 100% ethanol at -20° for later use. To combine protein and mRNA localization, the devitellinized embryos were first rehydrated by washing several times in 1x PBS with 0.1% Triton-X (PT). Embryos were blocked for 30 min in antibody incubation buffer (AIB: 1x PBS, 0.1% Tween-20, 20 units/ml Rnasin, 25 µg/ml tRNA, and 0.1% BSA). Both primary and secondary antibodies were diluted in AIB to the appropriate concentrations. The primary antibody was added, and the embryos were incubated 2 hr at room temperature (rt) or incubated overnight at 4°. Then, embryos were washed for 1 hr with PT, changing solutions every 15–20 min. Secondary antibody was added and incubated 1 hr, followed by washing. To visualize the labeling, 0.3 mg/ml diaminobenzidine (Sigma, St. Louis) and 0.01% H₂O₂ were added. When antibody labeling was done alone, the procedure was the same as that described above, except the antibody incubation buffer was replaced with PT + 2% BSA. When antibody staining was used in combination with in situ hybridization, the embryos were washed several times with PT before starting the hybridization protocol. In situ hybridization was conducted according to TAUTZ and PFEIFLE 1989 , with the following modifications: embryos were cleared in xylene by gradually transferring them from 100% methanol to 100% xylene and then incubated in 100% xylene for 1 hr at room temperature followed by transferring them back to 100% methanol. The embryos were rehydrated as in TAUTZ and PFEIFLE 1989 , but 4% formaldehyde was used, instead of paraformaldehyde. After rehydration, embryos were washed with PBS and 0.1% Tween-20 (PTw) for 25 min, changing solution every 5 min. Embryos were not pretreated with proteinase K before hybridization. The prehybridization solution consisted of 50% formamide, 300 mM NaCl, 10 mM Tris-HCl pH 6.8, 10 mM sodium phosphate (pH 7), 5 mM EDTA pH 8, 1x Denhardt's. For hybridization, 10% dextran sulfate was added. Hybridization and detection were conducted as described in TAUTZ and PFEIFLE 1989 , with one additional step: before adding the digoxigenin antibody, the embryos were incubated in 20 µg/ml RNase A in PTw at 37° for 30–35 min. Subsequently the embryos were rinsed with PTw for 30 min, changing solution every 5 min.

Molecular biology:
Genomic DNA flanking the E8-2-46 insertion was recovered by plasmid rescue (PIRROTTA 1988 ). Restriction fragments of the genomic DNA were cloned into Bluescript (Stratagene, La Jolla, CA) and used as templates for DNA probes and riboprobes. DNA probes were made from 17 kb of genomic DNA distal to the insertion and were used to screen three cDNA libraries (POOLE et al. 1985 ; ZINN et al. 1988 ). A 1.5-kb cDNA was isolated from this screen. Riboprobes were generated and used to probe Northern blots of total RNA according to standard protocols (SAMBROOK et al. 1989 ), with the exception of the hybridization solution (NorthernMax prehybridization/hybridization solution; Ambion, Austin, TX). Total RNA was isolated using the TRIzol reagent (GIBCO BRL, Gaithersburg, MD), as directed by the manufacturer.

To isolate additional transcript sequences of the bft locus reverse transcriptase (RT)-PCR was carried out as follows: RNA was reverse transcribed using SuperScript II reverse transcriptase (GIBCO BRL). To amplify PCR products >3 kb, the Expand Long Template PCR kit (Boehringer Mannheim) was used. PCR products were cloned into pT7Blue3 using Novagen's Perfectly Blunt cloning kit. The 5' ends of RT-PCR products from the bft locus were determined using two different 5' rapid amplification of cDNA ends (RACE) protocols (FROHMAN 1994 ; CHEN 1996 ). The protocol obtained from Chen was modified by using Gitschier's buffer. Flies generated from mobilizing the P element were analyzed using PCR and Southern hybridization. For PCR analysis, DNA was isolated from single flies as described in ASHBURNER 1989 , except that spermine and spermidine were omitted from the homogenization buffer and diethylpyrocarbonate was omitted from the lysis buffer. Southern hybridization and all other DNA manipulations were performed using established protocols (SAMBROOK et al. 1989 ).

Sequencing:
All DNA sequences were determined using the Big Dye Terminator cycle sequencing kit (Perkin-Elmer, Norwalk, CT) or the AmpliTaq Dye Terminator cycle sequencing kit (Perkin-Elmer) and an automated ABI sequencer. DNA sequences, chromosomal location, and gene structure of the bft locus were obtained from three separate sources: cDNA libraries (POOLE et al. 1985 ; ZINN et al. 1988 ), RT-PCR and 5' RACE products obtained from RNA isolated from white flies, and genomic DNA obtained by plasmid rescue from yw;E8-2-46 flies (see above). A few polymorphisms were observed between the reverse-transcribed cDNA sequences of white (w) flies and the genomic DNA sequence obtained from yw;E8-2-46 flies, and these polymorphisms were confirmed by sequencing PCR products obtained from genomic DNA of w flies. Only one polymorphism has a consequence on open reading frame (ORF)1 and one on ORF3, as indicated in Fig 5; the others are silent. Additional polymorphisms were observed when our sequencing results were compared to the published genome sequence (ADAMS et al. 2000 ), as indicated in Fig 5. These are bona fide strain polymorphisms, since they are the same when compared to sequences obtained independently from w and from the yw;E8-2-46 genomic DNA (yw;E8-2-46-derived genomic DNA sequence is available under GenBank accession nos. AF234639 and AF234640). The polymorphisms do not cause a dramatic change in the overall lengths of the ORFs. The two exons of the 7-kb bft transcript correspond to nucleotide positions 2489–3322 (exon 1) of AF234639 and 2–5875 (exon 2) of AF234640. The 3.5-kb transcript consists of three exons: the first corresponds to 2489–3322 of AF234639, the second to 2–141 of AF234640, and the third to 3404–5875 of AF234640.

View larger version (80K): In this window In a new window Download PPT slide   Figure 1. lacZ expression of the enhancer trap line bftE8-2-46 in support cells of es organs. Confocal images of stage 14/15 embryos stained for Cut (green) and LacZ (red). Anterior is to the left and dorsal is up. (A–C) lacZ and cut expression overlap in the PNS. (D) Schematic diagrams showing the arrangement of sensory organs in an abdominal segment. Anterior is left and dorsal, up. (D, left) cut-expressing PNS cells are in green (intensity coded). Round cells are part of es organs. Light green round cells show neuron (larger) and thecogen cells expressing cut weakly. Darker green round cells are the strongly cut-expressing support cells, trichogen (intermediate green) and tormogen (dakest green). Diamond-shaped cells are multiple dendritic neurons; some are Cut-positive (BODMER and JAN 1987 ; BREWSTER et al. 2001 ). Oval-shaped cells belong to ch lineages. (D, right) bftE8-2-46 lacZ expression is present in the support cells of es organs only (red). The strongly lacZ-expressing trichogen cells are dark red, and the tormogens are in lighter red. (E–G) High magnification of the dorsal abdominal PNS. Note the overlap of Cut and LacZ in the es support cells, but in reciprocal intensity (green and red arrowheads in E and F). (G) The four es organ cells are identified by their relative dorsal-ventral position in the dorsal-most abdominal es organ, desD (GHYSEN et al. 1986 ). Note the strongly LacZ-positive trichogen cell (tr, red arrowhead), the strongly Cut-positive tormogen (to, dark green arrowhead), and the weakly Cut-positive neuron (n) and thecogen (th, light green arrowheads).

View larger version (135K): In this window In a new window Download PPT slide   Figure 2. lacZ expression of the enhancer trap line bftE8-2-46 is regulated by cut, numb, and ttk. All embryos are stained for lacZ expression; B and D are double stained with 21A6, which labels ch organ-specific scolopale structures (BODMER et al. 1987 ; for orientation, black arrowheads indicate the position of the lateral ch cluster, lch5; GHYSEN et al. 1986 ). (A–D) Two abdominal segments of stage 15/16 embryos are shown. Black arrows indicate lacZ expression in es support cells of the lateral cluster. White arrows indicate the reduced or absent lacZ expression in lateral es support cells. Asterisk indicates non-PNS lacZ positive cells. (A) bftE8-2-46 is expressed in the support cells of es organs. An average of 26 PNS cells exhibit labeling in each abdominal segment. (A) Wild type. (B) ctdb7 mutant embryo. In cut null mutant embryos, lacZ expression is reduced to an average of 15 cells per hemisegment, and the level of expression is lower. White arrowhead indicates 21A6-positive scolopale structure showing an es to ch organ transformation in the dorsal clusters (others are out of focus). In addition, cut mutant embryos are positively identified by the transformed and mispositioned sensory structures of the posterior spiracles (not shown; BODMER et al. 1987 ). (C) The bftE8-2-46-driven lacZ expression in a nb1 mutant embryo results in high levels of lacZ expression of four instead of two cells per es sensory organ (black arrows). (D) ttk702/7 embryo, double labeled for bftE8-2-46-driven lacZ expression and 21A6. ttk mutants were identified by the disorganization of the chordotonal-associated scolopales. Reporter gene expression is reduced to an average of 12 cells per hemisegment. (E–G) bftE8-2-46-driven lacZ expression in the humidity-receptive Keilin organs (KO) of the thoracic segments in (E) wild-type (eight nuclei ± 1 SD, n = 15), (F) ctdb7 (seven nuclei ± 1 SD, n = 15), and (G) nb1 embryos. Note the lower level of expression in F.

View larger version (16K): In this window In a new window Download PPT slide   Figure 3. Map of the genomic region surrounding P-element insertion E8-2-46. The insertion site is indicated and the scale is centered at the site of insertion. Position 1 corresponds to nucleotide 2447 of AF234639 and nucleotide 146792 of the genome sequencing project (AE 003634). AF234639 starts at the most proximal BamHI site and AF234640 at the beginning of exon 2 (E2/E2'). E1, E2, E2', and E3 are the exons of the 3.5- and 7-kb transcripts. Intronic regions are indicated with lines. The arrow indicates the direction of transcription. D1–D4 are DNA probes, and R1–R4 are RNA probes used for in situ hybridization. PE1, PE2, and C91 cDNA are probes used for Northern hybridization. The extent of the deletions in the bft6 and bft24 alleles is indicated by dashed boxes. B, BamHI; X, XbaI; R, EcoRI; P, PstI; H, HindIII; G, BglII; S, SacII.

View larger version (62K): In this window In a new window Download PPT slide   Figure 4. bft expression correlates with PNS development. (A–E) Northern blots probed with the C91 cDNA (A and D), with the PE1 probe of bft exon 1 (B), or with the loading control RP49 (C and E). (A–C) Arabic numerals indicate the age of embryos in hours, Roman numerals indicate larval instar stages. Female, adult female flies; male, adult male flies. (A) C91 cDNA detects a 7- and a 0.8-kb transcript. (B) PE1 detects the 7-kb transcript only. The 7-kb transcript appears 6–8 hr after egg laying (slanted arrow), the time at which the cells of the PNS have begun to divide. (D) The C91 probe detects the 7-kb transcript in wild-type embryos and pupae, but not in adults or larvae, whereas the 0.8-kb transcript is expressed at all stages. In bft225 and bft24 embryos the 7-kb transcript is not detected at any stage (slanted arrows), but the 0.8-kb transcript is unaffected. A, adult; P, pupae; L, larvae; E, embryo; wt, wild type.

View larger version (37K): In this window In a new window Download PPT slide   Figure 5. Open reading frames found in the proximity of the E8-2-48 insertion. Top depicts genomic region with possible ORFs in the direction of bft transcription (boxed). Shaded boxes indicate ORFs within the 7-kb transcript. Open boxes indicate ORFs within LP06727 (ORF5 and -6), which is contained within the first intron. Conceptual translations of the ORFs are indicated in one-letter codes. ORF1 extends from nucleotide position 2526–2990 of AF234639. The nucleotide found at position 2892 in the genomic sequence (AF234639) is missing in the reverse-transcribed cDNA sequence, causing ORF1 to end after six nucleotides (region highlighted in yellow). ORF2 extends from nucleotide 252–580, ORF3 from 2164–2478 (66 nucleotides, 2363–2428, are missing in genomic DNA derived from white flies and RT-PCR products), and ORF4 from 3758–4153 of AF234640 (first methionines are indicated in red). Additional polymorphic regions with regard to the published genome sequence (ADAMS et al. 2000 ) are highlighted in blue (see MATERIALS AND METHODS). Differences in conceptual translation are given in blue letters in parentheses.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The E8-2-46 reporter gene is coexpressed with cut in external sensory organ support cells:
We sought to identify targets of both cut and ttk, on the basis of the expression pattern of candidate genes within the PNS. cut is expressed in all the cells of es organs (at higher levels in support cells), whereas ttk is found in three es and two ch support cells, but not the neurons (BLOCHLINGER et al. 1988 , BLOCHLINGER et al. 1990 ; GUO et al. 1995 ). Thus, the support cells of es organs express both cut and ttk, suggesting that genes responsive to these two pathways (i.e., the pathways leading to organ identity specification and lineage decisions, respectively) should also be expressed in these cells. We identified an enhancer trap line, E8-2-46 (BIER et al. 1989 ), in which the lacZ reporter gene is expressed primarily in the support cells of es organs within the PNS, on the basis of position, morphology, and cut expression (Fig 1, A–D). Although E8-2-46 is expressed in both es support cells (as identified by high levels of cut expression), the level of expression is lower in one of them (Fig 1, D–F). To determine which of the two cell support cells express the reporter gene more strongly, we examined the dorsal-most abdominal es organs (desD; GHYSEN et al. 1986 ), where these cells are aligned in a stereotyped linear fashion (BLOCHLINGER et al. 1988 ): tormogen, trichogen, thecogen, and neuron (from dorsal to ventral; indicated in Fig 1G). Strong reporter activity is observed in the bristle shaft-forming trichogen cell (Fig 1F), whereas cut expression predominates in the shaft-forming tormogen cell (Fig 1E), which is particularly evident in the merged confocal image (Fig 1G).

E8-2-46 reporter gene expression is modulated by cut, numb, and tramtrack:
The gene associated with the regulatory sequences responsible for driving E8-2-46 expression is potentially a target of both cut and ttk. To approach this question, reporter gene expression of E8-2-46 was first examined in homozygous embryos for ct^c145 and ct^db7, which belong to the lethal II class, are the strongest cut mutants, and are likely to be null (JACK 1985 ; BODMER et al. 1987 ). In these backgrounds, the expression of the E8-2-46 reporter gene in the PNS is severely reduced (Fig 2A and Fig B). Normally, the reporter gene is expressed in 26 es support cells in each abdominal hemisegment (Fig 1A and Fig B). In cut mutants, es organs are often transformed into ch organs, albeit this phenotype is not completely penetrant (BODMER et al. 1987 ). Accordingly, the number of cells expressing the E8-2-46 reporter in ct^db7 is reduced to 15 cells (±2 SD, n = 15 hemisegments from five embryos), and the level of expression is significantly lower (Fig 2B; ct^c145 data are similar). In the complex es organs of the thoracic segments, the humidity receptive Keilin organs, the number of E8-2-46-expressing cells is not decreased significantly, only the level of expression (Fig 2E and Fig F), consistent with the observation that these organs are less affected in cut mutant embryos than simple es organs (BODMER et al. 1987 ).

Next, we examined E8-2-46 reporter gene expression in mutations of the lineage genes numb and ttk. In numb¹ mutants, the number of cells expressing the reporter gene increases (42 ± 4 SD, n = 15) as would be expected if the neural pIIb secondary precursors are transformed into pIIa, the support cell precursors (Fig 2C). Similarly, Keilin organ cells are also increased (Fig 2G). In ttk mutants the opposite phenotype is expected, since ttk is required for support cell development (GUO et al. 1995 ). Indeed, in ttk^702/7 mutants E8-2-46 expression per hemisegment is reduced (12 ± 3 SD, n = 10; Fig 2D; see also SALZBERG et al. 1994 ), similar in extent to the reduction observed in cut mutants (Fig 2B).

Cloning bereft:
Although the data indicate that cut and ttk are not likely the only factors that contribute to the control of E8-2-46 reporter gene expression in es support cells, they clearly play a crucial role. Therefore, we cloned the gene immediately 3' to the enhancer trap insertion, which is, most likely, responsible for driving expression of this enhancer trap. To isolate bft, the P element from E8-2-46 was used as a tool to recover 19.5 kb of genomic DNA flanking the insertion site (Fig 3). By generating probes from the genomic DNA, we attempted to localize bft transcripts by in situ hybridization (see MATERIALS AND METHODS). Of eight genomic fragments used, all six located centromere-distally to the insertion (R1–R4, D3, and D4; Fig 3) mimicked the E8-2-46 expression pattern, whereas the two proximal probes (D1 and D2) did not detect any transcript pattern by in situ hybridization (data not shown). Sequencing of the genomic DNA revealed an overlap with a contig of P1 genomic clones (DS06189, DS04362, and DS07071; GenBank accession no. AC006240). Using the genomic DNA as a template, we screened several cDNA libraries but obtained only one species of cDNA [1.5 kb, including a poly(A) tail indicative of a 3' end] located 12 kb distal to the P-element insertion (C91; Fig 3). The in situ pattern obtained with this cDNA reflects precisely the expression of E8-2-46 (with the exception of the anterior spiracle primordia; Fig 1, A–C; Fig 6A and Fig B; data not shown), suggesting that this cDNA is likely part of the bft transcript(s).

View larger version (95K): In this window In a new window Download PPT slide   Figure 6. bft expression in the developing embryonic PNS. (A) lacZ expression in bftE8-2-46 stage 16 embryo. (B–E) In situ hybridization with exon 1 (R1) of wild-type (B) stage 15, (D) stage 9, and (E) stage 11 embryos. Note the punctate expression in the PNS and at the anterior and posterior regions. (C) Homozygous bft225 stage 13 embryo hybridized with a C91-derived antisense RNA probe. Note the absence of specific staining in bft225, suggesting that only the 7-kb (and perhaps the 3.5-kb) but not the 0.8-kb transcript is specifically localized in wild-type embryos. (D) Early bft expression in the acron (arrowhead), the ventral head region near the anterior lip of the cephalic furrow (solid arrow), and posterior spiracle anlagen (open arrow). (E) bft expression in the developing PNS of the gnathal segments (curved arrows). (F–L) bft expression visualized with the C91 probe in conjunction with Cut protein in wild-type embryos. (F) Stage 10 embryo showing bft and Cut expression in the posterior spiracle (open arrow) and Cut only in the anlagen of the Malpighian tubules (solid arrow). (G and H) Early stage 11, (I and J) middle stage 11, and (K and L) late stage 11 embryos showing coexpression of bft and Cut in the sensory organ precursors (G–L) or some of their progeny (I–L) of the PNS of the body segments (arrowheads) and the gnathal segments (curved arrows). Open arrow indicates posterior (G and I) or anterior spiracle (L). H, J, and L are higher magnification views of G, I, and K, respectively. All are lateral views; anterior is left and dorsal is up.

Using C91 as a probe on developmental Northern blots, a transcript of 7 kb (and of 0.8 kb) is detected (arrows in Fig 4A), indicating C91 is not full length. We isolated additional elements of bft's coding region proximal to C91 by RT-PCR and 5' RACE (see MATERIALS AND METHODS). Northern analysis and sequencing of these PCR products suggest the existence of three transcripts (Fig 3 and Fig 4). The smallest 800-bp transcript was detected with the original C91 cDNA, but not with the more proximal PE1 or PE2' probes (Fig 4). Since this transcript is present at all stages and not affected in bft mutants, it is likely the result of cross-reactivity and thus may not be associated with the bft locus. The longest, 7-kb transcript is detected using three different probes, PE1, PE2', and C91. It is most abundant in 6- to 8-hr embryos (stage 11), which corresponds to the time when the PNS precursors divide and show expression by in situ hybridization. The 7-kb transcript is also expressed at pupal stages, again correlating with the development of the adult sensory organs. RT-PCR products also suggested a potential transcript of 3.5 kb, which was, however, not reliably detected using Northern hybridization (i.e., only after long exposures; data not shown), suggesting that it may be less abundant than the 7-kb transcript. Analysis of bft's exon/intron structure reveals that the 3.5-kb transcript contains three exons, whereas the 7-kb transcript contains only two (Fig 3). 5' RACE experiments using two separate sets of primers indicate the 5' end of the transcript is located 37–41 bp upstream of the first exon's potential ORF and 39–43 bp distal to the P-element insertion. Using the neural network promoter prediction program (REESE and EECKMAN 1995 ; REESE et al. 1996 ), a putative transcription initiation site is predicted within 4 bp of the 5' end obtained with 5' RACE, providing additional support for having identified bft's transcript start site.

Bidirectional sequencing of the RT-PCR products corresponding to the 7-kb and the 3.5-kb transcripts revealed four short ORFs (Fig 5). The first exon (common to both transcripts) contains a 465-bp ORF, but no start ATG. Two putative ORFs are found in the second exon of the 7-kb transcript, which do not occur in the 3.5-kb transcript (242 and 305 bp), and the fourth (280 bp) is again common to both transcripts. All sequences and conceptual translations were blasted against the GenBank database but no homologies or motifs were found. Thus, bft either encodes a short, novel protein or perhaps may function as an RNA.

Part of bft's first intron matches the 5' and 3' ends of a 4.5-kb expressed sequence tag (EST; LP 06727) isolated by the Berkeley Drosophila Genome Project from a directionally cloned library. In this EST, the longest ORF (317 bp) also has no ATG (Fig 5), and no transcripts are detected on Northern blots (data not shown). However, anti-sense riboprobes from this EST by whole mount in situ hybridization do show a bft-like expression (data not shown), but in a punctate pattern that is typical of nascent transcripts. Our results do not rule out the possibility that this EST may encode a short protein or a noncoding RNA as well. The genome project has not identified any protein-coding regions within the bft locus in the direction of bft transcription (ADAMS et al. 2000 ). In the opposite direction, the genome project predicts an ORF that falls within the LP06727 sequence, but corresponding riboprobes do not detect any expression in whole mount in situ hybridization (data not shown). The next closest known protein-coding regions are 7 kb (distal, odorant receptor 33C) and 10 kb (proximal, - trehalase) from the bft locus.

Developmental pattern of bereft expression:
The course of bft expression during sensory organ development was visualized in whole mount wild-type embryos either by itself or in combination with Cut protein (Fig 6). bft transcripts in the PNS coincide with the onset of Cut protein expression in some PNS precursor cells, suggesting bft is already turned on in neural progenitor cells as cut is (Fig 6, G–J). bft transcripts appear to be punctate and perinuclear in the vicinity of nuclear Cut staining. After es organ precursors have begun dividing, bft expression levels are sometimes higher in cells next to strongly Cut-positive nuclei (Fig 6K and Fig L), consistent with the observation that reporter gene expression in the E8-2-46 enhancer trap line is higher in forming trichogen than tormogen cells (the opposite is the case for Cut; see Fig 1).

In addition to the PNS, bft is highly expressed in the head and terminal regions (Fig 6B). bft expression first appears at stage 6 in the cephalic region of the future posterior transverse furrow and of the acron primordia (data not shown), which persists until after the clypeolabrum has formed. At stage 8/9, bft-expressing cells appear ventrally in the head, at the anterior lip of the cephalic furrow (Fig 6D), which then appear to invaginate during head involution. At early stage 11, bft RNA is present in two stripes of cells corresponding to the anlagen of the pharyngeal ridges. Later during stage 11, the expression expands to include strong staining in the maxillary and labial lobes and weaker staining in the mandibular lobe (Fig 6E). Most of this staining in the gnathal segments persists throughout embryonic development and probably corresponds to PNS precursors (such as the antenno-maxillary organ), which also express Cut (Fig 6I and Fig K; see also BLOCHLINGER et al. 1990 ). As the hypopharyngeal lobes form, they also express bft (data not shown). In the terminal, proctodeal region of stage 10 embryos, bft transcripts appear in endodermal cells corresponding to the anlagen of the posterior spiracles (Fig 6D), which also express Cut (Fig 6, E–G). The primordia of anterior spiracles begin to express bft only at late stage 11 (Fig 6K and Fig L).

cut regulates bereft expression levels:
We explored whether cut function activates or modulates bft transcription in the PNS, by examining cut^db7 null mutant embryos. In the absence of cut function, E8-2-46 reporter gene expression is reduced in es support cells (Fig 2C). In wild-type embryos, bft is also expressed in the developing posterior spiracles, which is severely reduced or absent in cut mutants (Fig 7A and Fig C).

View larger version (105K): In this window In a new window Download PPT slide   Figure 7. bft expression in loss-of-function and gain-of-function mutants for cut. bft transcripts are as detected with a C91 probe in stage 15 embryos in lateral (A, C, and E) and ventral views (B, D, and F). (C) bft expression in posterior spiracles (open arrows) is absent in ctdb7 embryos (right) as compared to wild type (left). Homozygous cut mutant embryos are identified by the absence of FM7-PlacZ expression (brown CNS staining in the wild type on the left). (E) When cut is overexpressed in alternate segments using the Hairy-Gal4 driver, high levels of bft expression are induced, most noticeably in the dorsal and lateral cluster in abdominal segments 3 and 5 (arrowheads). (B) In wild-type embryos, bft expression is present in the cells of the ventral PNS, but absent from the CNS or ectodermal cells overlying the CNS (between braces). When cut is overexpressed ubiquitously using the 69B-Gal4 driver (D) or in alternate segments using the Hairy-Gal4 driver (F) high levels of bft expression are induced in ectodermal stripes overlying the CNS (arrowheads).

Next, we wanted to determine if cut suffices to activate bft transcription. We used the UAS-Gal4 system (BRAND and PERRIMON 1993 ) or a heat shock promoter to drive cut ectopically in embryos and to examine the resulting pattern of bft transcription. When we use hairy-Gal4 to drive cut expression in the odd-numbered segments, bft is expressed ectopically, most noticeably near the dorsal abdominal PNS clusters, which are the embryonic origins of the lateral chordotonal organs (Fig 7E). As previously noted, ectopic expression of cut causes es-specific gene expression in these chordotonal organs and prevents their lateral migration (BLOCHLINGER et al. 1991 ). Thus, the cell fate changes in the PNS induced by cut result in ectopic bft expression. Furthermore, cells that normally never express cut, in particular ectodermal cells overlying the central nervous system (Fig 7B), were induced to express bft when cut was ectopically expressed (Fig 7D and Fig F). Since cut can induce ectopic bft expression outside the PNS, it may participate directly in the regulation of bft. Consistent with this hypothesis, we have identified consensus Cut binding sites (ANDRES et al. 1994 ) immediately upstream of the bft 5' RACE products (data not shown).

Generating bereft mutants:
Since the E8-2-46 flies are viable and exhibit no visible phenotype, we sought to generate mutations in the gene responsible for the bft expression pattern by excising the P element (see MATERIALS AND METHODS). Often, these excisions are imprecise, resulting in the loss of flanking sequences. A total of 244 fly strains were generated in which the P element had excised or had excised and reinserted. Candidate alleles were detected by identifying those strains in which the DNA was disrupted (PCR screening), or they were detected by examining es organ structures for defects. Twenty-one mutant strains were recovered that contained small deletions or that exhibited defective sensory organs or both. The 7 strains that were chosen for further study have reduced viability (bft⁹⁷ is lethal), form a single complementation group (one allele, bft¹²², is complex in its complementation pattern), and exhibit a similar bristle phenotype (Table 1). In bft⁶ and bft²⁴, genomic lesions have been identified that eliminate the first (and longest) putative ORF and the transcript start (Fig 3). bft⁶ contains a deletion of 1.6 kb that removes sequences distal to the site of the P-element insertion, eliminating bft's first exon and 0.75 kb of intron 1. bft²⁴ lacks 2.8 kb of sequence, extending not only distal but also proximal to the insert, removing both bft's first exon and 1.5 kb of bft's first intron. Furthermore, in bft²⁴ (and in bft²²⁵), the 7-kb transcript is missing (Fig 3D and Fig E; data not shown). Thus, the 7-kb bft transcript is disrupted in the mutant alleles examined.

bereft is required for morphogenesis of the interommatidial bristles:
Since the majority of bft alleles are viable, although at a reduced level, we examined adults for defects in bristle morphogenesis. The es organs predominantly affected in bft alleles are the interommatidial bristles (IOB) of the eye (Fig 8). The Drosophila eye consists of 750 hexagonal ommatidia (CAGAN and READY 1989 ), each of which secretes a lens and is surrounded by three mechanosensory IOBs that project from alternate vertices of the hexagonal array (Fig 8A and Fig C). Similar to other simple bristle-type es organs of the fly, each IOB contains a shaft-forming trichogen and a socket-forming tormogen cell and is innervated by a sensory neuron.

View larger version (136K): In this window In a new window Download PPT slide   Figure 8. bft is required for interommatidial bristle morphogenesis. (A–F and H) Scanning EM pictures of IOBs in w (A and C; wild-type phenotype) and homozygous bft mutant eyes (B and D–F). Note that most of the IOB's bristle shafts are missing and sockets have an abnormal morphology. (H) Ommatidia of a fly carrying overlapping deficiencies in the 33B region (see text and Fig 9). The IOB phenotype and viability of these trans-heterozygous flies are comparable to the bft alleles examined. (G) Eye disc 24 hr after puparium formation of a bft24 homozygote stained with Cut antibody. The pattern of four cells stained per IOB es organ is indistinguishable from wild type. (Inset) Confocal image of an IOB es organ expressing Cut (green) and lacZ reporter of E8-2-46 (red). lacZ is expressed in three of the four IOB cells.

In the bft mutants, the majority of the IOBs are missing (Fig 8A and Fig B). bft²⁴ and bft²²⁵ are the strongest alleles, in that each fly lacks 50–90% of the normal complement of IOBs. In most cases, severely defective structures are found where the IOBs normally form (Fig 8, C–F and H). The most severe defect is the complete absence of shaft and socket morphogenesis, resulting in a slight bump or cap in a shallow pit, without any other distinguishing characteristics (Fig 7E and Fig H). Other structures found in bft mutants were a relatively normal socket and a round, spherical shape protruding from it (Fig 8D), reminiscent of mechanosensitive campaniform sensilla, found in other regions of the fly (see Fig 9C, p3 and p4). Another phenotype consists of discontinuous sockets, as if they are composed of two halves, without any remnant of a shaft, as if the shaft were transformed into another socket (Fig 8F). To determine if the precursors of the IOBs form in these flies, pupal eye discs were stained with Cut antibodies to visualize the precursor cells and their progeny. In wild-type flies, all four IOB sensillum cells express Cut; in bft mutants, these cells express Cut normally (Fig 8G), suggesting that bft is not required to produce the normal number of Cut-expressing progeny. Thus, bft must act at a later step in IOB differentiation. Interestingly, as is observed in the embryo, the presumptive trichogen cells within the forming IOBs express the E8-2-46 reporter most strongly (Fig 8G, inset). This prevalent expression in the shaft-forming cell may reflect the possibility that one of bft's crucial functions is in bristle morphogenesis.

View larger version (137K): In this window In a new window Download PPT slide   Figure 9. bft affects adult and larval bristle formation. (A and B) Head and (C and D) larval bristles are often missing in bft mutants (indicated by asterisks in B and D).

bereft is required for the morphogenesis of adult and larval trichoid sensilla:
We also examined bft mutants for defects in mechanosensory bristles of the head, thorax, abdomen, and legs. While wild-type flies occasionally lack vertical bristles (4% of flies lack one or more, n = 69), postvertical or humeral bristles were never missing. bft homozygous mutants lack bristle shafts on the head and thorax at a significantly higher incidence than wild type (Fig 9A and Fig B; Table 2). The vertical, postvertical, and humeral bristles were most often missing in bft mutants (up to 80% of the flies lack one or more of these bristles). Notably, the sockets of the missing bristle shafts in bft mutants are still present and normal in appearance, even when examined with scanning electron microscopy (Fig 9B; data not shown). These results again suggest that primarily bristle shaft formation is affected in bft mutants and that the observed high levels of bft activity in trichogen cells (Fig 1) may be required autonomously for this process.

Since bft is already expressed early during sensory organ formation, much before bristle morphogenesis, we examined the formation of different PNS cell types in bft mutant embryos. We find that not only Cut shows a normal distribution in all es organs, but also Su(H) (nuclear marker for socket-forming tormogen cells), 22C10 (shows characteristic es neuron morphology), and 21A6 (shows typical sensory structures associated at the tip of es neuronal dendrites; see BODMER et al. 1987 ) in bft²⁴, bft²²⁵, and bft⁹⁷ mutant embryos (data not shown).

Although no es organ defects can be detected in bft mutants during embryonic stages, a requirement for bft in bristle morphogenesis might manifest itself during larval stages. Indeed, third instar bft larvae often exhibit abnormal trichoid sensilla in which the shaft is missing, similar to the adult phenotype (Fig 9C and Fig D). In some cases, the sensory structure resembles that of a campaniform sensillum similar to what was observed with IOBs (Fig 9D and inset). The sensory organs established in the embryo further differentiate during larval stages. Thus, we reasoned that sensory organ defects associated with bft mutant alleles might be detected during larval stages. We began by examining first instar larvae, but found no defects in their sensory organ structure. However, third instar bft larvae often exhibit abnormal trichoid sensilla in which the shaft is missing, similar to the adult phenotype (Fig 8C and Fig D, compare with Fig 8A and Fig B). In some cases, the sensory structure resembles that of a campaniform sensillum (Fig 8D and inset). A function reminiscent of bft has been found for the paired homeobox gene pox neuro, which is expressed in one of the es support cells during larval stages. Interestingly, in pox neuro as in bft mutants, not only do the trichoid sensilla show bristle shaft abnormalities, but these defects do not manifest themselves earlier than in second instar larvae (AWASAKI and KIMURA 2001 ).

Bristle defects result from a molecular lesion of the bereft locus:
Although some of the E8-2-46 P-element excision alleles we generated do have molecular lesions at the bft locus (Fig 3), it is conceivable that the observed bft bristle phenotype of these alleles is caused by a background mutation in the E8-2-46 enhancer trap stock. To determine whether the observed defects in bristle morphogenesis are indeed associated with the bft locus, we crossed cytological deficiencies at chromosome position 33B (LINDSLEY and ZIMM 1992 ) to our excision alleles. Df(2L)escP3-0, Df(2L)esc10, and Df(2L) prd1.7 fail to complement the reduced viability of bft mutants (Table 1). The reported cytology of these deficiencies suggests that they overlap only in the 33B region (Fig 10; LINDSLEY and ZIMM 1992 ). Flies trans-heterozygous for our bft alleles and any of these three deficiencies exhibit a loss of IOBs, as is typically observed in bft homozygous flies (Fig 8; Table 1). Thus, the observed bristle phenotype is unlikely due to a background mutation, but rather due to a lesion at the bft locus. Df(2L)esc10 and Df(2L)prd1.7 both affect the bft locus but their deficiencies extend in opposite chromosomal directions (Fig 10). Therefore, we reasoned that the deficiency overlap may be small enough to yield some viable trans-heterozygotes that lack the bft locus. Indeed, survivors of the genotype Df(2L)esc10/Df(2L)prd1.7 do display the bft eye phenotype and show an almost complete absence of IOBs (Fig 8H, Table 1). The complementation pattern of the lethal allele bft⁹⁷ with the large deficiencies of the bft locus suggests it contains a larger deletion than bft⁶ and bft²⁴ (Table 1; Fig 10). As assessed by PCR, bft⁹⁷ lacks genomic DNA centromere-distal to the site of the E8-2-46 insertion corresponding to the location of the bft locus (data not shown). Furthermore, bft⁹⁷ but none of the other bft alleles fails to complement the lethality of Df(2L) escP2-0 (Table 1; Fig 9). These data provide strong evidence that the bft phenotype results from a loss of gene activity in the 33B region. Moreover, the fact that bft⁶, bft²⁴, and bft⁹⁷ contain molecular lesions of the bft locus, removing the first exon or likely the entire locus, and that they do exhibit the same phenotype in trans-heterozygous combination with deficiency flies, leaves little doubt that a defect in the bft locus is responsible for the observed phenotype in bristle morphogenesis.

View larger version (15K): In this window In a new window Download PPT slide   Figure 10. Diagram illustrating the approximate location and/or overlap of the deficiencies of the 33B region. Trans-heterozygote flies harboring the deficiencies Df(2L)esc10 and Df(2L)prd1.7 are semiviable (see Table 1) and exhibit the bft IOB phenotype (Fig 7H), indicating that the overlap removes the bft locus but no other genes that are essential for viability. Conversely, trans-heterozygotes containing Df(2L)escP2-0 and Df(2L)prd1.7 are fully viable and have no apparent phenotype (see Table 1).

bereft interacts genetically with cut and ttk:
We also explored the relationship between bft, cut, and ttk in genetic interaction experiments. For this purpose, we generated flies of the genotype ct^c145/FM6;bft⁶/CyO. These strains never yield bft⁶ homozygous females that were also heterozygous for cut (e.g., ct^c145/FM6;bft⁶/bft⁶, n = 300), although bft⁶ homozygotes are semiviable (data not shown; see also Table 1). Thus, mutating one copy of cut eliminates the viability of bft⁶. In an attempt to generate viable bft flies that lack some but not all cut function, bft⁶ was crossed to the viable ct^k allele, which by itself exhibits bristle and wing margin defects (BODMER et al. 1987 ). Similarly, a stock of ct^k/FM6;bft⁶/CyO flies never produces any bft⁶ homozygous females (n = 300; data not shown). In contrast, some ct^k/ct^k;bft⁶/CyO females do survive and exhibit a typical ct^k phenotype. To address the possibility that the interaction between bft and cut may be attributable to a background mutation in bft⁶, we crossed ct^k/ct^k;bft⁶/CyO females to bft²⁴/bft²⁴ homozygotes. Indeed, female ct^k/+;bft⁶/bft²⁴ survivors are observed, and they exhibit the bft eye phenotype (Table 3). However, male hemizygotes of this cross that are also trans-heterozygous for these two bft alleles, ct^k/Y;bft⁶/bft²⁴, are never observed (n = 98). Similar experiments were carried out with another bft allele, bft²²⁵. When bft²²⁵ is crossed into a ct^k mutant background, neither doubly homozygous females (ct^k/ct^k;bft²²⁵/bft²²⁵) nor males hemizygous for ct^k and homozygous for bft²²⁵ (ct^k/Y;bft²²⁵/bft²²⁵) are observed (Table 3).

To determine if removing ttk function augments the bft phenotype, we crossed bft⁶ into a ttk^702/7/+ background. Indeed, survivors of the genotype, bft⁶/bft⁶;ttk^702/7/+, were never observed (n = 98; data not shown). Thus, losing one copy of ttk is completely fatal for bft⁶ flies. Taken together, these findings demonstrate that cut and ttk exhibit genetic interactions with bft, consistent with the idea they affect some of the same developmental pathways.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

bft encodes a novel gene required for proper bristle morphogenesis. Flies lacking bft exhibit defects in the mechanosensory bristles of the head, thorax, and IOBs of the eye. Consistent with its function, bft is expressed in the sensory structure-forming support cells of these es organs, in particular at high levels in the bristle shaft-forming trichogen cell. Furthermore, bft expression depends on both cut and ttk, and bft expression can be driven ectopically by cut overexpression, suggesting these two genes are involved in regulating bft.

The cells comprising the IOB do form in bft mutants, but the cuticular structures they secrete are severely defective. These observations indicate that bft may be required to direct the secretion of the cuticular shaft (and socket) structures. The shaft is formed from a cytoplasmic extension of the trichogen cell, and its structure is provided by a core of microtubules surrounded by actin fiber bundles (OVERTON 1967 ). A number of different genes encoding actin-associated proteins have been shown to affect bristle morphology (VERHEYEN and COOLEY 1994 ; DYE et al. 1998 ); among them is sanpodo, a tropomodulin homolog, which also acts downstream of numb, as does bft.

Is the observed bristle phenotype in bft mutants due to a defect in the bft coding region?
Three lines of evidence indicate that the bristle phenotype observed in bft mutants results from a mutation in the bft gene. First, the tissues and cells in which bft transcripts are expressed are affected in bft mutant flies. bft is expressed in the precursor cells that secrete the sensory structures, consistent with bft being required for appropriate differentiation of these cells. Second, the alleles bft⁶, bft²⁴, and bft⁹⁷ contain molecularly characterized deletions of the bft coding region: bft⁶ and bft²⁴ contain deletions of 1.6 and 2.8 kb, respectively, that remove the first exon harboring the largest open reading frame, and bft⁹⁷ contains a larger deletion, probably removing the entire bft locus. Third, the 7-kb bft transcript is absent in bft⁶ and bft²²⁵ homozygotes. Taken together, this evidence strongly indicates that the bft phenotype results from a disruption of the bft locus and that it is likely that the absence of or a defect in the 7-kb bft transcript is the cause of the observed bristle phenotype. A further consideration is that bft alleles in trans to cytological deficiencies of the 33A-B genomic region do not noticeably increase the observed phenotypes, suggesting we have isolated strong bft alleles. However, without having corrected the phenotype using a bft transgene we cannot completely rule out the possibility that the molecular lesions of bft⁶ and bft²⁴ (also) affect a regulatory region of a distant gene. Centromere distal to bft are (or are predicted) odorant receptor 33C (Or33c; 7 kb 3' to the 7-kb bft transcript), Drosocrystallin (also known as Cry), and CG16964 (novel). Centromere proximal are - trehalase (similar to an enzyme involved in stress response in Saccharomyces cerevisiae, 10 kb 5' to the 7-kb bft transcript), CG6686 (predicted to be a cytoskeleton-associated protein with homologies to human and rodent tumor-rejection antigen SART-1), and CG12314 (novel). None of these genes, however, are predicted by the Drosophila genome sequence project to span the bft locus.

What does bereft code for?
Bft's longest transcript is 7 kb, but surprisingly the longest ORF we have identified is only 465 bp contained within the first exon (Fig 5). While other genes encoding small proteins have been reported, for instance reaper (WHITE et al. 1994 ), the size of this transcript makes it difficult to speculate which ORF is likely translated into a protein product. In addition, none of the sequences, at the nucleotide or amino acid level, exhibit any homology to other genes.

Considering the lack of an obvious ORF, the 7-kb bft transcript may not code for a protein product, but perhaps acts as an RNA. The mechanisms of action by noncoding, nonribosomal RNAs are poorly understood. A few apparently noncoding mRNAs have been proposed to act by hybridizing to the mRNAs of other genes, thereby preventing their translation. For instance, the lin-4 gene of Caenorhabditis elegans encodes small, noncoding transcripts that are thought to post-transcriptionally regulate the lin-14 gene (LEE et al. 1993 ; WIGHTMAN et al. 1993 ). In mammals, the Xist gene, involved in X chromosome inactivation (PENNY et al. 1996 ), appears to lack a coding region (BROCKDORFF et al. 1992 ; BROWN et al. 1992 ), and its transcript, like bft's, is quite large: 17 kb in humans and 15 kb in mice. In Drosophila, the rox1 and rox2 genes also appear to encode only RNAs and not proteins (AMREIN and AXEL 1997 ; MELLER et al. 1997 ), and they have a redundant but essential function in dosage compensation (FRANKE and BAKER 1999 ). Further experiments are needed to decide if bft acts primarily as a protein or as a noncoding RNA and what its mechanism of action is.

cut may participate in bft transcriptional activation:
By examining both Cut protein and bft transcripts in the same embryo, we have found that the es precursors express bft transcripts almost coincident with the onset of Cut expression. At later stages, bft transcripts are restricted to the support cells of es organs. Furthermore, bft transcripts are expressed in nonneural tissues that also express Cut, such as in the cephalic segments, and the precursors of both the anterior and posterior spiracles. In the absence of Cut activity, bft expression is reduced or absent. Conversely, the ectopic expression of Cut drives ectopic bft transcription. Moreover, consensus Cux/Cut-binding sites have been identified upstream of the bft transcript (ANDRES et al. 1994 ): ATC GATTA is found 600 and 660 bp upstream of the transcript start site, and a CCAAT motif, recognized by Cut repeat II, is also found near one of these sites. This, together with the overexpression data, suggests that Cut may activate bft transcription directly. However, Cut is unlikely to be the only factor regulating bft transcription, since in cut null mutants, bft expression is not completely absent.

bft may integrate information from the selector gene cut and the lineage gene ttk:
Our data suggest that bft may be responsive to both organ identity (cut) and lineage (ttk) information. Other candidate genes active in the Drosophila PNS that may respond to both the lineage and selector gene pathways include BarHI and BarHII (HIGASHIJIMA et al. 1992 ). These genes are also expressed specifically in es organs, as is bft, but in contrast to bft, they are present in the neurons and glia. The evidence presented here suggests bft is one of a group of genes that must be activated in es support cells to ensure their proper differentiation.

	FOOTNOTES

¹ Present address: Developmental Patterning Laboratory, Cancer Research UK, 44 Lincoln's Inn Fields, London WC2A 3PX, England.
² Present address: Carnegie Institute of Washington, 115 W. University Pkwy., Baltimore, MD 21210.

	ACKNOWLEDGMENTS

We thank Hugo Bellen, Seymour Benzer, Karen Blochlinger, Corey Goodman, Yuh-Nung, Lily Jan, and the Bloomington Stock Center for sending fly stocks or antibodies. We also thank Kenneth Cadigan for help with eye disc preparations and Krista Golden for excellent assistance in preparing the manuscript and the figures. This work was supported by a grant from the National Institutes of Health to R.B. Support for K.H. was provided by a National Science Foundation training grant to the University of Michigan.

Manuscript received December 8, 2000; Accepted for publication February 18, 2002.