Pattern formation in the limbs of Drosophila: bric à brac is expressed in both a gradient and a wave-like pattern and is required for specification and proper segmentation of the tarsus

Drosophila melanogaster has been used extensively as a model system for the study of development and pattern formation. Both the anterior-posterior and dorsal-ventral axis of the embryo are established by the formation of con-centration gradients of transcription factors which form by protein diffusion in the syncytial blastoderm of the embryo (for review see StJohnston and Nüsslein-Volhard 1992). In contrast, pattern formation in many other developmental systems occurs in a growing field that is cellularized. The mechanisms by which postional values along an axis are defined in a growing tissue is not well understood. One developmental system in which this problem can be examined is the proximal-distal axis of the limbs of Drosophila.

The legs of insects are composed of segmental units, which are, from proximal to distal, the coxa, trochanter, femur, tibia and tarsus. The tarsus of flies is further subdi-vided into five tarsal segments. The limbs of Drosophila develop from imaginal discs, small epithelial cell groups that invaginate from the epidermis of the embryo (Cohen, 1990; Bate and Martinez-Arias, 1991; Cohen et al., 1991), grow during the larval stages and give rise to the adult body structures during metamorphosis. In imaginal discs, the anlagen of the adult leg structures are concentrically arranged, with the anlage for the tarsus in the center and the anlagen for the more proximal structures toward the periphery (Schubiger, 1968). Morphologically distinct primordia of limb segments appear during the third larval instar, when the monolayered epithelium of the disc, while growing (for review see Bryant, 1978), forms concentric folds that represent the primordia of the segments, with the folds of the tarsal segments appearing last (Fristrom and Fristrom, 1975). Only 20-45 cells are initially incorporated into a leg imaginal disc (Madhavan and Schneiderman, 1977), a number that is too small to have already defined all the pattern elements of the adult leg. It has been suggested, therefore, that the early limb disc is provided with only a few positional cues that become more elaborated during sub-sequent growth of the disc (Schubiger, 1974; for review see Bryant, 1978; Cohen and Jürgens, 1989a).

A number of genes have been identified in Drosophila that are involved in pattern formation of the proximal-distal axis of the limb. These genes can be divided into two classes. The first class is characterized by an expression pattern in the form of a wedge-like sector or a longitudinal stripe in the leg disc, and by the fact that mutations in these genes cause loss of distal structures of the limb (for review see Bryant, 1993). Genes that belong to this class include the TGF-β homolog decapentaplegic (Spencer et al., 1982; Bryant, 1988; for review see Gelbart, 1989) and the segment-polarity genes hedgehog (Mohler, 1988) and wingless (wg; Baker, 1988a; Couso et al., 1993). wg has been shown to be required for dorsal-ventral organization in imaginal discs by expression studies and clonal analysis (Couso et al., 1993) and has been suggested to define a con-centration-dependent ventralizing signal based on experiments of ectopic wg expression in the leg disc (Struhl and Basler, 1993). The analysis of this class of genes suggests that the proximal-distal axis depends on the formation of positional values that are oriented circumferentially around this axis (anterior-posterior and dorsal-ventral values). A second class of genes, which are expressed in circumferen-tial stripes (rings) in the leg imaginal disc, appears to be required specifically for formation of positional values along the proximal-distal axis (for review see Bryant, 1993). One member of this class is Distal-less (Dll, also called Brista;Sunkel and Whittle, 1987; Cohen and Jürgens, 1989a), a homeobox gene (Cohen et al., 1989) that is required for the development of all leg segments except the most proximal one (Cohen and Jürgens, 1989b). The requirement for Dll is graded within the developing leg, with the tarsus being the structure most sensitive to a loss of Dll activity. A global pattern-organizing function has been suggested for Dll along the proximal-distal axis (Cohen and Jürgens, 1989a; Cohen et al., 1989).

The gene bric à brac (bab), which we describe here, belongs to the second class of genes and is specifically involved in the development of the tarsus. Mutations in bab cause homeotic transformation of the bristle pattern of tarsal segments as well as defects in or loss of intersegmental joints in the tarsus. The distribution of the bab gene product develops from a rather uniform pattern into a wave-like pattern with the peaks of expression level corresponding to the segments and the valleys corresponding to the interseg-mental joints during the last larval stage, a period in which the tarsal primordium more than doubles its size. The peaks of expression level form a gradient from distal to proximal in the tarsal primordium. This gradient will be discussed in relation to the graded requirement for bab function in tarsal segment specification along the proximal-distal axis.

The mutant bab alleles bab^E1, bab^P, bab^PRDS, bab^PR11, bab^PR23, bab^PR24, bab^PR72 and the deficiency Df(3L)bab^PG were isolated during this study. The mutant chromosomes carried recessive genetic markers and were maintained as TM3, Sb ry^rRK or TM6B, Tb e ca balanced stocks. Balancer chromosomes and genetic markers are described in Lindsley and Zimm (1992). Enhancer trap line P[lacZ, ry⁺]A30 is described in O’Kane and Gehring (1987), and P[lArB]A128.1F3 in Bellen et al. (1989). The translocation T(3;Y)A114 was isolated by Lindsley et al. (1972). Oregon R was used as a wild-type stock.

1. P-element mediated mutagenesis. The enhancer trap line bab^P was isolated as a recessive female semisterile mutation in a P-element mediated mutagenesis, using P[lacZ, ry⁺] as an enhancer detector (O’Kane and Gehring, 1987) and P[ry⁺(Δ2-3)]99B as a transposase source (Laski et al., 1986; Robertson et al., 1988) in a ry⁵⁰⁶ background.

2. P-element mediated reversion of bab^P. To ascertain that the mutant phenotype of bab^P is caused by the P[lacZ, ry⁺] insertion, the P-element was remobilized by transposase. Males of the genotype bab^Pry e^sca/ Ki p^p P[ry⁺(Δ2-3)]99B were crossed indi -vidually to ry⁵⁰⁶ females. Progeny were screened for reversion of the ry⁺ phenotype. About 200 independent male revertants were tested for complementation of the bab^P mutant ovary phenotype. 31 revertant lines failed to complement bab^P, among which, 24 showed a stronger mutant phenotype than bab^P. The new bab alleles are designated as bab^PR. Some of the remaining lines that complemented bab^P were analyzed and found to be phenotypically wild type; demonstrating that the bab mutant phenotype is caused by the P-element insertion.

3. Df(3L)bab^PG was isolated in a gamma-ray mutagenesis. 0-48 hours old homozygous bab^Pry e^sca males were irradiated by 4000 rad gamma-rays and immediately crossed to ry⁵⁰⁶ females. F₁ males were screened for reversion of the ry⁺ marker. Revertant chromosomes were checked for complementation against bab^P. From 7000 irradiated chromosomes, one deletion, Df(3L)bab^PG [61D3-E1;61F5-8], was isolated.

4. The bab^E1 mutation was isolated as an EMS allele based on its dominant leg phenotype. The EMS mutagenesis was performed according to Lewis and Bacher (1968). ri e males were mutage-nized and crossed to wild-type females. The progeny were raised at the non-permissive temperature (30°C) for the dominant phenotype. F₁ males were screened for ectopic sex combs on the prothoracic legs.

The temperature-sensitive dominant phenotype of the bab^PRDS allele was used for temperature-shift experiments. Wild-type females were crossed to homozygous bab^PRDS males and were allowed to lay eggs in vials at 25°C for 4 hours. The eggs were kept at 25°C for 24 hours and the 1st instar larvae transferred to either 18°C or 30°C (±0.5°C). A temperature shift was performed at different developmental stages (Fig. 2), and the flies were kept at the shifted temperature until they hatched. The heterozygous male progeny were screened for ectopic sex combs on the protho-racic legs under a dissecting microscope at 20-32× magnification. Both legs were screened because the phenotypic strength of two legs of a fly differ in the same manner as do legs from different flies at a given temperature. The penetrance of ectopic sex combs as well as the number of sex comb bristles per ectopic sex comb were counted.

Wild-type and mutant flies were raised at 25°C. Large numbers of flies per genotype were screened for ectopic sex combs and fusion of segments under a dissecting microscope. For a detailed analysis of the bristle pattern of legs and antennae were mounted in Hoyer’s medium (Van der Meer, 1977). The cuticle preparations of legs and antennae of 15-20 males and females, respectively, per genotype were viewed with a Zeiss Axiophot microscope.

Eggs were collected on apple juice agar plates (Wieschaus and Nüsslein-Volhard, 1986) in 2 hours intervals, and the hatched larvae were fed yeast while growing on the plates at 25°C. Staging of larvae and prepupae was based on the descriptions of Ashburner (1989).

The P-lacZ insertion of the bab^P allele was used as a tag for the cloning of sequences adjacent to the insert by the inverse PCR method (Ochman et al., 1990). The inverse PCR product was sequenced to determine the insertion site of the P-lacZ construct and was used as a probe to screen a lambda genomic library (Tamkun et al., 1992) using standard methods (Sambrook et al., 1989). The genomic 4.2 kb EcoRI restriction enzyme fragment E199, into which the P-lacZ element inserted, was used to screen a number of cDNA libraries. The cDNA clones were characterized by restriction enzyme mapping and DNA sequencing (Sanger et al., 1977). The DNA sequence was analyzed using DNA Strider (written by Christian Marck, Department de Biologie, Institut de Recherche, Commissariat à L’Energie Atomique, France) and the University of Wisconsin software package (Devereux et al., 1984). The 1.7 kb open reading frame was compared with sequences in the GenBank and EMBL data bases using the FASTA program (Pearson and Lipman, 1988).

A biotin-labeled DNA probe, using a genomic DNA fragment adjacent to the bab^P insertion site, was used for chromosome in situ hybridization (Ashburner, 1989). In situ hybridizations to imaginal discs were performed with the 4.2 kb genomic EcoRI fragment E199 and the 1.2 kb subclone 504 (Fig. 6A). The procedure followed a modified version of the protocol of Tautz and Pfeifle (1989) using the Genius Kit of Boehringer-Mannheim. Dis-section and fixation were done as described for antibody staining and Proteinase K treatment for 5 minutes. Hybridization with digoxigenin-labelled DNA probe (about 50 ng DNA/100 μl hybrid-ization solution) was performed at 50°C for 16 hours and followed by washing in hybridization solution, 1:1 mixture of hybridization solution and PBT, and PBT at 50°C each for 1 hour. The hybridized imaginal discs were dehydrated and mounted as described for antibody-stained discs.

The BamHI fragment from the subclone 504 (Fig. 6A) that encodes a 24×10³M_r protein was subcloned into the pGEX-1N (Amrad) expression vector. Expression of the glutathione S-transferase fusion protein and purification of the soluble 50×10³M_r fusion protein were performed according to Smith and Johnson (1988). The fusion protein (1 mg/ml 50 mM Tris-HCl pH 7.5) was sent to Pocono Rabbit Farm and Laboratories (PO Box 240, Dutch Hill Road, Canadensis PA 18325) for immunization of rats. The sera were provided with 0.01% Timerosal and tested by tissue antibody staining.

Imaginal discs were dissected from 3rd instar larvae and prepupae in PBS and kept on ice until fixation in 5% formaldehyde (Poly-sciences), 50 mM EGTA in PBS for 20 minutes. The tissue was incubated in methanol for 3-5 minutes and washed in PBT (PBS, O.1% Tween-20) for 30 minutes, followed by a 30 minutes incu-bation in PBTBS (PBT, 0.1% BSA (Sigma, globulin-free), 2% goat serum (Gibco)). Incubation with the first antibody, anti-β-gal (Cappel, polyclonal IgG) in a 1:3000 dilution in PBTBS or the polyclonal rat anti-BAB r2 antibody in a 1:5000 dilution in PBTBS, which was preabsorbed on fixed embryos for 1-2 hours, was done at 4°C overnight. The discs were washed in PBT for 4×15 minutes, incubated in PBTBS for 1 hour and subsequently in the secondary biotinylated antibody, anti-rabbit (Vector Laboratories) or anti-rat (Jackson Laboratories) in a 1:500 dilution for 2 hours at room temperature. After washing in PBT for 4×15 minutes the discs were treated with biotinylated horse radish peroxidase (HRP)-avidin complex (Vector Laboratories, ABC Kit) in a 1:50 dilution in PBT for 1 hour and washed again 3×15 minutes in PBT. HRP reaction was performed in staining solution containing 0.2 mg DAB (Sigma) and 0.01% H₂O₂/ ml PBT. The discs were dehy-drated and mounted in a 1:4 mixture of Canada balsam (Sigma) and methyl salicylate (Sigma) for whole-mount views or were embedded in an Epon/Araldite mixture (Polysciences, Ted Pella) for histological sections. A complete series of semi-thin sections (5 μm) of 8-10 imaginal discs were prepared for each chosen devel-opmental larval or prepupal stage.

Sections of anti-β-gal-stained bab^A128/+ imaginal discs were viewed using bright-field illumination on a Zeiss Axiophot micro-scope equipped with a image processing system (Hamamatsu Argus-10). The intensity of staining was linearly processed into different colors. For each cell, a value betweeen 0 and 10 was defined by reading the color directly from the monitor. 1-2 longitudinal sections from 10 different leg imaginal discs were examined.

The first bab allele, bab^P, was isolated from an enhancer trap screen as a female-semisterile mutation. The French name bric à brac refers to the disorganized structure of the bab mutant ovaries. Twenty-fouradditional bab alleles that have a stronger female-sterile phenotype were isolated by remo-bilization of the P-element insert in the bab^P line. Exami-nation of these alleles showed that they also cause defects in the leg and the antenna of the fly. An additional mutation, the EMS-induced bab^E1 mutation was identified as a bab allele on the basis of its mutant phenotype and because it fails to complement other bab alleles. A deficiency, Df(3L)bab^PG (61D3-E1; 61F5-8), was isolated that uncovers the bab locus which was mapped to the cytological position 61F1-2 (Fig. 6C). Two more P-element insertions, P[lacZ, ry⁺]A30 and P[IArB]A128.1F3 (the latter designated as bab^A128) map to the same chromosomal position (O’Kane & Gehring, 1987; Bellen et al., 1989) and express the lacZ reporter gene in a pattern similar to bab^P but do not cause a mutant phenotype in homozygous flies. The isolated bab mutations were classified as having either a strong, inter-mediate or weak mutant leg phenotype according to the severity of the leg defects. The homozygous mutant phenotype of strong bab alleles is slightly weaker than the transheterozygous phenotype of these alleles over Df(3L)bab^PG suggesting that these bab alleles are strongly hypomorphic. The bab alleles that were chosen for pheno-typic analysis are listed in Table 1.

Homozygotes and transheterozygotes of strong bab mutations show a change of the bristle pattern of tarsal segments (TS) TS2, TS3 and TS4 towards the most proximal tarsal segment, TS1. TS1, also called the basitarsus, can be distinguished from the other tarsal segments by specific bristle markers. The most prominent marker of TS1 is the sex comb on the prothoracic (front) legs of males. The sex comb is a longitudinal row of about eleven blunt, black sex comb bristles (SCB), which is located in the distal region of TS1 (Hannah-Alava, 1958; Fig. 1A,C). In addition to the normal sex comb on TS1, males homozygous for a strong bab mutation exhibit ectopic sex combs on TS2, TS3 and occasionally TS4 of the prothoracic legs (Fig. 1B,E). The ectopic sex combs were defined as such based on the color, shape, orientation and position of the bristles. The ectopic sex comb on TS2 usually contains 5-6 SCB, half as many as is present in the normal sex comb on TS1. The ectopic sex comb on TS3 is even smaller and usually contains 1-2 SCB, the one on TS4 not more than one SCB (Fig. 1B,E). The appearance of ectopic sex combs in distal tarsal segments indicates a homeotic transformation of the bristle pattern of TS2, TS3 and TS4 towards TS1.

This conclusion is supported by the observation that the homeotic transformation is not limited to the SCB but includes other characteristics of TS1, and that the transfor-mation is neither sex nor leg specific. In wild-type flies, the ventral bristles of the prothoracic TS1 are arranged in tightly packed transverse rows (Hannah-Alava, 1958; Fig. 1G). In contrast, the distal tarsal segments of the prothoracic legs have separated longitudinal columns of single bristles (Fig. 1G). In strong bab mutant flies, both male and female, the normal bristles on the ventral side of TS2 and TS3 are replaced by a larger number of bristles that are arranged in such transverse rows (Fig. 1H). The analysis of the bristle pattern of TS4 was difficult because of its small size and because it is fused to TS5 in strong bab mutants; however, the bristle pattern of TS4 is clearly affected by bab mutations as indicated by the occasional appearance of a SCB. In contrast, no alterations in the bristle pattern could be detected in the size-reduced TS5, which suggests that the identity of TS5 is not affected. Similar to the prothoracic legs, in metathoracic legs (third leg pair) the bristle pattern of TS1 is repeated in the distal tarsal segments TS2-TS4 (data not shown). The bristle pattern in the mesothoracic tarsal segments is indistinguishable and therefore not acces-sible to an analysis.

The size of the basitarsus in wild-type and bab mutant legs is much larger than that of the other tarsal segments (Fig. 1A). Although the bristle pattern is changed in TS2-TS4 of bab mutant flies, the size of these segments is not enlarged to the size of TS1, suggesting that the specification of the tarsal segments has changed upon a segment primordia of normal size. The analysis of the phenotype of strong loss-of-function bab alleles demonstrates that the bristle pattern of the distal tarsal segments TS2, TS3 and TS4 is transformed towards the bristle pattern of the basitar-sus, indicating that the bab gene is required for the specifi-cation of the three central tarsal segments.

In addition to its recessive phenotype, strong bab mutations cause a dominant mutant phenotype. Heterozygous males have an ectopic sex comb on TS2, indicating a homeotic transformation of TS2 towards TS1. The same dominant effect is also detected in hemizygous flies that carry a wild-type chromosome over the deficiency Df(3L)bab^PG (Fig. 1D) or over the terminal deletion of the translocation T(3;Y)A114 (61A1; 61F). This indicates that the dominant phenotype is caused by a reduction of the gene dosage of bab, and defines bab as a haploinsufficient gene. In contrast to the recessive phenotype, only the TS2 bristle pattern has changed, and on TS2 itself the ectopic sex comb is the only clear indication of a homeotic transformation. This shows that the dominant phenotype of strong bab mutations is similar to but weaker than the recessive phenotype of strong bab alleles. A com-parison between the recessive and dominant phenotype of strong bab mutations indicates that TS2 is more sensitive to transformation towards TS1 than either TS3 or TS4. This effect can be interpreted as a graded requirement for bab along the proximal-distal axis of the tarsus.

The graded requirement for bab activity is supported by the analysis of weaker bab alleles. Flies homozygous for bab^PR11 or bab^PR30 have small ectopic sex combs (2-3 SCB) on TS2. Therefore, the recessive phenotype of these bab alleles is comparable to the dominant phenotype of strong bab alleles. A stronger transformation of distal tarsal segments was observed with the alleles bab^E1 and bab^PR23. On the prothoracic legs of flies homozygous for these alleles, TS2 exhibits a sex comb with usually 4 SCB as well as transverse bristle rows. On TS3, however, neither a sex comb nor transverse bristle rows are observed. In the metathoracic legs of wild-type flies, TS1 contains transverse bristle rows with more than 7 bristles per row, TS2 contains somewhat irregular transverse rows containing 3 and 4 bristles, whereas TS3 and TS4 lack transverse row bristles. The larger number of 4 and 5 bristles in the transverse rows on TS2 of bab^E1 and bab^PR23 metathoracic legs suggests a transformation of TS2 towards TS1 as described previously for the prothoracic leg. The additional appearance of trans-verse rows on TS3 (2 and 3 bristles per row) indicates that TS3 also has taken on the identity of a more proximal tarsal segment. This observation might suggest that TS3 of metathoracic legs is more sensitive than TS3 of the protho-racic legs to transformation towards TS1. Alternatively, it might suggest that TS3 is transformed towards TS2, a trans-formation that would not be identifiable on the prothoracic legs because of a lack of distinguishing markers between TS2 and TS3. The analysis of bab mutants of different phe-notypic strength indicates a gradual transformation of the bristle pattern of the tarsal segments.

Strong bab alleles produce a complete fusion of tarsal segments TS5 and TS4 in all six legs as homozygotes (Fig. 1B,J). The segmental joint is missing between TS5 and TS4, and the fused double segment is shorter and thicker than that of TS4 and TS5 together in wild-type legs. In flies carrying a strong bab allele over Df(3L)bab^PG, TS4 and TS3, and TS3 and TS2 are partially fused as well (Fig. 1J). In weaker bab mutants, TS5 and TS4 are only partially fused and this occurs with incomplete penetrance.

The analysis of different bab alleles shows that the extent of tarsal fusion and homeotic transformation are related. These two phenotypic traits, however, overlap spatially only in the strong bab mutant phenotype. In weaker bab mutants, the transformation is only seen in the proximal region of the tarsus and the segmental fusion only in the distal region. This separation suggests that these defects are not dependent of one another.

In addition to the segmentation defects that are associated with loss of tarsal joints, bab mutant flies occasionally produce legs with a kink in TS3 (Fig. 1F) or legs where structures of the tarsus that lie distal to the position where the kinks occur are completely missing. The defects might be caused by the cell death, which can be detected in the imaginal primordium of TS3 (unpublished observations). They are observed more frequently in the metathoracic legs than in other legs and usually only in one leg of a fly. They appear rarely (<1%) in the case of the described strong bab mutations, which derived from the bab^P mutation, but occur frequently in flies of the genotype bab^E1/ Df(3R)bab^PG (19% of 130 flies).

Defects in the legs of bab mutants are restricted to a specific subdistal domain of the tarsus (Fig. 1B). Structures proximal to this domain as well as the most distal structure of the leg, the claw organ, are not affected. In addition to the leg, bab mutations affect the morphology of the antenna. The antenna of Drosophila is a structure homologous to the leg and similarly subdivided into different segments. The basal cylinder of the antenna, which consists of two small segments (Fig. 1K), has been shown to be homologous to TS2-TS4 (Postlethwait and Schneiderman, 1971). Strong bab mutations cause defects in the basal cylinder of the antenna suggesting that bab is required in a homologous region of the leg and antenna. Defects at the arista, the dis-talmost structure of the antenna, and at structures lying proximal to the basal cylinder were not observed. The two segments of the basal cylinder are fused to a variable degree to each other and to the arista in strong bab mutants (Fig. 1L). This segmentation defect is accompanied by loss of the segmental joints. Based on available morphological markers, no homeotic transformation of the basal cylinder was detected.

The penetrance and strength of the dominant leg phenotype is temperature sensitive. Flies heterozygous for a strong bab allele develop legs with a stronger homeotic defect at 30°C than at lower temperatures. The temperature sensitivity is not allele specific and can be seen in hemizygous Df(3L)bab^PG/+ flies as well. The strong allele bab^PRDS was used to perform temperature-shift experiments for which the penetrance of ectopic sex combs was chosen as the phenotypic parameter (Fig. 2). Along with its penetrance, the size of an ectopic sex comb changes gradually as summarized in Fig. 2A and B. The temperature-sensitive period of the dominant homeotic bab phenotype centers on the prepupal stage (Fig. 2C) and defines a critical period for tarsal segment specification. Because the temperature sensitivity is not an allele-specific effect, we cannot be certain that it is caused by the BAB protein. However, bab is expressed in the leg imaginal disc at the corresponding stage (see below), which suggests that the phenocritical period is the likely time of bab requirement.

The bab^P enhancer trap line expresses β-galactosidase (β-gal) in the leg and antenna imaginal discs. Flies of the line bab^A128, which have a similar β-gal expression pattern but are phenotypic wild type as homozygotes, were chosen for a detailed analysis of the β-gal expression pattern to avoid any effects that might be caused by haploinsufficiency asso-ciated with the bab^P mutation. Imaginal discs from the early third instar larva until the mid prepupa [6 hours post puparium formation (PP)] were studied. The distribution of the bab transcript was also examined by tissue in situ hybrid-ization to imaginal discs from mid and late third larval instar using bab-specific DNA probes (see description below and Fig. 6A). In addition, a polyclonal antibody (anti-BAB r2), which is directed to a protein that corresponds to part of the defined open reading frame (see Fig. 6A and Materials and methods), was used to study the bab expression pattern in imaginal discs in the third larval instar and throughout the prepupal stage. The patterns of expression of β-gal in the enhancer trap lines and of the bab transcript and the BAB protein were found to be equivalent in leg and antenna imaginal discs (Figs 3 and 5).

From mid third larval instar onward the bab product is present in a concentric domain around the center of the imaginal discs, a region that corresponds to parts of the tarsal primordium of the legs (Fig. 3A,C), and the subdistal structures of the antenna (Fig. 3E,F). No staining is detected in the center of the imaginal discs, which is the primordium of TS5 of the leg and the arista of the antenna (Figs 3, 4). At mid third larval instar, β-gal is weakly expressed in the furrow between the two central folds of the leg imaginal disc (Fig. 4A). The staining level becomes stronger during late third larval instar, when the domain where bab is expressed in the leg imaginal disc gives rise to three additional folds, the primordia of TS2 to TS4. In addition to the region of TS2 to TS4, bab product is also found in the distal margin of TS1 (Fig. 4B). In anti-BAB r2 stainings and RNA in situ hybridizations, we were not able to detect a signal prior to this stage. During evagination in the prepupal stage, when the tarsal segments have expanded, bab is expressed in the region from distal TS1 through TS4 (Figs 3B,D and 5A,B). The exact boundaries of the expression domain are difficult to localize because the staining drops strongly towards the edges of the domain. Differences in the distri-bution of the bab product in the imaginal discs for different leg pairs are not observed. The oldest antibody-stained discs that were examined were from 6 hours old prepupae, when the evagination is completed. To correlate the bab expression domain to the future segments is more difficult in the antenna than in the leg imaginal disc, however, strong β-gal activity is detectable in the two segments of the basal cylinder in the adult antenna of bab^A128 flies (data not shown), suggesting that the two rings of strong anti-β-gal staining in the antenna disc (Fig. 3E) correspond to the primordia of these structures. The analysis of the distribu-tion of the bab transcript and protein in leg and antenna imaginal discs shows that bab expression is restricted to those regions of the antenna and leg discs that are affected by bab mutations.

The analysis of bab expression with the anti-BAB r2 antibody reveals that the BAB protein is localized to the nuclei of cells (Figs 3D and 5B). Although the nucleus appears to be stained as a whole, the antibody detects spots of especially high antigen concentration inside the nucleus (Fig. 5B).

In the bab expression domain that extends from distal TS1 through TS4, the bab product is found in all cells but the amount differs from cell to cell. At mid third instar, the bab expression domain comprises less than 20 cell diameters along the proximal-distal axis and shows a higher level of staining distally than proximally (Fig. 4A). During the late third larval instar, when the tarsal primordium grows by cell division (Madhavan and Schneiderman, 1977) and becomes folded, bab expression becomes stronger and more differ-entiated. The staining level is highest in TS4 and TS3, lower in TS2 and lowest in the distal margin of TS1. Instead of a simple distal-proximal gradient, however, cells within each of the folds show different staining intensity (Fig. 4B).

At puparium formation when the tarsal folds are estab-lished and the evagination process starts, the bab expression pattern appears to be fully developed and remains stable throughout the early prepupal stage. In the expanding tarsal primordium, the bab expression domain comprises about 40 cell diameters along the proximal-distal axis. The complex distribution of the bab product has two characteristics. First, a wave-like pattern is observed in the expression domain. Each tarsal fold in the bab domain shows a bell-shaped dis-tribution of staining intensity. The cells at the ridges of the segmental folds show a much higher staining level than the cells in the furrows (Figs 4C and 5A,B). Secondly, there is a graded distribution of the bab product throughout the expression domain. The highest levels of the bab product are found at the ridge of TS3. TS2 shows lower expression levels than TS3, and the ridge of TS1 has the lowest levels. The staining intensity in TS4 was never found to be higher than in TS3 but either equal or lower (Figs 3B,D, 4C and 5A,B).

In order to facilitate a cell-by-cell analysis of the complex bab expression pattern, sections of anti-β-gal-stained prepupal leg discs of bab^A128 flies were examined with an image processor which translated the staining intensity of each cell into a color value (Fig. 5C). Fig. 5D shows the processed profile of anti-β-gal staining intensity which cor-responds to the imaginal disc shown in Fig. 5A. It illustrates the segmentally repeated high and low levels and the graded distribution of staining intensity. This pattern was repro-duced in all ten leg discs examined by this method.

The P-lacZ insertion of the bab^P allele was used as a tag for the cloning of chromosomal DNA adjacent to the insert. This DNA was then used as a probe to map the bab locus by chromosome in situ hybridization (Fig. 6C) and to screen both a Drosophila genomic library and a number of cDNA libraries (see Materials and methods). Conclusions from these data are shown in Fig. 6A. The sites of the P-lacZ insertions of bab^P and bab^A128 were mapped within the 4.2 kb EcoRI genomic fragment E199 (data not shown). Analysis of the 8 cDNAs isolated indicates that they are incomplete (data not shown). Compilation of sequence data from the cDNAs and the genomic clones allowed us to identify a 1.7 kb open reading frame that is adjacent to the P-lacZ insertion site (Fig. 6A; data not shown). This open reading frame may not represent the entire BAB protein-coding region, and the analysis of additional cDNAs is required. However, we believe that the sequences identified are part of the bab gene for three reasons. First, the P-lacZ insert of the bab^P allele is located immediately 3′ of the identified open reading frame. Second, the distribution of the detected RNA and protein is equivalent to the β-gal expression pattern of the enhancer trap lines (as shown in Fig. 3), and this expression pattern is consistent with the defects observed in the ovaries (Godt, Sahut, Couderc, Laski, unpublished observations) and limbs of bab mutants. Third, the anti-BAB r2 antibody which is directed to a portion of the open reading frame (Fig. 6A) does not detect the BAB protein in the imaginal tissues of flies homozygous for the strong alleles bab^PR24 or bab^PRDS but does detect it in wild-type revertants of the bab^P allele (data not shown). Computer analysis of the 1.7 kb open reading frame iden-tified a strong similarity to two other Drosophila genes, tramtrack (ttk) (Harrison and Travers, 1990; Read and Manley, 1992) and Broad Complex (BR-C) (DiBello et al., 1991). The amino acid sequences of the relevant region of BAB, TTK and BR-C are shown in Fig. 6B. There is a 57% identity over 127 amino acid residues between BAB and TTK, 59% over 112 amino acid residues between BAB and BR-C, and 55% over 113 amino acid residues between TTK and BR-C. Both ttk and BR-C are genes that encode a family of zinc-finger proteins that are thought to be transcription factors ( ttk: Harrison and Travers, 1990; Read et al., 1990 and 1992; Brown et al., 1991; Jiang et al., 1991; Read and Manley, 1992; BR-C: Galceran et al., 1990; DiBello et al., 1991). The highly conserved region (which we have named the BTB domain after BR-C, ttk and bab) is not part of the zinc-finger domain in the TTK and BR-C proteins, and its function is unknown.

Our present study shows that bab is required and expressed in a distinct proximal-distal domain of the limbs; the central region of the tarsus of the leg and the basal cylinder of the antenna. The domain of bab activity in limbs is apparently identical to the domain defined by the phenotype and expression pattern of rotund (Cavener et al., 1986; Kerridge, 1981; Kerridge and Thomas-Cavallin 1988; Agnel et al., 1989, 1992). In addition, this domain is characterized by the gene deadpan, which is expressed in a distal circumferen-tial stripe in each of the segments TS1 to TS4 (Bier et al., 1992). bab and rotund appear to act rather late in limb devel-opment, in contrast to genes that control the whole proximal-distal axis and appear to be required from embryogenesis onward, such as Dll (Sunkel and Whittle, 1987; Cohen and Jürgens, 1998a,b) and wg (Simcox et al., 1989; Baker, 1988a, b; Couso et al., 1993). The subdivision of the tarsal primordium is a late event in the pattern formation of the leg and is also an evolutionary recent step. Primitive insects only have one tarsal segment and the number of tarsal segments differs widely among more advanced insects (Borror and White, 1970).Taken together, this indicates that the bab/rotund domain is a distinct field for pattern formation during leg and antenna development.

Comparison of the bristle pattern in the tarsal segments of wild-type and bab mutant flies suggests that loss-of-function bab mutations cause a homeotic transformation of the three central tarsal segments (TS2-TS4) towards the basitarsus (TS1). This indicates a serial homology of segments in the tarsus. A similar conclusion has also been reached by Stern (1954) based on the finding of an extra sex comb on TS2 in extra sex comb-aristapedia double mutants. It may be assumed, therefore, that TS1-TS4 have the same basic pattern information that is modulated by bab activity in TS2-TS4 in order to give them a specification different from that of TS1. Our data suggest a role for bab as a homeotic gene that is required along the proximal-distal axis of the legs to direct the developmental fate of TS2, TS3, and TS4.

The haploinsufficient transformation effect of bab alleles indicates that the specification of the tarsal segments depends on bab dosage. Also, the phenotypic series of bab alleles can be interpreted as a consequence of sequentially reduced levels of bab expression. The comparison of different bab alleles indicates a graded requirement for bab activity along the proximal-distal axis that becomes apparent in the higher sensitivity of TS2 compared to TS3 and TS4 transformation towards TS1. Given the dosage dependence, a simple explanation for this observation would be a graded distribution of the bab gene product along the proximal-distal axis. This explanation is corroborated by the different levels of bab expression that are observed in TS1-TS4. TS1 contains the lowest level of bab product and is considered the ground state. TS2, which is most sensitive to homeotic transformation to TS1, has a lower level of bab expression than TS3 and TS4. We propose, therefore, that the sensitivity to homeotic transformation towards TS1 is correlated with the concentration of the BAB protein in the different tarsal segments. It is unclear if this correlation holds for TS4, which contains equal or slightly lower levels of bab expression than TS3. In TS4, whose small size seems to allow the production of only one SCB, the usage of this morphological marker might not be sensitive enough to always detect a transformation to TS1.

There are two possible explanations for how bab may act in a concentration-dependent manner for the specification of tarsal segments. bab may promote a binary decision between a TS1 fate and a non-TS1 fate, or bab may act as a morphogen and provide different positional values for the specification of TS2, TS3 and TS4. The latter possibility is supported by the graded distribution of the bab product. In addition, the analysis of metathoracic tarsal segments of weak and intermediate bab mutants shows a change in the bristle pattern of TS3 that can be interpreted as a transfor-mation of TS3 towards TS2 rather than TS1. We suggest, therefore, that bab may provide, depending on its concen-tration, different positional values for the specification of TS2 and TS3. It will be possible to assess the ability of the BAB protein to act as a morphogen by studies of bab over-expression and by examination of segment-specific molecular markers in bab mutants.

bab is the first gene for which a graded expression along the proximal-distal axis is described. The BAB protein dis-tribution in the tarsal primordium reflects the distribution of the bab transcript, which indicates that the pattern of bab expression is regulated at the transcriptional level. The polar coordinate model proposes the formation of distal limb structures under the control of a circular coordinate which integrates anterior-posterior and dorsal-ventral positional values (French et al., 1976; Bryant et al., 1981). Mutations in wg, for example, which encodes a secreted and diffusible protein (for review see Nusse and Varmus, 1992) and is believed to define circumferential positional values, have a drastic effect on the proximal-distal axis (Couso et al., 1993). However, it is not clear how genes like wg could induce a gradient of a nuclear protein along the proximal-distal axis. Whether Dll, which likely acts upstream of bab and is proposed to provide positional information along the proximal-distal axis (Cohen et al., 1989), directly regulates bab expression is not known, but considering that the tarsus is a distinct Dll expression domain in the third larval instar (Cohen, 1993), Dll might be involved in defining the domain and/or the pattern of bab expression.

The analysis of bab mutations indicates that bab has, in addition to its requirement for segment specification, a second function in limb development. bab mutations cause segmentation defects in the central region of the tarsus and in the homologous part of the antenna, the basal cylinder. Because our strongly hypomorphic bab mutations cause more severe segmentation defects as hemizygotes than as homozygotes, we assume that amorphic bab mutations might lead to even stronger defects in segmentation.

At the onset of metamorphosis, the bab product is dis-tributed in a wave-like pattern in the tarsal primordium. Each tarsal fold shows a bell-shaped expression pattern of bab with the maximum level of expression at the ridge and the minimum in the furrow. Considering the segmentation defects in the tarsus of bab mutants, we propose that the wave-like pattern may be involved in the segmentation process of the tarsus. A number of theoretical models have utilized chemical wave patterns to explain how segmenta-tion could occur (for review see Wolpert, 1989). In these models, the waves compose a prepattern that reflect the structures that will develop, and the segment boundaries are proposed to be specified by the troughs or peaks of the wave pattern. Analysis of the molecular mechanisms of segmen-tation in the Drosophila embryo has shown that wave-like prepatterns are not part of the segmentation process (for review see Akam, 1987). Our finding that a wave-like pattern of bab exists in the developing tarsus suggests that segmentation in the tarsus occurs by a mechanism that differs from segmentation in the embryo.

In this context, we have to ask whether the wave-like bab expression pattern can be considered a prepattern for seg-mentation of the tarsus. The wave-like pattern of the bab product at the onset of metamorphosis develops from a rather uniform distribution in the mid third larval instar. The wave-like pattern appears to form in parallel to the devel-opment of the tarsal folds. This, together with the observation that the segments of the tarsus seem to be already defined by the mid third larval instar (Schubiger, 1974), suggests that the wave-like distribution of the bab product is not likely to be a classical prepattern. In contrast to the embryo, however, pattern formation in the tarsal pri-mordium occurs in a proliferating epithelium where a stable molecular prepattern is not expected. It has been suggested that new positional values in the growing imaginal disc are generated by intercalation between values that already exist (Bohn, 1970; Cohen and Jürgens 1989a). The development of the wave-like bab expression pattern may reflect this process. Early folding defects in the tarsal region of bab mutant imaginal discs (D. Godt, unpublished observation) indicate that bab is involved in the morphogenetic folding process. We therefore propose that in the growing field of the tarsal primordium, the morphogenetic process of seg-mentation and the development of the bab gene expression pattern may occur in a mutually dependent manner.

The genetically defined bab locus corresponds to genomic DNA that we have cloned from the region 61F1-2. The β-gal expression pattern of three bab enhancer trap lines and the expression pattern of the transcript and protein from a gene in this region are similar. Further support for the iden-tification of this gene as bab was gained by showing that strong bab mutations do not express the prospective protein at a detectable level.

Part of the open reading frame encodes a domain of about 115 amino acids, named the BTB domain, that shows 57% and 59% identity to the aminoterminal region of the tran-scription factors encoded by BR-C and ttk. BR-C is an ecdysone-inducible ‘early’ gene (Ashburner, 1972; Chao and Guild, 1986; Galceran et al., 1990) and is required for proper morphogenesis during imaginal development in a number of different tissues (Fristrom et al., 1981; Kiss et al., 1988; Restifo and White, 1991, 1992). ttk acts as a repressor of pair-rule genes in the embryo (Read et al., 1992; Xiong and Montell, 1993; Brown and Wu, 1993) and is required during eye development (Xiong and Montell, 1993). The BR-C and TTK proteins contain both a BTB domain and a zinc-finger domain (Harrison and Travers, 1990; DiBello et al., 1991; Read and Manley, 1992). Due to the incomplete nature of our cDNAs, we do not know whether bab also encodes a zinc-finger domain. However, the observations that bab encodes a nuclear protein and that bab mutations cause homeotic transformations in legs suggest that bab functions as a transcriptional regulator.

We are grateful to Eddy DeRobertis and Ulrich Tepaß for encouragement and thank them and other members of the UCLA fly groups for stimulating discussions. We thank Volker Harten-stein, Gerd Jürgens, Ulrich Tepaß, Andrew Tomlinson and Susan Zollman for critical comments on the manuscript. Especially we express our thanks to Volker Hartenstein who has generously made his microscope and image processor available to us. We thank Isabelle Sahut for defining the location of the P-element insertion A128, Silvia Yu for maintaining the fly stocks and technical assis-tance, Todd Laverty for advice on chromosome in situ hybridiza-tions, and Valérie Laget and Clara Paik for help during mutagen-esis experiments. We thank Cahir O’Kane for the two bab enhancer trap lines A128 and A30 and the Bloomington and Bowling Green fly stock centers for providing other fly stocks.

J. L. C. received an International Research Foundation Award from the John E. Fogarty International Center at the NIH. S.E.C. was supported in part by NIH training grant GM 07185, Research Training in Cellular and Molecular Biology. F. A. L. is a Pew Scholar in the Biomedical Sciences and a Basil O’Connor Research Scholar of the March of Dimes Birth Defects Foundation. Funding was also provided by the NIH and the California Institute for Cancer Research.

The DNA sequence that encodes the BTB domain of the bric à brac protein can be found in GenBank under accession no. U01333.